Listen on Spotify, Apple, Amazon, and Podcast Addict | Watch on YouTube.

Today, Investor Sabrina Wu hosts Bobsled Co-founder and CEO Jake Graham for the latest episode of Founded & Funded. Jake had stints at Neo4j and Intel before joining Microsoft, where he worked on Azure Data Exchange. Jake is revolutionizing data sharing across platforms, enabling customers to get to analysis faster directly in the platforms where they work. Madrona co-led Bobsled’s $17 million series A last year, which put the company at an $87 million valuation.

In this episode, Jake provides his perspective on why enabling cross-cloud data sharing is often cumbersome yet so important in the age of AI. He also shares why you can’t PLG the enterprise, how to convince customers to adopt new technologies in a post-zero interest rate environment, and what it takes to land and partner with the hyperscalers.

This transcript was automatically generated and edited for clarity.

Sabrina: Jake, it’s so great to have you with us today. Maybe to kick things off for us, why don’t you start by telling us where the inspiration came from to start Bobsled?

Jake: Absolutely. I’ve always wanted to start a company because I love the competitive aspect of business, the idea of trying to go out and win and bring something into the market. Over the first 15 years of my career, I found that you can get that in pockets in larger organizations, but it’s hard to get that feeling that it’s only the efforts of you and your team that stand between you and success.

I’ve always been on the lookout, but I’ve also always had two tough requirements before deciding to jump in. I wanted to believe in the idea. I wanted to feel, especially in my data infrastructure space and focused on the enterprise, that I understood the problem and that a product in that space deserved the right to exist. It was critical to me to have at least one strong co-founder. For me, especially, it was a CTO I deeply believed in, and I thought we could win together.

As far as how Bobsled itself came about, I sometimes joke that life has three certainties. There are death, taxes, and reorgs. And it was a reorg that made Bobsled exist. I’ve been at Microsoft Azure in a role that I enjoyed. I spent about 18 months building the strategy and plan to create an Azure Data Exchange and an Azure data ecosystem to make data significantly easier to share across Azure consumers. And we finally got a significant budget to start and hire in earnest to start building.

When that happened, a significant reorganization resulted in the birth of Microsoft Fabric. Microsoft Fabric was a fantastic decision. It was and is a product I’m excited about, but it made what I was building not make sense at the time. It’s finally starting to make sense as Fabric has gone GA. But I remember thinking, I took a run, and I realized that I’d finally uncovered a problem specifically in this case; it was around not just data sharing but changing how data is accessed for analytics and evolving data integration into being cloud and not AI-native. I was motivated to solve that problem. I’d finally found something that deserved the right to exist as a successful business.

It would also be totally unfair for me, not to mention that my wife had been pushing me to start something for a while. She realized that — I wouldn’t say that I would be happier founding because this is a really hard job — but that I would be a lot more fulfilled. So, my wife, Juliana, was the secret co-founder who pushed Bobsled to exist.

I spent a couple of days maturing the idea. I started thinking, who do I know I can start to bounce this off of? I called the person who is the best engineer I’ve ever worked with. The gentleman I worked with at Neo4j. We hadn’t spoken in probably a couple of years. I gave him this idea, and he said, “Jake, are you asking me to be your co-founder?” And I said, “I mean, I’m starting to think about this.” And his response was, “I’ll quit tomorrow. Let’s do this. How long do you need? I think we should get going quickly.” I was really taken aback, and I said, “Tomorrow feels a little fast, but we can talk again tomorrow.” At this point, I sat back and realized I was starting a company.

Sabrina: I love that, and I love the co-founding story. It’s definitely a testament to you as a founder to be able to attract great talent and build Bobsled. Can you help us set the stage more and explain what exactly Bobsled does? What traditionally were companies doing before Bobsled to access and share data?

Jake: Bobsled is a data-sharing platform that allows our customers to make their data accessible and available in any cloud or data platform where their customers work without having to manage any of the infrastructure accounts, pipelines, or permissions themselves. Fundamentally, the product is pretty straightforward. You grant Bobsled read access to wherever you store your data, whether in files in an S3 bucket or indirectly within Databricks in BigQuery, Snowflake, et cetera. You reach out to our API to say, “This data needs to be consumed by this individual in this cloud, this data platform, and be updated in this way.” We manage intelligently and efficiently replicating that data to where it will be consumed for white-label infrastructure. Whoever’s consuming that data feels like they’re getting a share from ZoomInfo or CoreLogic, or any of our other customers, but they didn’t have to build any of those pipelines. It allows them to move from putting all the work on their consumer to making it easy without suddenly having to manage an infrastructure footprint in every platform where their customers work.

It’s still crazy to me that the volume of data used for analytics, data science, and machine learning has grown a couple of orders of magnitude over the last decade. The actual mechanisms for that data to be accessed are almost exclusively the same as ten and even 20 years ago. The overwhelming majority of data that’s used to drive any form of data pipeline is either pulled out of an API or an SFTP server. That doesn’t make sense in a world where so much of the value being generated by modern enterprises is in data and in which you need that data to be consumed by others, whether in your organization or others, to extract that value. We needed to see the cloud-native data exchange mechanism take off.

The reality is that the right way to solve this problem is generally by the actual data platform itself. The best way to access data within Snowflake is Snowflake Data Sharing. They have built a fantastic experience where you don’t have to go through and ETL data from another platform. You can query live, well-structured, up-to-date data as long as it’s already in the same cloud region and in Snowflake.

That sharing mechanism was pioneering, and every other major platform followed it. The problem is that it puts a significant infrastructure burden on the actual producer. We want to move away from a world in which every data consumer has to ETL the data out of whatever platform it’s in. The issue you get with that is that sharing protocols aren’t connectors. It’s not just taking the traditional model of we’re going to toss data over the wall. They provide a better consumer access experience because you have to bring the data to where it will be consumed. You have to structure it in a way that is ready for analytics and then grant access to it.

I believe data sharing would be how data is accessed for modern analytics and ML pipelines. But in order to make that happen, the data producers needed a way to interact with all of these systems without having to do it all themselves. That’s fundamentally what Bobsled is.

Sabrina: As you alluded to, data sharing is a critically important part of the stack. I love how easy Bobsled has made it for data providers actually to share the data and how you’re also agnostic, to your point, across the different cloud platforms. If you’re on Snowflake or you’re on Azure, maybe it isn’t easy. But you’re allowing companies to share across different cloud platforms.

You’re flipping this idea of ETL on its head, which is one of the parts I love the most about what Bobsled can do. I’m curious, though, to this point about different cloud providers, what role does cross-cloud data sharing play in this new age of AI and the new way that companies are starting to build?

Jake: Someone recently told me, “I really hope this ML wave sticks this time.” I’ve been working in ML for over a decade, and it gets bigger every year. This just seems kind of like a natural evolution of what we’re doing.

Something you talk about a lot is the idea of intelligent applications. That term applies to an enormous amount of where the market is going. There’s something there’s a really strong definition around. The way I think of them is they’re applications that leverage data in order to better automate whatever workflow they actually solve for, and then also generate more valuable data as a part of that. I think that’s true whether you’re leveraging an LLM, or you’re leveraging more traditional machine learning, or whether this is built on more human-in-the-loop analytics. In order to continue to move toward this more data-driven and now AI-focused age, data has to be able to move between the systems and organizations where it’s generated.

One of the things that people are starting to realize is that often, in any application, a lot of the value it provides is actually in the data generated by running that application. There’s an enormous amount that you can do in that workflow to use that data to improve it and continue to automate it. Another thing we’ve learned about data over the past decade is that it becomes valuable when it’s blended with other data sets. Within data, almost always, the whole is greater than the sum of its parts. When you realize that if you want to think intelligently and predict, and I think even more if you want to do that in an automated fashion using LLMs, you have to be able to bring in the data that represents different aspects of a problem, and that is never sitting in one system.

We saw a push led by Salesforce around the Salesforce Data Cloud. Well, if we can get everyone to bring all of their data into our application, we can solve all the problems. And the answer is no, you can’t. You might be able to answer many questions, but in reality, data is being generated across this enterprise and its partners and other vendors. It needs to flow into the systems where it’s going to generate insight. I fundamentally believe that data sharing will be the mechanism to do that.

How I think this shift enables the move toward the age of AI is we’re going to allow every company to create data products and to have them be accessible wherever they need to work without, again, having to manage an infrastructure footprint and have an army of people who understand how does clustering work differently in Databricks versus Snowflake? How does the permission protocol work differently in an S3 endpoint versus an Azure Blob store? How do I replicate this data? There are better things for people to be working on. Bobsled is going to be a lot of the plumbing for how the world becomes AI-driven.

Sabrina: Data is key to what everyone has referred to for years as a critical part of the modern data stack. You’ve hinted that you think the modern data stack is at a turning point. We’ve talked about this a little bit, but I’d love you to walk me through your thought process. What is this turning point? How do you think that might impact the tech sector and startups overall?

Jake: My general feeling about the modern data stack is it’s no longer a valuable term because it won. There was a time when the modern data stack described a few companies and categories that were bringing analytical infrastructure into becoming cloud-native. That was a relatively exclusive set of companies, so it didn’t include any of the previous vendors who weren’t cloud-natives, whether that was Informatica or even Microsoft, in a lot of ways at that time. And it also didn’t include any part of data that wasn’t pure analytics. There was a separate data science and an L stack, and there’s been a separate operational stack.

I think what has happened is that those incumbents have caught up and are now cloud-native. The companies that weren’t purely in the analytics stack have started to move in there. The more successful companies in the modern data stack are branching out beyond it. The discrete categories within the modern data stack are blending.

There’s been a lot of unnecessary angst around the death of all the companies in the modern data stack. I view it as a victory. The modern data stack is just now a key part of technology. We’re moving from a purely software-centric technology market to a data-centric one. That’s the idea for me of intelligent applications, or if you want to call it the age of AI. The software is incredibly important, but it needs the data. It’s no longer enough to build for a very specific set of users in a very specific category. We now have a much larger field to play in, but also, it’s a much more competitive field.

I think that’s a little bit of what people have been shying away from in the modern data stack. It is like: But wait, this company we thought was going to be a unicorn isn’t going to get there just by solving this slice of the problem.

I go back to the first thing I said: I love the competitive nature of this. I love the fact that every day you can wake up and figure out, okay, how do we execute our way to win, and how do we make sure that we’re solving real problems and that we can bring those solutions to market, and that we can get that feedback loop going? A lot of modern data stack companies are going to be incredibly successful. I don’t think that term is valuable anymore.

Sabrina: Do you think there’s going to be more consolidation of the players within the data stack? Do you think people are going to start to bleed into the different swimming lanes?

Jake: It’s going to happen for good reasons because the most successful company is going to want to keep growing back. Again, this is a competition. You have to win more every day. You’ve earned the right to expand beyond existing categories. I think that’s really good.

I use a joke term: the enterprise strikes back. The vast majority of software spending in the United States is done by the Fortune 1000, and that’s true in the world. The way in which enterprises buy technology is fundamentally different. The idea of fully adopter-led tying together many different solutions is just not working.

Benn Stancil is probably the best writer in our space. I don’t want to parrot his words and take credit for them myself, but he wrote a really great piece on this, with the example of Alteryx. About a lot of startups are looking at Alteryx doing all of these things, and they can only say they do their best in class. It became easy to say, “We’re going to attack this part of it.” Until you get in and realize Alteryx is selling to the enterprise and that complexity, it’s not for nothing. It’s because they built it to meet their customers’ needs. And that breadth is in and of itself a feature.

We’re seeing a little bit of that enterprise strikes back — of the way in which software was built and brought to market for a long time. There’s still a lot of value in it. We need to learn to blend some of the PLG and more pure means of software and product development with some of the realities of your enterprise is going to get you to the promised land. If you can’t add value to the largest companies, you’re putting a pretty low ceiling on yourself.

Sabrina: One phrase that I’ve heard you say before, and we’ve talked many times before, is that “You can’t PLG the enterprise.” Maybe you could talk a little bit about what that means and how you view that statement.

Jake: It means a couple of big things to me. Holistically, as an industry, we’ve lost respect for the enterprise sales part of enterprise software. The pendulum has swung a little bit too far toward the product should sell itself in some ways that’s for really positive and great reasons. It has pushed us to think about product design and user experience. A lot of it has been pushed by individuals within organizations being much more empowered to adopt technology. There are hundreds of millions or billions of dollars in truly product-led growth revenue happening every day. I’m not saying that’s not real, but it doesn’t consider how large enterprises make decisions around technology.

If you think of a few things, often, your buyer and your user are not the same person. Generally, if you’re building something that’s of strategic value and is looking for, you’re not starting small; you need to be attached to a strategic initiative in which there are multiple decision-makers, not just the person who will be using your product. Creating not just a sales motion that allows you to get in front of those people, understand their requirements and goals, navigate their organization, and transfer your excitement about your product to them. That’s a big part of what people have missed: the art, craft, and need for actual enterprise selling.

The second part is that you also need a product development process that feeds into that. Every company, but really every startup, you live and die based on your feedback loop. One interesting thing is that, as an industry, we’ve all internalized the idea that our product ideas are not that good. Focus on a problem, not your solution; ship quickly, get feedback, and iterate. That is awesome in a PLG world where the cost of getting that feedback is incredibly low. It’s challenging in the enterprise space because there is a gap between your buyer and your user. It’s often easier to get time with executives and the users who are going to implement. You’re not getting perfect information there. If you are building something entirely net new, like there is no direct equivalent to Bobsled, even your user will think they’re going to understand how they’re going to use their product. And it’s going to be somewhat wrong once you actually get into production.

The biggest part of it is that the gap between my interest in using this product, I’m evaluating using this product, and actually using it in production is unpredictable and longer. What you end up with is if you don’t build a product development process that views your sales process as a key component of feedback, you end up going back to building in a vacuum. You really need to get to a place where, one, you trust and listen to your sales team or your go-to-market team in general. Two, you’re much more actively asking your customers questions. Third, you’re much more willing to ship even more iteratively and recognize even faster that, a lot of times, what I talk about the team a lot is how we win by fixing the problems our customers identify with our product faster than they think it would even be possible.

If you take the existing agile processes we’ve all built over the past five to 10 years and try to apply them to this more challenging feedback loop, I don’t think you’ll figure out how to interpret the signals. When I say that you can’t PLG the enterprise, it’s not just that you can’t put a trial on your product and hope people come in. It’s not just if you don’t figure out how to navigate the enterprise and get to real decision-makers, budget, and stakeholders. If you build your product in a way that doesn’t take into account how feedback is generated from enterprises in your development cycle, you’re not going to build the right product.

If you’re a product like Bobsled in which larger customers experience our problem more acutely, that’s where we’re starting; you won’t build a product they can adopt. I think you’ll also have the other thing that many companies, especially from the modern data stack age, is we know we solve a real problem, but the product and the go-to-market don’t seem to quite fit where the actual money is. How do we get over that? I really think the answer is to fall in love with enterprise sales again. Really care about it.

Sabrina: Many companies that we talk to are struggling with this because PLG has become so favorable and fashionable, to say the least. But in reality, it sometimes comes back to basics. That’s what you’re alluding to here: How do you get that feedback loop going? How do you listen to your customer, especially when there’s a disconnect between the user and the buyer?

One critical component for Bobsled is being friendly with all the hyperscalers, but you partner with many of them, including Snowflake, Databricks, GCP, Azure, et cetera. How have you gone about managing these partnerships? Especially as an early-stage company, I think it can be very challenging. These companies are very large. Do you have any advice for entrepreneurs who might be navigating some of these relationships?

Jake: I’ve always been inspired by Apple’s 1984 commercial where they made clear they had an enemy, in which case was IBM, to try to motivate the team. I assumed it would be one or a few of these platforms because they were building walled gardens, and we were building a product to bring those walls down and connect these different platforms. I was shocked when none of those platforms oriented Bobsled that way. They all oriented to Bobsled for, “Oh, wow, this solves a real problem for our customers. It’s one that we don’t want to solve ourselves.” I go back to our specific problem, it involves managing infrastructure across all these different platforms, and that’s a hard thing for them to do themselves, although there are plenty of examples of them doing it.

We were pleasantly surprised. There was a warm reception from executives early on to partner. I was fortunate to be on the other side of partnership motions both at Microsoft and at Neo4j, from both the startup and the hyperscaler itself. I think they’re super dangerous for early-stage startups and that they can suck an enormous amount of your time, energy, and mental thought into them in ways that will eventually pay off, but not in the timeframe that you need them.

My advice for the early stage would be to focus on partnering with individuals at hyperscalers. So all of these companies have effective machines that move billions of dollars in revenue for partners, and almost none of those billions of dollars in revenue come from early-stage startups. You need to get to a place where you’re at scale, and then they can help you scale more effectively. At that point, you need to be able to be relevant to the individual AEs across every platform with an incredibly repeatable sales process, pricing motion, and mature integration. I look forward to that day, but it’s not series A.

The way that we’ve approached it, which has been effective, is we’re now starting to clip into the machine a bit more and start training sales teams at some of our partners like Snowflake and Databricks. The first two years have been finding executive champions willing to help us navigate and take calls with prospects and propose Bobsled, who are willing to individually send us leads. I can think of a few of our first enterprise customers where we’d be on, I was sharing a cell phone, actually my head of sales was sharing a cell phone with an executive from, in this case, Databricks. Or we were having joint meetings with executives at Snowflake and having them really reinforce the story, not just that the problem we’re solving was real, but that we are the company to help solve it.

You’re not winning by convincing 5,000 companies at once or by getting 50,000 Microsoft sellers to sell your product. You’re winning one by one. It’s like a day by day, you go out and convince these customers that your product is worth betting on. Focus on those individuals, get those wins, and that will earn you the right to scale.

Sabrina: You thought about all these different partnerships. Were there any ways that you’re prioritizing them? Or thoughts on, obviously you said it can take up a lot of time, and so if you’re an early-stage startup, how do you think about who might be the most valuable partner to you? Or how do you stack rank as you’re thinking about building? That’s a critically important part of the process for founders, so I’m curious how you thought about it.

Jake: I mean, part of that goes back to the last piece of advice that you’re partnering with individuals, not with organizations at that point. So, where you find individuals who really want to lean in and have influence in their organization, and really the specific part of their organization you want, you should be a lot more opportunistic, I think, than anything else.

For us personally, what we found, and despite my coming from Azure and having a lot of close contacts at AWS and GCP where we are partnered, was that the breadth of their portfolios made it much harder. So for us, Snowflake and Databricks, because everything that we do helps them achieve their goals. We’re in one of the few positions where it’s by working with Bobsled, a customer does two things immediately. It enables them to share their data on every one of these platforms, which directly drives consumption for every one of those platforms. So, getting data shared in is a good thing.

The other thing it does is it allows you to centralize more and more of your data in the platform of your choice because you are now able to access these other platforms, and you don’t need to worry about lock-in. So we’re in a bit of a unique position with most of these platforms where they all win when we win. It’s one of the few times where we could go to an enterprise and say, “Hey, Databricks would like to talk to you with us because they’re going to win if we win. And Snowflake would like you to talk to us because they’re going to win if we win.”

Anything that you can do to focus on where you are driving value they care about is similar to enterprise sales in the same way. If you’re trying to sell something, you must attach it to a strategic initiative that people care about. You’re not going to close your first half-million-dollar deal on a science project. Over time, it needs to tie to something. It’s the same thing with a partner. Find an individual that you attach to those strategic priorities. As much as you can, find the actual part of the business or possibly the entire company where you’re helping to drive their strategic priorities. You’ll find that they pay a lot more early dividends than trying to figure out if I could get in front of GCPs 30,000 sellers; I know that I wouldn’t have to hire my own sales team.

Sabrina: That’s great advice and I’m sure founders listening today will find that to be very helpful.

As we wrap up here, Jake, just a couple of questions left for you. One is I’m curious, what’s an unexpected challenge? Maybe one of those oh my gosh moments where something just didn’t work out the way that you thought it would, and how did you deal with it?

Jake: I think the one that’s coming to mind right now is that I believe in executive coaching, so everybody at Bobsled has a coach, and I have two. One of the challenges of being a CEO is figuring out the right timeframe for you to focus your energy on. You are responsible for the strategic vision. Especially if you’re building a VC-backed business, you can’t just be focused on, well, if we just win these customers. It needs to be a part of a larger master plan. Moments where I drove the least clarity were when my mind was focused on what it would look like for Bobsled to win two years out or even a year out and not focused on what exactly we needed to do to win right now.

I have an amazing leadership team that is often much more experienced than I am in the roles they bring. I can’t just say, “You figure out now, and you go figure out where we go next.” That’s not how this works. Make sure you’re actually defining what winning looks like for your team and talking constantly about how to win right now.

Make sure you’re still separating the space for yourself to evaluate: Are we going in the right direction? Are we building ourselves into a corner? Is there enough total addressable market here? When is the right time for me to start thinking about expanding our TAM? When is the right time to think about bringing on that next leader? But don’t get caught up in constantly building and thinking about the future. Make sure you’re focused on your actual wind condition today.

Sabrina: I love all that advice. It’s critically important to stay focused on what’s happening in the moment but also have that broader vision, knowing that the 10-year plan is potentially to get some other big wins, maybe that IPO down the road, or whatever else you’re looking forward to.

Jake: You’ve got to be able to convince yourself that there’s a path for you to get there. Otherwise, don’t take this path.

When we had a customer win, I must’ve had at least two minutes of real excitement around that before I thought, okay, we need 100 and some odd more of those to IPO. Like you just got to, you can’t celebrate the touchdowns. That’s the reality of VC. Madrona didn’t invest in Bobsled because they thought we could get to the Series B, you’re investing in Bobsled because you think we can go significantly further. It makes this decision point harder. Well, it’s like, well, I’m building for IPO, and I need to validate that. If we don’t do what we need to do to get to series B, it’s all for nothing. How do you constantly live in that time shift? It’s a mental challenge I hadn’t thought about, and I now spend a lot of time thinking about it.

Sabrina:

Well, it has been a pleasure and honor working with you, Jake, and the rest of the Bobsled team. We’re excited about what you guys are building and know that the best is yet to come. So thanks for allowing us to be part of the journey. And thanks for joining us today. Really appreciate it.

Listen on Spotify, Apple, Amazon, and Podcast Addict | Watch on YouTube.

Today, Managing Director Tim Porter chatted with Craig McLuckie, who’s most known as one of the creators of Kubernetes at Google. Madrona recently backed Craig’s company Stacklok, which is actually the second company of Craig’s Madrona backed (Heptio in 2016).

Stacklok is a software supply chain security company that helps developers and open-source communities build more secure software and keep that software safe. Tim and Craig discuss Stacklok’s developer-centric approach to security, the role of open source in startups, the importance of open source and security, and they also hit on some important lessons in company building — like camels vs. unicorns – and where Craig sees opportunities for founders. It’s a must-listen for entrepreneurs out there. But, before I hand it over to Tim to take it away, don’t forget to subscribe wherever you’re listening.

This transcript was automatically generated and edited for clarity.

Tim: I’m very excited to be here today with Craig McLuckie. It’s the second time I’ve had the privilege of working with Craig on a startup. So, Craig, It’s awesome to be with you here today, my friend. How are you?

Craig McLuckie: Oh, I’m doing great. Thanks for having me on.

Tim: Absolutely, It’s our pleasure. Well, I didn’t do it justice. Tell us a little bit about Stacklok and what you’re building now and a bit of the founding story, and then we’ll double back and talk more about how some of your experiences in Kubernetes and Heptio, et cetera, led you to build this company now.

Stacklok & Securing Open Source

Craig McLuckie: The story behind Stacklok is it’s a little company, a series A company, Madrona backed, that was started by myself and my friend Luke Hinds, who was the founder of a project called SigStore. The story behind Stacklok goes back several years. I’ve been known for my work in the cloud native computing space, and I had some success with open-source efforts like Kubernetes and many other projects that we built on the back end of Kubernetes to operationalize it and make it more accessible to enterprises. Open source has served me incredibly well as a professional, and I’ve spent a lot of time in open source building open-source communities and navigating those landscapes.

One of the things that occurred to me is that it seems obvious, but open-source is important. It is the foundational substrate for what is driving a substantial portion of human innovation right now. We spend a lot of time talking about generative AI, and you look at something like ChatGPT, and we dig into what’s behind the scenes, and there’s Linux, there’s Kubernetes, there’s Python, there’s PyTorch, there’s TensorFlow. All of those were precursor technologies before many ChatGPT-specific IPs even lit up. That’s a significant stack, and it’s all open-source technology.

The question that went through my mind and continues to echo with me is, we’re so contingent on open-source, but how do we reason about whether open-source is safe to consume, as a fundamental building for almost everything we consume? Historically, the currency of security has been the CVE, so there’s a critical vulnerability associated with a piece of technology. However, it’s increasingly been challenging for organizations to deal with this. My interest in this space predates things like the SolarWinds incident, which got people to think about supply chain security. This only predates the Log4J incident, which continues to plague enterprise organizations. It comes down to this: we are contingent on these technologies, but we don’t necessarily have a strong sense of the security or sustainability of the organizations producing them. We’re not necessarily creating the right amount of recognition for organizations going above and beyond to produce more open-source security.

What we’re doing with Stacklok is two things. One is we’re building a set of intelligence that helps organizations reason about open-source technology, not just in terms of whether it looks safe, whether it has a low number of vulnerabilities or static analysis being run against it, et cetera, but also increasingly whether the community that’s producing it is operating in a sustainable way or not. So, producing that signal is a good starting point. Still, we also want to make sure that on the consumption side, when organizations are looking to bring that technology in-house and continue to invest in it and build around it, we have the apparatus to drive sustainability to create control loops that enable folks that are both producing open-source and consuming open-source to institute policies to make sure they stay within the lines of safe consumption. It’s about bringing those two things together in a set of technologies backed by open-source sensibilities to help organizations navigate this incredibly critical landscape.

Tim: Stacklok is built on open-source, or at least around Sigstore for open-source, as companies and developers ingest lots of open-source, put that together to build their product, and then have to keep that updated over time. So those open-source projects that fed it continue to be updated themselves. So you get to track that whole chain. Talk a little bit more about how that works and how Sigstore plays into the strategy here.

Knowing where open-source tech comes from

Craig McLuckie: One of the most important things to know when consuming an open-source piece of technology is where it came from. It’s less obvious than people might think. If you go to a package repository and download a package, and then you install the package and run the package, you’ll look at the package repository’s metadata for where that package was published. So hey, it was built in this repo by this organization, et cetera. But how do you know that that’s the case? Because most of that information is self-reported, it’s not being generated. The answer is that folks aren’t even publishing the signatures associated with the public keys associated with signatures.

You need something that can deterministically tie a piece of software back to its origins. This is what Sigstore has done. It’s effectively a bridge between an artifact and the community that produced it, and it’s a set of tools that make it easy for a community that’s producing something to sign that thing and say, hey, we produced it, here, and this is what it contains.

Now, Sigstore is one little piece of the story. It ties the software to the origins but doesn’t tell the origin story. It does not necessarily give you insight into the community behaving responsibly or show what the transitive dependency set will look like. We’ve created this technology called Minder, which helps communities or consumers of software or producing software operate in more intrinsically sustainable ways. Let’s say, hey, I want all of these policies applied to all of the repos associated with producing this piece of technology and make sure that those repos stay in conformance.

When a new merge request comes along with something sketchy, let’s block that merge request and recommend an alternative instead. If someone misconfigures one of your repos so that branch protections are off, let’s catch that in the moment and make sure that that can be remedied. In so doing, you’re producing valuable information about how that community produced that piece of software, which then feeds the machine so that all that information then becomes a context that can get written into a transaction log like Sigstore so that the next person who is coming along to consume that piece of software now has that intelligence. It can make informed decisions about the consumption of that software. It’s turtles all the way down because when you look at a piece of software, it’s composed of other pieces of software that are composed of other pieces of software. Enabling these organizations that are in the process of composing things together and enhancing them to document their practices and write them into a record that someone can then consume subsequently is an incredibly powerful metaphor.

Tim: You mentioned supply chain security and talked about some of the major exploits that have occurred, like SolarWinds. For the audience, put Stacklok in the context of the broader supply chain security market. If it’s a market, it’s an umbrella term that gets used by a lot of different companies. It might be a little confusing for a lot of people and confusing for customers, too. Can you help frame this?

Craig McLuckie: First, Let’s look at the security landscape and consider how the world has worked. Historically, it used to be a world where hackers were effectively like burglars sneaking around your neighborhood, and they would look for an unlocked window and then open the window and sneak into your house and steal your silverware. That was the world that existed.

With the SolarWinds incident, we’ve seen that it’s not enough to lock your windows and turn on the alarm system with some of these active security practices. The burglars are now getting jobs at window manufacturing companies or are breaking into window manufacturing companies and tampering with the apparatus that produced the latches on windows so that the windows that are being produced and installed are intrinsically insecure, so when everyone’s out at a town banquet, they can go and clean out the entire neighborhood. That’s the sea change that we’re seeing in the security landscape.

It’s not enough to look at a piece of software today and say, hey, this thing has no known vulnerabilities, it’s good. No known vulnerabilities don’t necessarily mean anything. It could be that it’s good, or it could also mean that it’s perfectly malicious, meaning it was produced by someone sufficiently sophisticated to make it look exactly like what you want to consume. Still, they added a little naughty something that will ruin your day when it gets deployed into your production system. This idea is about understanding and knowing where your software is from and its origins, like understanding the farm-to-table journey around your software.

In terms of our positioning, where does this start? It starts with your developers. It’s not enough to say we will insert this into the production environment because, by the time you put something as a control in a production environment, your organization is incredibly invested. Failing a quality scan in a production environment is the first time you discover a Log4J instance in something you’re trying to deploy. That’s very painful because you must return to the development team. You have to figure out who produced that thing and go back to them and say, hey, this is not okay. They then have to go and reproduce it, revalidate it, redeploy it, and get it to you, and it takes an inordinately long time to deal with.

You want to intercept developers not just at the time when they’re submitting a merge request but when they type that import statement and give them signals saying, yeah, what you’re doing is good, or, what you’re doing is probably going to get you hurt later down the pipeline. Next, instituting further controls, starting with the Git repository, moving into the pipelines, the container registry, and the production environment. You have these controls along the entire journey, and you can look back on the farm-to-table story of a piece of software you’re looking to deploy.

The Role of Developers in Security

Tim: You mentioned being focused on the developer. Stacklok’s a very developer-first, developer-centric company and product you’re building. A lot of times, you think of security software as something that’s more top-down; it’s a necessary evil, and it’s being imposed on you from above within management. Talk more about why it’s important to start developer-first and how it can become something equal that a developer wants to embrace and ultimately create a better experience for their users.

Craig McLuckie: One of the things that is true about developers is they generally want to do the right thing. It’s not like developers are sitting there saying, you know what? I want to produce insecure code, or you know what? I want to mess with my ops team’s day and produce something insecure. They don’t want to go back and forth with the operations team, but it’s also worth recognizing that, at the end of the day, they’re primarily being metric-ed on producing software that creates business outcomes. The thing that’s going through their head is that I need to create customer delight through my code, and in creating customer delight, I’m going to create money for the company I’m working for. If you’re starting to produce capabilities that inhibit that, they may still want to do the right thing, but they will find ways to work around you.

The later you leave the discovery of something in the left-to-right flow of a developer’s workflow, the more intrinsically painful it’s going to be, and frankly, the less likely people are going to be to accept what you’re doing. That’s from the pragmatic side of getting this technology adopted. This is a constant challenge for the CISO, which is, hey, I want to introduce this, but the minute I do, my business counterparts yell at me because I’m slowing down their production world. So starting with technology that’s developer-friendly and developer-engaging is a good story.

Now, it’s important not to confuse developers as buying centers. Developers don’t have budgets by and large, but increasingly, when you look at the IT landscape, developers are just a disproportionately significant cohort, not because they’re necessarily going to buy your technology directly, but because all of the people that do buy your technology care about them, and want to enable them, and are going to see you as being a more intrinsically valuable capability if you appeal to the developer.

To Open Source or Not

Tim: A lot of the listeners are thinking about, hey, I’m building a new company. Should I open-source? Should I not? How do I capture developers’ attention, hearts, minds, and community? I’ll back up just a little bit. So Craig, you co-created Kubernetes at Google and left a while later and started Heptio, which was all around helping enterprises adopt Kubernetes. Kubernetes became arguably the most popular open-source project in the world. There was a point where only Linux had more daily interactions than Kubernetes. Of course, then at Heptio, you built it for two years, sold it for 600 million to VMware, and it continues to live on there. I know there’s a lot here and a lot of founders come and talk to you about it, but what are some of the principles that you thought about in building out Kubernetes and now are thinking about Stacklok or other companies that are focused on developers and using open-source to build this adoption flywheel and community?

Craig McLuckie: There’s a hard truth here, and this is important for the listeners to embrace, which is that open-source is effectively mortgaging your business future for lower immediate engagement costs. So, you get lower activation energy. It’s much easier to get the flywheel turning with open-source particularly if you’re a small company. it’s going to create a virtuous cycle with a lot of the individuals that you want to engage. They may even contribute directly to the project. They’ll certainly make it their own. They’ll give you feedback in near real time; they’ll build the project with you. It’s a wonderful way to build technology. It reduces your barriers to entry in enterprise engagements, particularly if you’re a rinky-dink little startup. If you have an Apache-licensed open-source project, particularly if the IP is Foundation Homes, you’re far more likely to get through procurement because, at the end of the day, they know they can just run it themselves if things get weird. So your path is easier.

When I was thinking about Kubernetes, what distinguished us at Google was that I didn’t need to commercialize Kubernetes. I just needed it to exist because I had something to commercialize it. It was called Google Compute Engines, which had decent margins. I needed to disrupt my competition with the technology to run better myself on Google’s infrastructure, which motivated us to drive the Kubernetes outcome. With the success of that project, Heptio was enabling enterprise, this idea of enabling enterprise organizations to achieve self-sufficiency, to bridge the gap between them and the community to fill some of those operating gaps associated with the consumption of the open-source technology, and that worked out well.

When I look at what I’m doing with Stacklok, I recognize that over time, if I’m successful with the Minder technology, I will have to accept a lower value extraction narrative than if it were proprietary technology. But realistically, the probability of me succeeding and getting something that’s consumable out there is far higher if I can embrace the community. You have to have a plan. What is your plan for commercialization?

In the case of Stacklok, I’ll be very open with our customers and the community. Our plan is to create incredibly high-value contextual data, which we’re manifesting as a trustee right now, to support the policy enforcement that Minder has. That represents a differentiated thing. It’s hopefully something that our customers will value over time as we bring more and more richness to that data set. It represents something that’s differentiated from the community open-source technology. That’s my broad plan. I’m very open; I wear my heart on my sleeve. I don’t plan to change the licensing model. I have to stand by my commitments to my customers and to the community that I stand by. But the point is, I do have a plan for commercialization. It’s not just, I’m going to be the RedHat of this thing because it turns out RedHat is a pretty singular creature.

There is another story here, and we see a lot of this, which is that people are happy to pay you to operationalize something. If you have built a system that’s reasonably complex and you’re able to operationalize it better than other people. You’re able to set the boundaries of where commercial starts and ends and open-source starts and ends; you can navigate building a SaaS business around a single vendor, open-source technology. We’ve seen great companies emerge in that sort of context.

Tim: You built Heptio over a couple of years and it was bought, great journey, faster than probably anyone anticipated when you started. What were some of the things you learned? It could have been about some of these open-source threads we were just talking about, or there’s a whole host of other just general building startups successfully. What things did you take from that experience that you’re making sure to bake into Stacklok? Were there any things that you didn’t want to repeat having done it once before?

The Importance of Culture in a Startup (Camel v. Unicorn)

Craig McLuckie: I enjoyed the Heptio journey, and it’s hard to complain about the outcome. It’s also worth recognizing that we were riding on the momentum of Kubernetes. It was a very luckily timed undertaking. I don’t claim to be able to create that kind of luck for myself again. We need to approach this from a bottoms-up perspective.

What is different with Stacklok is that I definitely had a bit more of the unicorn versus camel mindset. A common narrative around my leadership circles is no, that’s a unicorn thing; that’s not a camel thing. We’re camels, not unicorns. We’re building something that is incredibly lean and purposeful, that’s going across the desert to an oasis, and if we don’t make us die, that’s how we think. We know we have to get this far on this much money, and there are no alternatives.

Opportunistic hiring that’s a unicorn thing; that’s not a camel thing. Getting crazy with swag, that’s a unicorn thing, not a camel thing. I think one essential delta is just a reality of the environment that we’re in. It’s like you have to have a plan for commercials. Like the days of being able to raise on promises and a winning smile are over, you better have a business plan. We’re more thoughtful in terms of our approach. We’re much smaller than I was with Heptio at this point. We’re twenty-one engineers, but they are crazy efficient and really hardworking. I’m very proud of our two products in nine months. So that’s the one difference.

The second difference is that I’ve approached this with what I think of as a hardwood growth model. Heptio was like a linear growth model, resulting in a lot of fog of war. We struggled with folks feeling disconnected because we were growing so quickly. What we’ve instituted with Stacklok is a grow hard and grow hard and grow hard, like a growth execute and run the team really hard, produce an outcome, then grow, run the team really hard, produce a result, then grow. I haven’t seen that referenced or talked about much. Still, I’m finding it to be incredibly beneficial just to the culture and to the organization because you establish the team, by establishing the norms. Then you use those norms to direct your next wave of hiring.

The other important thing is that in Heptio, we were a very culture-forward company. I think that I’m retaining that culture-forward perspective with Stacklok, but I’m now very purposeful in this case that the culture isn’t the mission, the mission’s the mission, and the culture supports the mission, and making sure that as we embrace our hiring practices, we’re very, very diligent about not just hiring the kind of people we want to work with, but hiring people that have that sense of purpose, that sense of mission, that willingness to engage in the way that we need to. That camel mindset is very powerful.

Tim: I love the camel versus unicorn mindset. It’s absolutely essential in today’s market. The days of the unicorn are distant memories in many respects. You’re talking about culture and how you’re really intentionally and thoughtfully building it at Stacklok. You wrote this great blog early on in the company called The Operating System of the Company. I like your point about how it supports the mission but is different from the mission. You have these great company tenants that you sort of alluded to. Maybe just say more about how you’re building culture and how you thought about those company tenants. This is the most common thing between all startups and all founders, is that you’re building this culture, building it right for authentically you and your team, how that carries forward for the long run in the company’s just absolutely essential foundation.

Three Jobs of a Founder

Craig McLuckie: Yeah, I heard a founder, say this, and I’ve adopted it as my own story.,which is I have three jobs, right? I have to effectively set the strategy for the company, I have to execute the strategy for the company, and I have to define the culture of the company. And of those, the third is probably the least understood, and most important. So culture for us is it’s our operating system. It’s not just the warm and fluffy things you write on the whiteboard to make you feel good about yourself. It defines how we hire and how we make decisions, and It’s indistinguishable from things like our brand. Our culture is our brand. Our brand is our culture, it defines us in ways.

In my previous company, I built with Joe Beda, and Joe and I are old friends. We’d worked together for ten years and done impossible things together. You would not find any air between us. You could present us with the same data, and we’d always formulate the same conclusion, which was easy. So when Joe and I started Heptio, we first sat down to write the culture and then went off.

Now, with Luke, an amazing human being whom I’ve come to respect in the most fundamental ways possible, we were relatively new. We had yet to work together. We had spent many hours together and gone hiking but had yet to work together. So, we both did this exercise and wrote down what we believed, just drawing on our lived experience. These are the things that define us and how we make decisions. Those little statements Luke says, like you always have short toes. You can’t step on my toes because I have toes. You can say anything, and I’ll just take it at face value. I’m not going ever to take defense. So that’s an example of something he came up with.

We both wrote those down and then I did a day exercise where I built an affinity map and went through all of the tenants, everything that Luke and I felt. I tried to distill it into a set of five things that I could then articulate that would represent what I believed to represent the way we operate because you can’t fake culture. The minute you try to fake culture, at some point, a hard decision’s going to come up, and you’re going to make a decision that’s not aligned with your culture, and then hypocrisy is going to creep and your culture’s dead and you have to start from the beginning.

I wrote down the five tenets. One: We’re a team. No one gets lifted by pushing someone down. The organization is invested in community-centricity. Community is a symmetric advantage. Two: We’re mission-oriented, we’re a startup, and we’re a camel-based startup. There will be some long, dry patches, and you better have the world to make it to that oasis. The only thing that will get you there is if you believe in the mission. I want people that feel that burn, that really want to engage in this mission with us. Three: We’re a startup, so we have to be data-centric and find truth in data. We cannot gloss over a hard truth staring us in the face because we refuse to engage with and believe the data. Four: We have to be humble but determined. The camel is a humble but very determined creature. Some of the best leaders I’ve ever worked with have embodied this. Five: We’re a security company, so we have to stand vigilant. Bringing that all together and then operationalizing it. When discussing hiring candidates, guess what exists in our candidate management system? It’s these five cultural virtues. Our interviewers should provide feedback on those five things. We talk about how to assess this in people that we’re talking to. When we’re talking about promoting individuals, we reference back to the cultural elements. Whenever you’re having a complicated conversation, and you have to make a decision, you point back to these tenets. So these are the things that are informing our decisions. That’s how I’ve approached building a culture.

Tim: Fantastic articulation. It’s energizing to hear you talk about it, Craig. Of course, you referenced your fabulous co-founder, Luke Hinds, and he was a long-time senior engineer at RedHat, co-creator of the Sigstore project that Stacklok is drawing on significantly, and is also based in the UK. So you also have a bit of a geographic/ time zone that you’re working through, which in this world where companies are at least to some degree hybrid, this cultural blueprint and operationalizing it, I think is that much more critical, compared to when everyone can just be together in the same office.

Craig McLuckie: In this kind of remote first world that we find ourselves living in, canonizing that and expressing it and being very deliberate in your communications is so critical as a startup founder.

The Future of Stacklok and the Cloud-Native Ecosystem

Tim: Awesome. Let’s come back one more time to Stacklok. So you just formed a company and founded it earlier last year. An incredible pace of building this great team of predominantly engineers that you referenced, launched the first two products which you referenced Trusty and Minder. What’s up for 2024? Maybe without giving away anything confidential, what are you looking forward to from a company’s standpoint this year? If there’s a call for customers, who should be interested, what type of developer or what type of enterprise, the developers there should be looking to Stacklok for help in the areas that we were talking about earlier?

Craig McLuckie: There are a couple of things that we’re looking to bring to market. One is, there’s Minder, which is a system that helps you mind your software development practice. This year, we want to help organizations embrace and engage with Minder. We want to make it free to use from an open-source community perspective. We want to support and engage open-source communities that are looking to improve their own security posture by providing this technology. They should feel like it’s, hey, Stacklok’s bringing us something that’s really useful. So if you are an open-source project and you have, say, 500 repos, and you’re worried about licensing taint showing up for some purpose. Having something that you can run reconciliation loops across 500 repos and reason about the trans and dependency set and make sure that nothing’s showing up in your world that shouldn’t show up, is useful. We want to engage with communities to help support their use of this technology.

We think of this as a way to give to these communities to enable them to start operating securely but also to be able to show their work so that they can produce Sigstore-based signatures and generate the attestation signal that the people who are looking to use their projects are starting to expect. We’ve focused almost exclusively on GitHub integrations, but through the year, we’ll add other critical integration points, such as Bitbucket, GitLab, Kubernetes, and pipeline integration.

On the other hand, we wanted to start the flywheel going with some high-value information. So, while Sigstore’s gathering ahead of steam and these communities are taking time to start producing consumable attestation information, we wanted to precede the segment with some very high-value intelligence. We started by doing some statistical analysis using Trusty, which is data science against open-source packages, looking for signals that tend to indicate both vulnerabilities and health. You can expect us to continue to enrich that. I’m not ready to announce, but keep watching the space. We’ll start to introduce some very, very sophisticated and cool ways of thinking about open-source technology that complements a lot of what’s already out there in the ecosystem. We’ll make that all well integrated into the Minder capabilities so that you can start to define and enforce policies based on those signals.

Tim: Awesome, looking forward to this year and beyond. A lot of people wonder your point of view on Kubernetes and the Cloud Native ecosystem. You’re firmly focused on building the security company now. Of course, lots of interrelated work with development that takes place around Kubernetes, but what’s your view on the state of that community, Craig? Is it at a maturity phase? All the big companies have their hosted and managed services. Do you still see room for more innovation for startups broadly across the Cloud Native ecosystem? Are there any pain points that you’re continuing to hear about? Any advice for founders who are looking to continue to build in your old world?

Opportunities for Founders

Craig McLuckie: There’s tremendous opportunity in that space. Call me crazy. I don’t want to be that ’80s band with that one hit song, like singing at corporate events for money in my 70s. I want to branch out and try some new things, and the supply chain stuff has been something I’m very passionate about.

Honestly, I think one of the things I’ve observed is bringing platform sensibilities to the security space, it introduces a very novel way of thinking. Look at the security ecosystem, why isn’t everything just running reconciliation because they work so damn well for platform sensibilities. It’s like, why is this not just a Kubernetes-esque integration platform? Why doesn’t this exist? We should just build it.

I think there’s a lot of work to be done. I mean, obviously, generative AI is hot, and from firsthand experience of building and running large language models that we use behind the scenes to produce some of the value that Trusty offers, there’s a lot of fragility and brittleness there. I think there’s an almost unfettered demand for capabilities to simply operationalize and deliver as a service-like experience for some of these community-centric large language models. I think in the Gold Rush of generative AI, the operationalization pickaxes are going to sell very, very well. I think that’s something that I would certainly be interested to see where it goes. I would certainly be looking to consume that myself because right now, we just find that operationalizing those models is brittle and can be a bit of a challenge.

I think there’s still unfinished business in the gap between the platform as a service and the Kubernetes layer. So, the gap between Heroku’s and Google App engines and the pivotal cloud foundries and the world of Kubernetes still exists. We haven’t seen a lot of really great turnkey experiences. I think the work that we started doing with the Tanzu portfolio was a nod in the right direction, but I definitely think there’s a wonderful opportunity to continue to explore and play with the idea of the service abstraction and basically looking at service dependency injections for modern workloads.

I’m also acutely interested in the way that WebAssembly is going to start to shape up and represent ways to bring new atoms of computational resources into new form factors using and borrowing a lot of the distributed system sensibilities that Kubernetes created. I think there’s tons of opportunity. If I hadn’t met Luke Hinds and fallen in love with supply chain security, I can think of three or four great startups that I’d be happy to do tomorrow, but I’m very happy with the one that we’re doing.

Tim: We’re as well. Thanks so much, Craig, great insights. We could talk about these things for days, but I really want to thank you for spending time and sharing these insights. It’s just a great pleasure and fun to be able to work together here and do a little bit to help you and Luke and the team as you build Stacklok. So, thank you.

Craig McLuckie: Thanks, Tim, really appreciate it.

 

In this episode of Founded & Funded, Madrona Managing Director S. Somasegar and General Counsel Joanna Black discuss the fundamental role of a board in a startup’s growth and development with Madrona Digital Editor Coral Garnick Ducken. They touch upon the importance of aligning strategic views with board members, managing disagreements, and effective board governance to ensure that the organization is run efficiently. The duo offers numerous insights into the intricacies of board structure at different stages of a startup lifecycle, drawing parallels from recent events at OpenAI. The conversation covers the need for transparency, both in sharing good and bad news, and the necessity for a functional board reflecting a functional culture.

This transcript was automatically generated and edited for clarity.

Coral: What we’re trying to do here is set the stage before we dive into all the advice and startup board governance and structure and everything that we’re going to talk about here. So, to kick it off for us, Joanna, can you give us the quick and dirty with the problems surrounding the board structure that we saw at OpenAI not that long ago?

Differences between Nonprofit and For-profit Boards

Joanna: I think what’s interesting about OpenAI is that it is really a nonprofit. And the nonprofit owns a for-profit subsidiary. That for-profit subsidiary is fully controlled by the OpenAI nonprofit. Most people don’t realize that a nonprofit board is very different from a for-profit board. We all understand that a for-profit board tries to maximize shareholder interests. However, in a nonprofit board, they are guided by a stated mission. So that’s a very different output. The nonprofit board does have the same duty of care and duty of loyalty, but they have this separate duty that’s not found in a for-profit organization, which is a duty of obedience. They have an obedience to the laws and the purposes of the organization.

Coral: Before we unpack a lot of that and then apply it to startups and founders more broadly, why don’t we just define the term governance a little bit? I know that in a lot of the media reports and other things that we’ve all been hearing and reading, it’s gotten thrown around a lot, and I don’t know that people fully understand what that means. Okay, governance, yeah, okay, we all get it, but I don’t think a lot of people do.

Defining Governance and its Importance

Joanna: Governance, from a legal standpoint, really refers to how an organization is managed or governed. So, by who and how. For most entities, governance starts with their governing documents. These are the foundational rules for the entity. They’re typically a charter and the bylaws. The charter is where you will find the rules by which everyone must follow when it comes to governing the company. Now, the charter is typically what sets up a board of directors or whoever is going to manage and govern that organization. They have the ultimate management authority over the organization. And then, the authority of the leadership team, the CEO specifically, derives from the authority of the board.

Coral: So then, Soma, if we pivot to you a little bit, you’re obviously the guy with all of the board experience here. I’ve heard you say before that, basically, a board is half company building and half governance, especially in the early stages. So why don’t you break down for us? What’s the purpose of the board and some of the experiences you’ve had in early-stage board building?

The Role of a Board in Early-Stage Companies

Soma: In general, I think a board is all about how you provide the right level of checks and balances to ensure that an organization is being run or managed in the most appropriate way. You follow the laws. You ensure that the right things are happening. You also keep in mind the fiduciary responsibility that the board has in terms of putting the shareholder’s interests first. As much as I say that, you know, governance is important right from day one. Usually, what happens is in a very early-stage company, the responsibility of the board is twofold.

One: Focus on ensuring that the right startup board governance is in place. But it’s also being a trusted partner to the CEO and the founding team and helping with what I call the company building — whether it’s building the team. Whether it’s helping with product building, product strategy, go-to-market plans or go-to-market strategy.

There is a bunch of things that need to happen in early-stage companies where everybody is learning as they go along. And it’s the board’s responsibility to be a trusted adviser and a trusted partner to the founding team and to the CEO. In the early days, it’s actually probably more time spent on company building. And as the company scales, as the business scales, then more of the governance comes into existence. By the time a company reaches a level of scale and becomes a public company, then the board is predominantly about governance and about what I call strategic alignment.

Coral: You made a good point there. Startups are growing — it is basically their purpose. And the board needs to grow along with it, while making sure all of the missions and everything are aligning, which obviously is where we got into a little bit of trouble with the OpenAI stuff.

So why don’t we talk through the board at those different stages? When you bring in independent board members and that sort of thing, and I think you can, both of you can tackle that a little bit. Joanna, why don’t you start by setting that stage and tell us how things evolve as more funding comes in.

Board Evolution with Company Growth

Joanna: I do think, exactly what Soma is saying, although the ultimate role of the board is oversight, that oversight changes over time — just like when you’re a parent. Your parenting duty over a toddler is very different than your parenting duty over, say, a teenager or even an adult child.

I think very much that, at the end of the day, the board’s role is oversight. They oversee strategy. They monitor risks. So, when you launch a startup, you might have a sole director on the board. Like when you don’t even have any investors, you are literally having a board of maybe one, maybe two, maybe both of the founders, if there’s two founder. So at this stage, it’s literally the founders company, and the board directors and officers are often exactly the same people. Once you raise some money at this point, it’s often comprised of three people. One is the CEO. One would be what we’d call the preferred director, which is usually the representative from the lead investor at that Seed stage. And then a third really depends. It could be an independent member. It could be another founder. It could be another investor representative. Something to note here, Coral, and I think most people know this, but since decisions are made on a majority basis, for the most part, it is good practice to have an odd number of board members. Having five members at this stage is a bit unwieldy, so we usually don’t go quite a five. As a company raises more and more money and more investors, such investors may ask for a board seat. We might want to change the composition as they grow and tasks change. And so, the number of board members you will have might increase, but we usually keep it odd-numbered and balanced.

Coral: And Soma, what should founders and CEOs keep a lookout for? Any sort of markers or moments that, okay, maybe it’s time to reevaluate what the board looks like. Do we need independent directors? Are there moments that you look for specifically when it might be time for some of that?

Soma: I think it’s twofold, Coral. It’s a balance, right? You don’t want a nine-member board for a company that has two employees. You also don’t want to have a thousand people in a company and like two members on the board.

But having said that, they should focus initially on getting people on the board that are really, really aligned with you — both from a strategic perspective and from a creating value perspective. The reason the firm that leads your Seed round of financing wants a board seat is — they are putting their money where their mouth is, so to speak. And, they are responsible. They have fiduciary responsibilities to their LPs. And they want to make sure that they have the right level of oversight, governance, and visibility to ensure that all the right things happen.

So every founder has to think about which venture capital firm they’re taking money from and which partner is going to be on their board. And is that partner somebody that they’re excited to work with? Because there is a meeting of the minds, so to speak.

The closest story I can tell you, Coral, is the following. If you are sitting in Seattle, let’s say. And, let’s say you’re the board member, and you have a CEO who wants to take a road trip to San Diego. Both of you need to be aligned that you want to go from Seattle to San Diego. Then you can argue over whether to take I-5 South. Should I take 101? Should I take some other road? But let’s say the CEO wants to go to San Diego, and the board member wants to go on I-90 East. Then you’ve got a problem.

Coral: So, how do you navigate that? Obviously, you do your best to pick those people that are aligned with you and those best partners. But at some point, if you get into a place where things, you’re not aligned anymore, obviously, like we saw in OpenAI, what’s the actual route that should be taken to handle that sort of conflict?

Choosing the Right Board Members

Soma: That’s why it is important for you to know who you’re going to be working with. And you know that like, Hey, this is a person I’m ready to work with because it’s not like you’re always going to be in agreement. There are going to be disagreements. There are going to be what I call debates. There are going to be some ferocious arguments along the way. But as long as you know that you are aligned on doing the right things. And people have different perspectives and different points of view, then I think you can work it out. Sometimes a board and the CEO might just decide “This is one thing where we disagree.” But they have enough trust in the other person that they are going to disagree and commit.

That, I think, is a good way to solve things sometimes because it’s not like everybody’s going to say yes to everything all the time. That’s not what a board is all about. And that’s not what a CEO should be expecting. And if a board expects a CEO to behave that way, then probably they’re not the right CEO in the first place anyway, right? So having arguments and having different differing points of view is okay as long as you can work through them and get to a common place of understanding about how you’re going to move forward. But if you realize that the relationship is so strained for whatever reason and you can’t navigate through it, then some change needs to happen.

Coral: Is there a process by which that change has to happen whether in terms of what the rules are that have been put in place, Joanna, or otherwise?

Navigating Disagreements and Conflicts within the Board

Joanna: There’s usually rules in the charter and the bylaws that talk about how you would remove a board member or how you would add a board member or you would switch out the board members. But you know, again, all of those rules, like almost all legal documents, are there just to be a fail-safe for what to do if you have to figure things out. Ultimately, it’s about the relationship, like Soma is saying. You have to work things through. A very functional board, I think, reflects a very functional culture. And the startup does best when there’s a functional board that knows how to work with its management. Ultimately, it involves both sides figuring that out.

Things that I have seen in my past where it hasn’t worked out great is when the CEO or the management don’t want to bring things to the board. They don’t think the board will understand. They don’t think the board will consider it. And I think oftentimes that can cause some issues when the board does find out. I think it’s really helpful for the founders and the board to have a good trusting relationship and vice versa. The board having honest open conversations with the leadership and the management really go a long way toward having good startup board governance and not having sudden decisions that come out of nowhere and surprise either the leadership or the board.

The Importance of Transparency and Communication in Startup Board Governance

Soma: The thing I would add to that is that it is a two-way street. As much as you could argue that, you know, uh, the CEO reports into the board, the CEO and the board should really think of themselves as like sort of partners. Right. And partnership works when even both sides, like, you know, operate with the same level of, uh, uh, what I call integrity, transparency, communication, and willingness to work together. The other thing is — as a CEO, sometimes you have good news to share, you have bad news to share, you have lousy news to share with the board. Rather than try to be super nuanced about it, ensure that communication happens as real time as possible, whether it’s good or bad, doesn’t matter. Because if you want to build a trusted partnership, it’s really important that you communicate as and when something happens that you think the board needs to know. But things that you think ought to be shared with the board have a different velocity for good news versus bad news. In some sense, I would say, the velocity for bad news should be higher than the velocity for good news.

Coral: Never wait to tell any news, just keep it open and transparent, all lines of communication always.

Well, I think that this will be really helpful for people, you know, thinking about starting a company or those sitting in companies right now. And I thank you both so much for joining me today.

Joanna: Thank

Soma: Absolutely, Coral. Thanks for doing this. And, I think, this year has been particularly interesting with what happened with like FTX and what happened with OpenAI. And all these conversations about like, hey, how do we ensure that the right startup board governance is in place, until we realize it’s too late kind of thing. And so this is a good reminder for every founder and every CEO, no matter whether they are day one, or they are looking to go public tomorrow, or they’re a 1-year-old company — governance and oversight is really, really important. Have the right level of energy and focus and attention on that. Don’t go overboard, and don’t underinvest in that.

Coral: Perfect. Thank you guys both so much.

This week, Madrona Managing Director Karan Mehandru hosts Madrona Operating Partner Anna Baird. The operating partner role is a new one for Madrona, and we’re excited to have her on the team to advise our founders. Before transitioning to board positions in the last couple of years, Anna was most recently the CRO at Outreach. Before that, she was the COO at Outreach, and before that, she was the CFO at several companies. That trifecta of C-Suite experience gives Anna a unique perspective to help founders navigate any blind spots they may have. Anna and Karan, who is on the board of Outreach, dive into how the sales profession has changed, how the role of the CRO has changed since COVID, how to talk to customers, how to build high-performance sales teams, go-to-market strategy, AI’s role in sales, and so much more. These two tackle it all, and it’s a must-listen.

TL;DR

• Maintain focused consistency: When running a business, especially in the sales domain, avoid changing strategies too frequently. Find a core focus and maintain it for sustainable growth.
• Adapt and stay close to customers: The market might change but always understand the pain points your product solves. Continuously evolve by staying close to the customers and their changing needs.
• Balance growth and profitability: Understand the balance between growth and profitability. It’s crucial to stay focused, not just on expanding but also on maintaining a profitable approach.
• Hire for traits, not just experience: When hiring for sales teams, seek traits like intellectual curiosity, a desire to win, and collaborative skills. These traits often matter more than specific experiences when navigating new or challenging markets.

This transcript was automatically generated and edited for clarity.

Karan: Hello everyone. My name is Karan Mehandru. I’m one of the managing directors here at Madrona out of the California office. It is my pleasure and privilege to welcome Anna Baird to this podcast, and Anna and I go back a long way. I’ve had the pleasure of calling her my friend for almost a decade, and I’ve had the privilege of working with her for almost five years out of the 10 at Outreach. So, super excited to have you with us, Anna. Thank you so much for making time.

Anna: Hey, Karan. Super glad to be here and part of the Madrona family.

Karan: Awesome. We’re excited to have you. We have a lot of great topics to cover with you. Maybe we start with how your journey and your career took off and how did you get into tech in the first place?

Anna: I’ve been in tech since I started. I was an accounting major in college and got into one of the big accounting firms, KPMG, but I wanted to do tech. I wanted to do the accounting side and help tech companies. So, I moved to Silicon Valley, San Jose, at the time and was with KPMG for 17-and-a-half years.

I started by helping startups go public. I worked with Google pre-IPO, and then Intuit, and a bunch of others for the next seven years. There’s probably not a boardroom in Silicon Valley I haven’t sat in at one point or another after 17-and-a-half years. I moved to the consulting side partway through that and really loved helping set those foundations to operate effectively.

I left and became a senior vice president at McAfee, running finance governance risk. When they were bought by Intel, I decided to be CFO. I like to do a lot of different things. I became CFO and then a COO and then a CRO. Those are all the C’s I’m covering. No more. I’m done. That was it.

Karan: Great. Not many people we encounter have taken the CFO, COO, and CRO roles, making you unique in and of itself. Tell me why you transitioned from being a CFO to wanting to be A CRO. Did you just wake up one day and say, well, I’m done counting the numbers, and now I’m going to start driving the numbers? How did that mental switch come about in your head?

Getting into tech from accounting

Anna: Well, it’s funny. People don’t realize this, but as a partner at KPMG, I had a 6 million quota a year, so I was already selling.

Karan: We’re all salespeople.

Anna: That is so true. As a CFO, you’re trying to get investors. You’re selling the company. I loved that aspect. I loved customers. I loved understanding what the customer pain was. My favorite thing, even when I was in accounting, was what is product building and how are they solving the customer pain?

I loved understanding those things. I think it’s what made me a good consultant and an accountant at that time. I think it’s also what made me a good CFO, but even as a CFO, I kept ending up taking on other things because I loved the operational side. Everybody’s like, oh, Anna, take this, and Anna, take that. Then I said, you know what? I’m going to do the COO side. I have more flexibility in the breadth.

Then, of course, in 2019, you, Manny, and others asked if I would take on the CRO role. When you are building a company, and it’s all hands on deck, and you’re growing like crazy, and there’s a change that you need to make sure stabilizes, it’s the right thing to do, right? Sometimes, there are things that are just the right thing to do for the business.

One thing I’ve been pretty good at in my career is always putting the business first. I was helping a lot with sales anyway as A COO, so it wasn’t like some crazy transition, but I loved the team, and I knew we needed to stabilize and make some different changes, so took that on from 2019 until 2022.

Karan: We’re so glad you did because you took Outreach. I remember, when you came in, we were still under 20 million, and you took it all the way to 250 million, so it was a wonderful journey. I’d love to ask you about that, but just to clarify for listeners, the decision for us as the board was simple because we just got you to ask for the budget, approve the budget and then drive the numbers, so we didn’t have to do anything. So that was awesome.

Anna: I was the COO and the CRO for too long. It took us a while to find a CFO. That was the easiest job I ever had. I was like, is this a good idea? Yes, it is. Let’s go do that.

Karan: That’s awesome. Let’s talk about how the sales profession has changed over the last decade. I mean, there’s so much that has happened, new tools in tech. There are so many changes in the actual way things are sold, whether it’s product-led growth or sales-led growth. Then, things changed even more when we went through COVID.

I’d love to talk to you about your experience in this profession. You’ve been around CROs, you’ve been a CRO, you’ve been a CFO, you’ve been a COO. So, from your purview, what are some of the biggest changes that you’ve seen play out in the market in the sales profession, and then how did that get either accelerated or exacerbated when COVID hit?

Changes in the sales profession since Covid

Anna: It’s been a fascinating journey, especially these last few years, right? Everybody says, oh, how did you lead through a pandemic? Well, no one’s ever done it, so there’s no playbook. We were all figuring it out as we went and trying to make the best decisions.

I think one of the things that changed, and I’ll start at the top, is the CRO role in general, as it expanded significantly to what you need to understand, the skillset you need to have. I always explain that CRO became part CHRO from a human resource side, part CIO, and part CFO.

CHRO to deal with the mental health and the remote work, and how did you make sure that you kept your team focused and effective and addressed issues quickly so that you could make sure that everybody was healthy and in a good place, CIO, to understand the technology available to you to work with remote employees and what tech was available and how did you think about utilization of that tech for your go-to-market strategy?

Then CFO, the data, I mean data is so key when you have the technology, you have to understand what data you need to be able to run the business and how you are viewing it and what is the rhythm of your reporting and operating so that you could make sure you were making data-centric decisions because you couldn’t see everything anymore.

That changed for CROs, obviously, but it also changed for the teams for their understanding of customer engagement, what was happening, and how they wanted to be educated. When you think about the world being so remote when we first hit COVID and still quite a bit remote today, there’s so much self-education happening with customers.

We’ve talked about the number of touchpoints you have to get a deal approved and went from 10 to 17 to 21. You know what I mean? It was insane. They’re all working in different locations, and they’re not sitting in some conference room talking about your deal, your product every day. So, how do you educate that group? How do you create central locations where they can access business cases and videos on the product and those sorts of things?

If they’re already a customer, how do you do some product-led growth with showing them what’s available? Those things became way more critical than they used to be. For the account executives and the go-to-market teams, people didn’t want to wait to have three calls to get their question answered anymore either. They didn’t have the time. They were back-to-back with scheduled meetings because that’s how we’ve had to operate and still do.

So, they can’t sit there and go, yeah, let me get back to you. Let me get back to you on that question. Let me come back. We used to do things like that, I think, wtih go-to-market strategy, and it was okay. It was accepted as part of the ecosystem, and it’s not anymore. How can you educate a customer before they show up on a call?

Can you send a video to give them some visibility into the product and talk about, “Hey, like you, here are the four things from leaders that we hear?” So you make it, understand their role, and you understand who they are. Here’s some things we hear from leaders like you. We’re experts in the industry, how could we help you and how can we make this next call the most effective? Here’s a video of what our product does. Those kinds of things are game-changing now.

Karan: It’s interesting. As I listened to you, it makes total sense, and it sounds coherent, it sounds logical. That said, put in the context of companies that are starting out and the founders and their experiences. There’s so many founders that we work with these days that are product founders or tech founders, and a lot of times go-to-market isn’t a very natural trait, that skill or experience that they’ve had.

A lot of times, they don’t have any go-to-market people around them, so help us understand, when you advise founders now in this role as an operating partner at Madrona and being on boards of companies and you’re looking at early-stage founders that just got their product out, how do you advise them to interact with their go-to-market teams?

How do you think about advising them about go-to-market strategy? At what point should they start thinking about it? If you build it, they will come, which is usually the model that a lot of them operate on. So, help us understand what do you advise them when you’re looking at a new founder or a founder of an early-stage company that’s starting to think about go-to-market strategy.

Go-to-market strategy

Anna: I think there’s a couple of things. Getting your go-to-market organization, getting them to have a foundation, a process that is consistent across the teams, you have to build it, right? You got to build it from that, again, that foundation up. Part of it is, you’ve got to make sure you let them focus. Don’t change strategy every quarter. You cannot have new messaging, new strategy. You need to make sure you’re getting the message out to the market and watching how the market responds, and that can be pretty fast, but don’t change strategy every quarter. I think, especially when you are a product founder, you can change product sometimes quickly, and nobody knows. They don’t see what’s happened in the background.

When you’re changing your messaging all the time, you definitely see it, right? Each of those go-to-market people is the best marketer you have. They’re the voice on the street every day, talking about who you are, what you stand for, and what pain you solve.

I think that is so critical because, especially as product founders, you get so engaged in the engineering and the tech, and you want to talk about the features that you built and the functionality that you have, and your customers don’t care about that. They care about the pain you solve for them, the problem that you make less painful. So, how do you make sure that you are creating the language and keeping that consistent?

Here’s why we exist as a company, here’s the pain we solve for you, and here’s why we know your pain. We’re totally empathetic because we see other leaders like you all the time because that’s who we talk to every day, and we know that these are the types of issues that you run into. Then you get that whole empathy, I’m in your shoes, you’re selling with them, not to them, because they’re trying to solve a problem in their organization, not buy a feature or functionality.

Karan: Love it. I love that framework — why we exist, what we do, and how we do it — which is a great framework for all of us to remember, and more so, I guess, in a time when there’s massive change, just to remind everybody to rinse and repeat that message. You used the word focus, and I want to come back to that.

Obviously, scarcity does breed focus, but then companies grow up, and then they raise a lot of money, then they become multi-product, and then they hire a lot of people, then they have multiple locations, and then they have the tyranny of choices in front of them.

We went from scarcity to abundance in the market, and now we’re going back to a little bit of scarcity because the market is correcting in front of us. One of the things I’d love to understand from you is, you manage the sales team through scarcity, then to abundance, then back into scarcity.

You’ve been a CRO, and you’ve been a COO, and you’ve been a CFO, so you understand the difference between growth and profitability more than most CROs who are just thinking about growth. It would be great to hear some of your anecdotes or lessons learned as you manage that massive growth engine within Outreach and Livongo and all these companies. How do you balance that growth and profitability? How do you breed focus when there’s so many choices in front of you?

Balancing growth and profitability

Anna: We say it all the time, and it’s like what you say no to, right? You do have to say no to things. You can’t have 20 key priorities for the business for a go-to-market team or engineering. There has to be three things that we are trying to accomplish in this 12 and 18 months, whatever your timeframe is, and you have to be maniacal about that focus.

That’s where it’s not changing every quarter because it takes time to build it. If you’re looking at building competitive moats and you’re looking at solving that customer pain, whether the market has an abundance or in scarcity, you control what you can control, which is why do you exist? It’s back to that again, right? Why do you exist as a company? Because you solved something, and don’t ever forget that.

Even when it’s abundant, you must stay focused on solving a pain. How do we solve it even better? How do we make it more effective? How do we make it faster? How do we solve the next piece of the pain, and building that roadmap of understanding? We talk about our product roadmaps and really needing something that is 12, 24, and 36 months out of what would we do next. If we said this is the customer’s pain, and here’s what we’re seeing, what would we take next?

Always stay close to the customers, always stay close to understanding what’s happening in the market, and make sure you are controlling what you control, which is what your company does, and what you build, what you put in the marketplace. That is always a recipe for success. What happens is people get scared, and I get it. When you watch your child suffer, it is hard to be objective.

You must also surround yourself with leaders who will help you. It’s too personal. So, when you have leaders around you who can help you go, let’s all breathe for a minute together and let’s talk this through. What will be the most impactful thing for our business and our team?

Always remember the most impactful thing for your customers is going to be the right strategy to go after, and obviously, based on the tech that you built and where you have the assets to take those major next steps.

Karan: That’s great advice. I want to talk about culture and as part of that, hiring in particular, and you mentioned one of the first things you can do is build leaders around you that embody the cultural values that you and the company espouses.

When you think about building a high-performance sales team and when you think about hiring your next-generation leaders, maybe even the first rep or the first five reps, what are some of the organizing principles or traits that you look for when you’re lighting up a sales team?

Building a high-performance sales team

Anna: That’s a great question. There’s something that a lot of people don’t do. Everybody’s like, I just hired a salesperson, and they were great, and they came from this other company and they did great there, so they’re going to do great here. Anybody who’s been in this long enough, you know that’s not true.

I still remember when I first came to Outreach. We hired some of the top salespeople from companies that had been crazy successful, but they’d never introduced a new category. They’d never done an educated sell because the company had a great product that people came to them, or they just needed to talk about it. When you’re trying to educate for something that never existed before, that is a whole different selling strategy.

So, it’s like what are the core attributes you’re looking for in your salespeople, and those understanding the business you have and what they need to do in the marketplace is the first step. What are the traits that you need? But I’ll say three traits that are always critical when you think about the skill sets you’re hiring for.

It’s intellectual curiosity. You want them to understand the customer’s pain. That’s critical. Understand the customer strategy. How do we align our product so that we’ll win every time? It is a desire to win, Which comes from a lot of different places: overcoming adversity in their background, big families is a good one. They’re one of seven children. They had to compete to stay ahead.

Not just athletics, but competitive arenas in general, like chess, and lots of different areas where people are like, I really like winning. That adrenaline is that desire to overcome and to get to that outcome. Then you also need somebody who knows how to quarterback but is an incredible collaborator because one of the things they know is how to bring the power of a team to a customer, not just themselves.

If they’re all about being the hero and winning every time on their own, that works okay in some of your early days sometimes. That is not a strategy for long-term success. When you bring the power of the team, you win faster; you win more efficiently; you win bigger.

So, somebody who knows how to get out of their own way and they don’t have to be the star. They know how to take a step back, bring the right people in, and prepare them for a great customer conversation. I think those three.

Karan: I remember you telling us all at one of our sales offsites at Outreach when you got up on stage and said if you want to go fast, go alone. If you want to go far, go together. That’s one of the African sayings you reminded everybody, and I still remember that.

Anna: Yeah, and it’s totally true. It, just like history, repeats itself over and over and over again. We see that. You forget sometimes. You have to remember the basics.

Karan: So, Anna, now let’s fast-forward to today. You are our first operating partner at Madrona. You’ve collected a wealth of experience, lived every role in a company that exists from early to late, worked in different industries, such as sales and healthcare. I would love to hear your reasoning for why you joined Madrona and why you’re an operating partner here today. And then secondly, to the next Anna that is just graduating from college and potentially getting out of KPMG and aspires to be a CEO, CRO, COO, what have you, what would you tell the 22, 23-year-old Anna that’s graduating today?

Why Madrona

Anna: Oh, great questions. I joined Madrona because I’m a big believer, and this was part of the stage of my career. I think it’s important, and you see this a lot at Madrona. How do we help fill in the skill sets and the gaps that some of our founders might have? They’re not all coming from 10 different backgrounds, so how do we bring to the table skills and capabilities to make them even more successful?

When you’re a founder, and you’re looking at, especially those ABC rounds, you need a partner who will bring to the table things that help you figure out those corners you can’t see around because you’ve never been around them.

This stage of my career was about giving back. It was about teaching all those things that I learned in all those roles you mentioned and how to avoid the mistakes and take advantage of the opportunities faster because you do learn a lot of that. It’s like getting a mortgage. By the time you get good at it, you don’t do it anymore. So, how do I make sure I help other people avoid that mistake or those mistakes?

In this marketplace in particular, when you are looking for a great partner, people like Madrona are thinking about bringing executives onto their teams, prior executives onto the teams, and we have quite a few at Madrona. Those people have your back. They’re there to make sure it’s one phone call away to say, hey, can I talk about X or Y?

At one of our CEO’s all-hands the other day, I talked about this maniacal focus on focus. Stop looking at the market and focus on the pain to solve, what you’re building, and how you’re taking that to market. If you do that, you will get through this. Getting distracted by all the noise is one of the challenges that you’ll deal with, and we’re learning it well right now. That’s really critical.

One of the other things that is key, and you hit on this earlier, is the culture you build. We were talking about performance culture, and I just wanted to hit on it for a second. One of the reasons I joined Madrona was because it had an incredible backbone of ethics, culture, and incredible skills on this team.

That is important as you think about the partners you’ll work with because part of something that your board and your investors will do for you is help be culture bearers and think about what’s going to make your company successful. When I look at founders and some of the founders that I’ve worked with and said, okay, what are the things that help lay some of that foundation?

It’s okay to say, I don’t know. As a leader to say, it’s okay to say I don’t know, but it is not okay not to try, not to come with a perspective, and not to come with a point of view. When you start to create that openness, you get the diversity of thought. I always try to make sure I say it. It’s, I don’t know, I don’t know the answer to this, but let’s talk it through. Let’s work this out.

As a founder, when you do that, it’s even more powerful because you are the one who created this product, this company, so bringing that openness and opening the floor to diversity of thought gives people the freedom to do that. Fear is a tactic, not a strategy. When you put fear in play, it’s because you are trying to emphasize how dire a situation might be.

Sometimes that’s okay, but you use it at a point in time, not as a strategy every day, because people who respect you and want to please you and want to make sure that they’re working hard for you work 10 times harder if they don’t have that fear because fear makes them do things that are not thoughtful and are not strategic and are the opposite of really what you want.

When they admire you, they are going to work that much harder, and I think it’s a foundation you have to build from the early days of who you want to be, what you want your company to be known for, and when you have those things, you also recruit incredible talent. That is something I see with some of our founders, and what would I tell the 22-year-old Anna?

Karan: Yeah, 23-year-old Anna. I mean, that’s only five years ago, so you should have no problem remembering.

Advice: don’t try to make failure look pretty

Anna: I wish it were only five years ago. I think I’ve been pretty good at this, but being an executive and a founder is about taking risks. It’s about saying no to things, but it’s also about what you say yes to. But don’t try to make failure look pretty. Call it a failure and move on.

It is one of the most critical things that we all can do, and it is easy for us to say it wasn’t that bad. It was okay. If we just do this with it, if we just do that with it, or maybe we just need to try harder. Sometimes, it is just failure, and that’s okay. You learn so much from that, but you don’t if you don’t pivot and take another direction quickly. If I’d known that earlier, it would’ve been super helpful.

The other thing that I just hit is, what pain do you solve in the marketplace? It’s why you exist. Don’t ever, ever forget that. I say this to our founders, too. You have to sit in front of customers, no matter what leader you have on the team. I’ve heard CROs go, oh, when you get to be a CRO, you don’t have to do customer meetings anymore. I was like, that’s insane. Everybody should be doing customer meetings, everybody.

Heads of product, heads of engineering, heads of marketing, obviously CEOs and CROs every day, right? That is how you make sure you are staying in touch with what you’re solving in the marketplace, and is the market changing on you because, if you miss that, you miss those cues, then it will cost you. It’s going to cost you time. It’s going to cost you money, so stay close to that pain, and don’t make failure look pretty.

Karan: I love it. Talking about failure reminds me of a wonderful speech that JK Rowling gave at the Harvard commencement a few years ago about the power of failure. Your comment around failure reminded me of that, and I would encourage everybody to listen to that, as well.

Anna: You are the king of commencement speeches. I know that is one of your things. It’s funny, I just read something the other day. Apple was talking about one of the things they look for in all their interviews, and I was like, okay, I’ll click on this. What do they look for in every interview? It’s for somebody to say, I don’t know.

Karan: That’s great. Awesome. Well, I have two other questions that’ll hopefully be short. I don’t think I can do a podcast in 2023 if I don’t talk about AI. So, is AI going to take all the sales jobs away, Anna?

Is AI going to take all the sales jobs away, Anna?

Anna: No. Hopefully, it just really helps. I think we’ve talked so much about how do I enable you to be faster, better, and smarter. That’s what customers want. They want you to be faster, better, and smarter, so how do we do that with AI? AI is going to be game-changing in really improving time-to-value for customers and in time-to-value for go-to-market teams, as well, right?

On both sides of that, that’s a win-win. Will there be changes? Absolutely, and you see it already, but I think there’s so much there that will be positive, and AI is not as sophisticated as we’d like it to be yet. We all want it to be the end-all-be-all.

As you all know, you have to tell it the 10 things to get to the answer you want, and it’s only as smart as what it can find out there. That will obviously improve. That will get better. It will learn, and so will we. There’s just, we are going to get into a culture where go-to-market stategy is a lot more effective than it used to be.

Karan: That’s great. Well, I’m sure many people will breathe a sigh of relief after hearing that. Are there any books you’ve read recently that you would recommend to our listeners?

Book recommendation: “Simplicity”

Anna: Oh, wow. The one I just started is called “Simplicity.” It’s by Edward De Bono, and it’s about how you boil things back down again? How do you get back to simplicity? It resonated for me, and I’m just starting it. Somebody had recommended it to me, so I’m going to recommend it out, but it was about the world being so complicated, and we are trying to do 25 things to impact, and sometimes it’s not 25 things, sometimes it’s three.

We over-engineer ourselves. We over-engineer the problem sometimes. I loved this concept because, as I step back and I look at my career and the not five years that it has been, it is so much. There are three key lessons. I come back to the same core. I come back to the same principles.

The problem is we forget to focus on them, and we don’t focus on them and weave that thread through everything we do, and that’s what creates more of the challenges. We over-complicate the environment versus focusing on what’s really critical here for us to win, to be successful, and to move our company to the next stage.

Karan: That’s great. I love it. Well, on that note, I want to thank you, Anna. This has been awesome. Obviously, I was sad when you left Outreach, and I figured we’d never get a chance to work again, but I’m so glad you chose to come to Madrona as an operating partner. Our founders are lucky to have you as an advisor.

We are lucky as investors to have your insight and experience guide us in our investment decision-making, and I’m just so happy that we get to have lunch right after this in the office. So, thank you again for your time. Thank you for sharing all these pearls of wisdom with our audience today.

Anna: It has been such a pleasure. I love working with this team. Thank you.

Coral: Thank you for listening to this week’s episode of Founded & Funded. We’d love you to rate and review us wherever you get your podcasts. Thanks again for listening, and tune in a couple of weeks for our next episode of Founded & Funded.

Today, Madrona Partner Aseem Datar hosts Tigera CEO Ratan Tipirneni. Tigera is in the business of preventing and detecting security breaches in cloud-native applications, and its open-source offering Calico is one of the most widely adopted container networking and security solutions out there. Aseem and Ratan dive into whether or not you should open source, the three business models founders should evaluate when thinking about commercializing an open-source project, how to compete with free, how product-led growth can really kickstart your business compared to traditional go-to-market, and so much more.

This transcript was automatically generated and edited for clarity.

Aseem: I’m incredibly excited today to talk about open-source in the context of one of our portfolio companies, and I’m excited to have the CEO of Tigera, Ratan, with me. So, Ratan, very excited to have you and a warm welcome. Would love to maybe just get a little bit of your background in your own words and tell us how you got started with Tigera.

Ratan: Thank you, Aseem. So first, I have to say I’m very excited to be here. Thank you for the opportunity. My background is — I’m a software entrepreneur. I’ve been working in the Valley for a while, and that’s what I enjoy doing: building businesses and software. At Tigera, we are in the business of preventing, detecting, and mitigating breaches in cloud-native applications. We started several years ago with some really smart engineers and architects who got together and decided to solve some pretty tricky and gnarly problems around container networking and container security. That was the company’s genesis, and since then, we’ve been on a really exciting ride.

Aseem: This is a common pattern we see today — really smart people hunkering down on a potential challenge they see in the ecosystem and going after it. The question that comes to my mind is how should founders, engineers, and those seeing these set of problems think about it in the open-source context? When do you start building something, when you decide whether to open-source it or not, like how do you think about real demand, and what are some of the early signals that folks should watch out for as they’re thinking about going down the open-source path?

Ratan: The best way to get started is to get out of the starting gate, build something, put it out there, and get some conversations going with the community. The conversations are super critical to building that sense of camaraderie within the community and, more importantly, to get some feedback and participation from the community where they feel like the community has a voice in terms of providing feedback, contributions, engaging the community, and really seeing where that goes and letting the community actually steer their project. That’s probably the best way to get started.

Aseem: Are there any specific metrics to look at? Like people track all sorts of things. They track stars, they track the number of contributors, and they talk about how many repos are getting created. What are some concrete metrics to sort of get behind and understand if there’s real demand?

Ratan: Sure. So, I’d say there are probably quantitative and qualitative metrics. In terms of the quantitative metrics, I heard you say the stars — this is a personal opinion — but it’s a vanity metric. I’ve literally seen that being gamed, where I saw one of the founders at KubeCon offer to hand out some gifts for users to go liking some stars.

But signs of daily usage of your software are probably one of the best metrics and most objective metrics you can get. Remember that the metrics are as much for you, as they are to help you build some credibility in the community. You really need to know that someone’s actually using it and getting some value from that.

In terms of the qualitative stuff, it is equally important to have the dialogue and get some feedback from users on how they’re feeling about it. That could give you plenty of clues on perhaps what you may want to build in open-source or perhaps what you want to build in your commercial products. But honestly, just being curious about it and engaging with users at every single opportunity is probably the easiest and fastest way to get some of the qualitative feedback.

Aseem: Continuing down the path of commercializing an open-source project, Ratan, what’s the point you arrive at when you really decide, “Hey, how do I commercialize that?” Is there a stage gate that one should think about? So, talk to us a little bit about that journey and how Tigera got to that stage.

Ratan: There are two parts to my response to that question. First is when you start to think about it, and second is when you actually pull the trigger on that.

When you start to think about it is — you start to think about it on day one, right? Even when you’re launching the open-source, you have to be thinking about your commercial offer also. It doesn’t mean that you’ll launch the commercial offer, and you should not launch it on day one. But you certainly have to be thinking about it because any decisions you make about product functionality and packaging in open-source will have an implication on your commercial product strategy. So you’ve got to do that right away.

The second part of the response is when you actually pull the trigger and when you start to launch a commercial product or monetize it, I’d say you have to wait until you have some minimum critical mass in your community where you feel like there’s enough momentum where you hit escape velocity before you actually think about launching and monetizing your commercial product.

The other interesting concept, at least my mental model, is you have to be thinking about building two separate and distinct operations or businesses from day one. You have to think about building that momentum in open-source, the open-source project, the community, the users, the value they get, the accelerators, and all that. That’s one part of the operation. You also have to simultaneously think about how you actually build a commercial business around that. From a timing, it’s difficult, but you have to be doing both in parallel because if you over-index on any one of them, it may not be an optimal outcome for the team and for the company.

Aseem: Pulling on that thread a little bit, at what point would you say a company is not ready to transition over? At what point do you say, “Hey, you’re not there yet, let’s not turn commercialization on?” Is there a moment like that, or is there a gate that you absolutely must cross, which could be a binary one? Or is this more art and a science to go figure out like, “Hey, do we turn commercialization or revenue on?”

Ratan: One simple heuristic I’ll offer is that if you’re struggling to build community and are not seeing acceleration in the number of users month-over-month, you really have to question whether this model is even worth pursuing. That’s a moment of truth where one may have aspirations of building an open-source-based company and a go-to-market model based on that, but if you aren’t able to get acceleration in the open-source adoption, perhaps it’s not even feasible, and maybe it’s time to go try something else. Be a little bit premature to launch a commercial product to layer on top of the open-source product.

A good time to be thinking about launching a commercial product or figuring out how to monetize your open-source is if you hit a point where you feel like the open-source community is growing. Also, if there’s some word of mouth and a level of acceleration, you see month-over-month in the open-source usage.

Aseem: Shifting gears to a topic relevant to the commercialization chasm is really around business models. What are some of the business models? What are some of the considerations that one must make to transition over to commercialization, especially in the open-source world, that the founder needs to be aware of? How should they think about monetizing?

Ratan: From a historical perspective, there have been three business models which have evolved over time. The first generation of business models in open-source was you give everything away for free, and then you try to monetize it through commercial support. There’s been one company that has done it extremely well, but it’s yet to be replicated since then. I’m not quite sure I would recommend that to anyone at this point. Although I do see some companies in open-source make the mistake of putting everything in the open-source bucket, and, at that point, they have no other option but to try to monetize it through commercial support. That’s pretty dangerous.

I’ve seen someone in our ecosystem do that recently, and that’s a one-way street where you’re eager to build community and momentum. You dump everything into open-source because you want to try and do one better than the other companies in the ecosystem, but you can’t recover from that. Your only option is providing technical support, and it’s not a very defensible way of building a great business. Even if you succeed initially, what you’ll end up seeing is you’ll end up seeing quite a bit of churn in later years. You just will not get that cumulative momentum.

Aseem: How do you compete with free, and how should you think about competing with free? Or is there something where you compete with free that’s viable?

Ratan: I think the broader uber issue behind your question is, “Hey, how is a company able to do this for free?” I would offer to say it goes back to capital markets. Suppose the capital markets are throwing money at companies, and there’s an endless supply of capital. In that case, that is when we saw a lot of these company adopt this approach of putting everything into open-source. Giving it away for free with no accountability, and some of them maybe had unicorn status.

It’s safe to say that those days are past, and we are back to fundamentals. Even if a company does something like that, it will only sustain for a short time. Trying to compete with free by offering more free stuff is a losing battle.

The other part is it also depends on the quality of customers you’re trying to see because the really high-quality logos that you’re trying to build a business on — they actually want to pay you money. They want to see you successful because they understand that open-source companies, they can’t really abuse the system. If they do that, it’s not sustainable for them, and they’re going to end up losing in the long term.

Even if you’re competing for free, I would say that you want to go find customers like that where, even though they may have a choice of something free versus someone else, an open-source company that’s offering a paid version, they know that the right long-term option is the paid version because that’s the only sustainable solution.

Aseem: Absolutely. I think that’s very helpful. Let’s get back on track. I think we talked about the first business model. I know you had a couple of others, so maybe let’s just continue on that path.

Ratan: The second business model, which was tried about a decade ago, was being able to layer on a commercial product on top of open-source. Essentially, you’re selling something open-source, and you have a project that you offer. Then you provide some additional commercial features on top of that that a customer can buy. Several companies have done a stellar job monetizing and using that approach. If I were to pick a few names, GitLab did a brilliant job. MongoDB built a very successful business model using that. I didn’t give you an example of the first generation, but Red Hat is the most obvious example of the first generation of those business models we talked about.

The third generation, which we have seen more recently and has had a lot of success, is offering a SaaS version to be able to offer a service as a commercial product. The best example in recent history that certainly had a big impact on our decisions was Databricks, where they started at the second-generation business model and then they switched to the third-generation business model I talked about and started to offer a service. They’ve seen really, really good success.

Tobany entrepreneurs who starting today, jump to the third generation right away out of the gate. It’s difficult to switch from the second-generation to the third-generation business model I described. If I were starting today, I’d switch directly to a SaaS model because it gives you so many levers and a better ability to experiment rapidly and really sets you up for success if you’ve got this motion of leading with open-source to create a wedge and following through with the SaaS offer.

Aseem: That’s very well articulated, and I’m sure a piece of gold right there for some of our founders as they think through commercializing an open-source project and which model to go back to. I think there’s a pretty obvious one in our discussion, but not a one-size-fits-all kind of thing.

What surprised you as you were building Tigera and going through the motions about commercialization, signing enterprise customers, maintaining the open-source, and working with partners? What’s the one thing that has been a little bit like, “Gosh, didn’t expect that?”

Ratan: There’s perhaps a little bit of a fallacy and assumption about open-source because it’s free that you assume that users will adopt it. That’s just something to rethink.

Another way of saying it is, what is the cost to a user to adopt your software, to install it, to deploy it, to derive value from it, to maintain it, to operate it? You got to think about that entire life cycle. What is the cost involved? And if that’s not a good equation, the odds of you building a great community that is and an adoption base that’s growing rapidly, the odds of that are going to be pretty low.

The kind of rigor any entrepreneur would apply to building a commercial business is you have to apply the same level of rigor to your open-source community. Just because you give it for free, you don’t get a free pass on all those other things.

What makes it challenging is that you’re going to do all that in a very cost-effective manner because, to state the obvious, you’re not getting paid for any of that stuff. Unless you’ve got an incredible amount of leverage in your operating model to accomplish all those things, it’s going to be very difficult to pull off.

Aseem: Well, that’s certainly very true, at least from the perspective of surprises that you want to avoid.

Tell us a little bit about getting scale. I think we all think about new motions around PLG, self-serve channel. Like from an open-source perspective, is there one go-to-market type that works better than the others? What’s your perspective on that?

Ratan: It depends on the space you’re working in, so I’d be very reticent to generalize a single playbook that would work for everyone. It depends on where you are in the stack or what domain you work in. The models work very differently, but since it touched upon PLG, let me talk about that.

I’m a big proponent and a big fan of PLG for a few different reasons. The first is, fundamentally the buying behavior compared to even five years ago has changed dramatically where most users really want to actually try out your product, use it, and test it in production before they decide to make a long-term commitment to that. The best way to make that happen is to offer them a free tier of product that they can activate, start to use, derive some value, and evangelize to others. So that’s the first thing.

The second aspect of buying behavior I feel that’s changed dramatically is that the traditional purchasing model of top-down, with the CIOs making all the decisions, has shifted quite a bit. The power is now in the hands of the people in the trenches: the DevOps engineers, the security engineers, the DevSecOps engineers. They are the first to have an intimate understanding of the challenges, and they are the ones out there looking and shopping for tools. They have a very tight-knit community of peers who they talk to to get information. They’re certainly not talking to famous analysts to get recommendations for any of the tools they use, and there’s a disproportionately large amount of power in the hands of those people.

And the third is a lot of these buyers, the new generation buyers, the DevOps engineers, DevSecOps engineers, security engineers, they would rather not talk to anyone. They would rather go self-serve, educate themselves on the product, and enjoy the process of getting to know new tools. They enjoy the process of going through tutorials, reading stuff, downloading stuff, playing with it, and using it. For all those reasons, PLG or product-led growth is a really natural motion that aligns with that buying behavior.

Now, that’s easy to say. It’s extremely hard to operationalize the PLG model. It is not incrementally harder. It is harder by a few orders of magnitude compared to a traditional go-to-market model. The reason is you may have a lot of warts in your go-to-market model that you can hide using people. In a PLG model, it’s all out there. There’s no place to hide, and so it’s extraordinarily difficult to do that. If you can pull it off, some of the fastest-growing businesses that we have seen — that’s the one common denominator. They’re all using a PLG motion, and if you actually nail that part, the operational part, the business is off to the races.

Aseem: Two interesting things to sort of draw out from that. One is this notion of — so much of PLG is of the essence of how open-source started. It’s community-led growth because you’re letting the community tell you more about what works and what’s not working in the product.

The second thing is in a traditional go-to-market, your limit or your bottleneck is going to be people. In a PLG type of motion, there is no physical bottleneck that you’re going to hit. For better or worse, I think that’s where the fastest growth is going to come from, or that’s where the fastest learning is going to come from.

Ratan: Absolutely. There’s a tremendous amount of learning. I’d flip it around and say there is no single way to do PLG. You have to think about it as a process, a process of learning, and you’re learning through some rapid experimentation. The only playbook is how rapidly you can learn with a series of small experiments, so the solution may manifest slightly differently for each company.

Aseem: So much of what you’re saying is resonating with signals that we see across the board. What is something you wish you knew going into all of this? Is there a fun story there? I always think that there are nuggets of brilliance that come out of this question, so I have to ask you that.

Ratan: You know, that’s an interesting question, so my response may not be what you’re looking for, Aseem. Being an entrepreneur, you have to be slightly insane to be an entrepreneur. So, let me just state that it’s not a path for rational people.

Aseem: So I am taking notes, Ratan.

Ratan: One of the reasons I think people like me chose this path is because what’s interesting on a day-to-day basis and the breadth of challenges that you have to deal with that keep you really in a state of flow and engagement. It’s been a lot of fun decoding everything that we have done here over the last several years. I don’t wake up any day and say, “Hey, I wish I knew this before.” Because if I knew it before, it wouldn’t be fun, and, honestly, I won’t be doing it.

Aseem: I am taking away a “no regrets” learning there, which is fascinating. Ratan, any parting thoughts or advice for founders who are playing in this space, who are thinking about open-source, who are along this journey of decoding, as you call it, like the challenges they’re faced with?

Ratan: Here are a few things I’ll share because we are seeing some of that play out in the ecosystem that we compete in.

The single most important thing is to stick to fundamentals. What I mean by fundamentals is to be clear about who your end-users are, what their pain points are, and serving them high-integrity manner, and everything else will take care of itself. And try to misrepresent things either to users or customers in the short term — it may feel like it’ll help some companies, but it really backfires in a big way in the long term.

For anyone trying to pursue open-source, my advice is to reiterate what I said earlier: they’re two businesses or two operations, so you’re building in parallel, and you don’t get bonus points for nailing one of them. You’ve got to nail both of them together. From a timing perspective, you’ve got to be thinking about both simultaneously.

The third part, which may sound a little redundant, but it is worth stating, is to not over-index on one aspect, especially the open-source. In the early days, especially if you’re trying to come from behind, we are seeing someone in our ecosystem do that where they’re trying to come from behind, and you throw everything into the open-source bucket, and it feels really good for a while. It’s like eating sugar; it’s going to feel really good for a while, and then you pay a heavy price afterward because you don’t really have a repeatable monetization strategy.

Other than that, it’s a fantastic time to experiment with this go-to-market model. If you are, I’d say jump to the third-generation business model, which is to take your open-source, layer it on with a SaaS service, and if you have a strong stomach, try PLG with it.

Aseem: That’s awesome. I think you summed it up really well. Ratan, thank you so much. I’m sure that our listeners will enjoy every moment of this conversation, and so much to learn from this. I’m taking away a few keywords that I will repeat: insane, have the appetite for having too much sugar, and don’t worry about commercializing an open-source project in the sense of being obsessed with it.

Ratan: Thanks, Aseem. It has been fun. I appreciate the opportunity. Thank you.

Coral: Thank you for listening to this week’s episode of Founded & Funded. We’d love you to rate and review us wherever you get your podcasts. Thanks again for listening, and tune in a couple of weeks for our next episode of Founded & Funded, where we’ll get to know Madrona Managing Director Tim Porter in one of our investor spotlights!

Today, managing director Steve Singh sits down with Troop Travel Fo-founders Dennis Vilovic and Leonard Cremer. Madrona invested in Troop and its corporate travel meeting management platform in 2021 and was particularly interested in how Dennis and Leonard were effectively creating a new technology for a market segment that was largely served through manual processes. These three talk about looking for the right problem to solve to launch a company, and how you decide if it’s the right time. They also discuss when you need to switch from bringing on generalists to bringing on experts, how to identify the right investors to work with, why working closely with customers and potential customers is so important, and so much more.

This transcript was autogenerated and edited for clarity.

Steve: Welcome to another episode of Founded & Funded. As a direct result of a global pandemic, the nature of the workforce has changed. Where we live and where we work has evolved. What has not changed is our desire to get together to collaborate and improve the quality of life, so how do globally distributed teams come together to collaborate? To explore that question, we’re going to be speaking with Dennis Vilovic and Leonard Cremer, the co-founders of Troop Travel, a next-generation meetings management travel platform.

Dennis, Leonard, why don’t we start with just speaking to your experiences prior to Troop? What was it, as far as a pain point, that inspired you to create Troop? I’d love it if you could explain how two people who were unlikely, perhaps as a pair, came together to solve this problem.

Dennis: A quick introduction about myself. I’m Dennis, one of the founders, German originally, economist by profession. I’m based in Spain, and I used to work my whole life as a consultant, driving efficiency and how government institutions, or organizations in general, operate. So introducing new business processes and technology, etc. And there were two key moments in the time before founding Troop, which basically led to identifying that there’s a big challenge we are solving here right now. One was in 2014 when actually I was trying to meet up with a friend who was living in a different country. I hadn’t seen this person for a year, and we said, “Hey, let’s just come together somewhere,” It didn’t really matter where. We just wanted to spend a weekend together.

So I remember I literally went into Google, and I put in, “Okay, I’m based in Madrid. My friend is based in Munich, where do we meet?” And I thought there would be this service to help us — to tell me, “Well, if you meet in Paris, that’s the cost, and if you meet in Lisbon, it would look like this, etc.” Well, it didn’t exist, so I manually gathered these data points in a Google Sheet. It took me about a week to figure out how to bring two people together, and in the end, we realized, Milan would be the best option. So we met in Milan, and it was great. Fast-forward a few years, and the second part of the story is that I started to work for a global organization, which is a membership organization, having a lot of members all over the world, bringing them together frequently, and I was part of a remote team at the time already. I was wondering, whenever we organize our internal team meetings, why would we always go to locations that, in my opinion, didn’t make any sense.

For example, we kept going to Paris. Paris is a beautiful city, but it’s very, very expensive. And I thought, why are we not going to Prague, for example? And that’s where I had this frustration in that organization, and that’s where I met Leonard, who was a member of that organization, that’s where we realized we should maybe try to solve this.

Steve: Fantastic. And Leonard, what’s your background, and how did you get interested in solving this problem?

Leonard: Now Steve, so my background is from a technology perspective. So, I studied engineering, and then I worked in telecoms for a number of years. Our previous business that we founded was integrating with mobile operators and doing high-volume transaction switching for them, and we grew that business across Africa and to a number of countries. But then that industry started to change, and I joined the organization EO, Entrepreneurial Organization, to network internationally and to look for other opportunities. I think it shows you also the power of in-person meetings where I met Dennis in person two times at some of these events organized by the organization. And it’s really through those events and discussions with Dennis that he introduced me to the idea.

Initially, I didn’t know if there was opportunity or not. But after the second time I came back to South Africa, I started doing some software around it to test the scenarios, and then we saw that actually there was opportunity to save quite a lot on the cost side of the travel, but also the travel time. And then, from there, we started exploring the idea properly and talking to funders, etc.

Steve: That’s incredible background. One of the things that is common across both of you is that you both worked at fairly large organizations in the past, had very comfortable jobs, and could meet your needs and your family’s needs easily. It’s risky. It’s challenging to go start a business. How did you get comfortable with that idea? What drove you to say, “Yeah, this is the right time to go do it, or this is the right thing for me to do?”

Dennis: For me, it was really, I was working in this organization which was working with entrepreneurs. I had a direct contact with these entrepreneurs who were kind of a different breed, and it was fascinating working with them, the energy and always solving problems. So I really got excited about being an entrepreneur myself, but I was looking for this opportunity, I was looking for this one idea, and I couldn’t find it. And the light bulb went on at an event in Bangkok in 2016, I think, where I was having a drink with the founder of Trivago, and I was talking about this frustration I had in that organization with how we are organizing our meetings. And I asked him, Rolf is his name, “Rolf, why does a service not exist where you bring in your starting location, it tells you where we should we meet?” And then he looked at me, and he said, “Well, what a fantastic idea.” I was like, “Wow, oh really?” So is that the idea I was looking for? And then, the next evening, I was having dinner with Leonard, and I was telling him that story.

Leonard got very excited, and then two weeks later, Leonard was back in South Africa. I was back in Kenya at that time. Leonard gave me a call and said, “Hey, I actually looked into that idea, and I think there is something.” So there were some key moments which led us into, well, at least spending our free time working on this concept, and then the real game changer came as then a few months later actually the founder of Trivago said hey, he wants to invest in that idea we have, and then I said, okay, if someone gives us money to build this business, then we have to go all in. And that was, for me, the reason then to say, okay, let’s actually become an entrepreneur, and have enjoyed it very much so far.

Leonard: For me, the journey was a little bit different. I think for me because I’ve gone through the journey myself, I understand what the adventure is of building a company. And I think it’s just, in my previous company I’d been there for 15 years, and it was an established company, and I think that excitement was gone, and I was really looking for something to get involved in again. So at that time when I met Dennis, I was actually exploring a few different ideas around where you can apply technology to make the business processes better. I think when we started exploring this idea, I was still continuing with a few of these ideas, but then as the idea crystallized and we saw there was a real opportunity, and then also I think when we raised the funding from Rolf from Trivago, that’s when I stopped all the other opportunities and went full time into this and then started the journey to where we are now.

Steve: And so far, I’m assuming it’s been a fantastic experience.

Leonard: I think for both of us, it’s really, every time you do it, you learn new things, and I think it’s just such a unique opportunity that we have here that we’re excited about the future.

Steve: So you gentlemen started the company in 2017, and a couple of years later, there was a global pandemic. Now, most founders, especially those operating in the travel industry, would’ve looked at the impact of the pandemic and said, boy, maybe we should rethink our strategy and maybe think about a different area to focus on, but you saw it as an opportunity. Tell me a little bit about why.

Dennis: So, actually, the time before the pandemic was quite exciting. 2019 was the first year where we had paying customers, and immediately we signed Fortune 100 companies. So it was quite exciting. So we went into 2020 really with a lot of expectation, a lot of hope. We were at the business travel show in London actually just two weeks before the pandemic started, and you could feel at that show already like, wow, something is happening. People heard some news from China and then some people got worried. But I don’t know, at least on my side, I never really thought that that would impact us so much because what I’ve always believed in is that we need in-person meetings, we need connection, so this will happen. So the pandemic is a temporary thing, and as human beings, we have to or we required to come together more often, and it’s way more complicated now to meet in person. That’s a bit of what we saw.

As the pandemic started, people stopped to travel, people stopped to have meetings, but people were thinking already, okay, how will the future look like and will it be different? There were some additional data points that no one thought about in pre-pandemic, like what entry requirements if you travel to a country, for example. And so those were all functionalities that we were able to bring into the platform. And then, as we were getting a bit out of the pandemic, that was one of the main reasons why we were able to grow the business in that time. So in 2020, we tripled the business, we tripled the customer base, and we continued to grow in 2021 as well.

Steve: Incredible. So when you think about the impact of the pandemic, it seems like that was actually obviously coming out of it, a potential growth vehicle for the company. Is that a fair way of thinking about it? Leonard, what do you think?

Leonard: Yeah, for sure, Steve. I think, as Dennis said, I think the data that we brought in about Covid restrictions, that was a big positive for us. So I think really it complicated the whole travel experience, especially for larger corporates where compliance and safety of the people are important. And I think the other thing that happened is during Covid, their employees couldn’t travel. So the people responsible for the travel program at times actually looked more strategically at the products that they used and their strategy going forward.

But I think a third thing maybe just to mention is that during that first lockdown period, we also called some of our customers and industry players to a strategy session around travel in the future. We’ve included people from the flight industry, hotel industry, and some of our customers, just for a workshop around it. So for everyone, it was so uncertain of where it was going to go that everyone was looking for answers and just people to discuss this with. I think some of the ideas actually came from that session on what created some thoughts on our side of how to move forward with the strategy of the business.

Dennis: And just to add maybe to what Leonard was saying, in pre-pandemic already, there was some remote work happening. Actually, in our case, I have been working remotely since 2016, but it wasn’t that common. But then, obviously, with the pandemic where within a couple of weeks companies had to move from working on-premise into working fully remote, people actually started to enjoy the benefits of it. Obviously, there are some implications to it, but it really showed that, well, as we’re coming out of the pandemic, there are certain things people don’t want to give up anymore. People don’t want to give up that flexibility of deciding, “Hey, do I go to the office, or can I work at home”? There are a lot of things we can do actually virtually, but then there are very key things we can only really achieve when we come together in person.

And one of the key things is building this connection, building culture within a company, which is the backbone of any organization, the connection between people. And I think a lot of companies have realized now that as they’re coming back on their, maybe turning into a hybrid work arrangement that hey, we have to facilitate building these connections between our people, and one way of doing it is by having way more internal in-person meetings.

Steve: So before Troop Travel, how did companies, whether you’re talking about some of the largest companies in the world or even small and medium-sized companies, how did they determine where to meet? And when you think about that now in the context of Troop, what’s the value that they’re getting out of it? Maybe you can give me some examples.

Dennis: So it’s quite interesting that when we talk to potential customers, and we tell them about the value of actually asking yourself where should we meet, that people have never thought about that that it would make sense to ask that question. And then we share some concrete data with them. And I think there’s a very interesting use case we have with a technology company based in San Francisco. They contacted us, and they wanted to do their first all-team meeting after the pandemic — 1,000 people roughly distributed over 76 different places from all over the world. And they said, okay, we want to bring everyone to San Francisco, and they asked to help them by basically letting them know the impact in terms of cost, etc. So we ran it through our technology, and we told them, okay, if you send 1,000 people to San Francisco, it’ll be roughly $2.5 million in cost. But then the real interesting data point no one ever thought about is that it would take the group roughly 14 years of working time — they would spend 14 years of working time on planes traveling to San Francisco and traveling back home. And that event would actually produce 2,500 tons of C02. So we said, hey, if you now do the same meeting in Paris, you would actually be able to reduce the costs by $400,000, reduce the travel time by four years, and avoid 800 tons of CO2.

Steve: That’s incredible. So massive economic savings, massive savings of time, and frankly, lower reproduction of CO2 or greenhouse gases. So one of the things that I find interesting about this market is that unlike some markets where there are existing technologies and cloud-native technologies are delivering oftentimes 10X better value props, here we’re talking about a business need that really wasn’t solved for. Why do you think that need was not solved for before? Obviously, it’s wonderful to be able to go after what effectively is a greenfield opportunity. Why do you think the need wasn’t addressed in the past?

Dennis: One key thing is really as well that technology was actually not able to manage all these different data points bringing together, and I think Leonard can talk to that, how much is really happening in the platform when we run those analyses and as well just simply, people in the past were going to their TMCs, to the travel management companies and ask them hey, we want to do a meeting, can you give us a location? So, in the end, then the people in the agency were doing a manual exercise, which is extremely time-consuming, and extremely difficult to do, and I think there are a couple of components why people haven’t addressed that in the past.

Leonard: Let me just add to that quickly, Steve. I think the one that Dennis mentioned about the technology, that’s a big point. I think it’s a massive amount of data that you need to manage and process and make sense of. And I think it’s just on both the back end and the front end side, the web browsers and the laptops that we use nowadays, it’s much more capable and can process that amount of data much easier where previously it was a challenge. And then secondly is access to the data. We saw that some APIs like Skyscanner, Booking.com, exposed all the APIs that weren’t previously accessible so easily. So I think there are a few factors that came together and just made the opportunity possible, maybe at the time when we started exploring that opportunity.

Steve: Yeah, what’s interesting, even in the legacy model, these are totally disconnected processes. Determining where to go was a very manual process to your question of where do we meet. Then, booking it was effectively re-keying a whole bunch of information into a booking tool, and then there’s no real itinerary management. Talk to me a little bit about what is the vision for Troop’s product suite. It’s obviously more than just answering the simple question of where do you go?

Leonard: Yeah, exactly. I think what we saw initially, our idea was more around finding the best meeting place, which was very much a manual process, and as soon as the meeting got too complex, it just didn’t happen because it was too much data processing manually. So people didn’t do it and they picked a spot, like Dennis mentioned, like Paris, and you just go there again and again, and it’s personal opinions, and I think we filling that piece of the need, we filling with our tool like Troop, but then I think the interesting thing is that when you do that exercise, we said right at the beginning of our process of executive meeting, so it’s exactly what you say, you need to book it, you need to manage it, there are expenses associated, etc, and because we right at the beginning of that process, you can then mold the product out to service the whole journey of a meeting planner and also attendees. So I think it’s just a very interesting position.

I think our strategy is really to be able to develop our product, to be able to cover all aspects of meeting planning, but not only the product or the platform side, I think for us, it’s important to build productivity tools for the ecosystem so that the different players within the system could collaborate. For instance, the person organizing the meeting should be able to collaborate with meeting planners in the city where they want to go and get proposals and collaborate with them around the proposals. Also, the data providers in the industry, like meeting space providers, visa providers, and local transport, we wabt to enable them to become part of this ecosystem through a marketplace that they could integrate into our platform and make their services available to either our customers that are planning the meeting or the local meeting planners that want to do proposals. So I think for us, it’s the platform, the whole journey, but also this whole ecosystem and to enable them all to be more productive and solve the meeting planning process in a better way.

Steve: So, in effect, what you’re doing is you’re building the booking and a full travel experience for meetings as opposed to individual travel, which obviously has been automated by companies like Concur in the past, but there’s really not been anything to automate the group travel experience.

Leonard: Exactly. I think that the group planning process is quite a bit different from individual travel. And I think the complexity for a meeting planner is that traditionally it was different people booking in different places and then collaborating all that and making sure who’s booked, who hasn’t booked, what the status is of the hotel booking, flight bookings, etc. So we enable all that with the future vision of Troop Travel, which just makes the whole process a lot easier.

Steve: And maybe Dennis, you can expand a little bit upon the meeting planner concept that Leonard just mentioned. Certainly, the way I interpret it is that, hey, if we’re picking a location for the group to meet, you want local experts to be able to advise not just on locations to use, but also maybe even coordinate as part of the trip. Is that the right way of thinking about it? And then how do you see that ecosystem expanding?

Dennis: So I think the big challenge when the difference of a meeting versus a normal business travel is that there are a lot of different components which are part of the things you have to consider. It’s not only the flight or the hotel, it’s the meeting room, it’s activities, restaurants, local transport, a lot of different things. So then when you pick or when you look at a hotel option for your meeting, it basically defines the other components around it depending a bit on the size, but let’s say if you have 30 people and you look at a hotel, well, ideally you don’t want to have dinner than on the other side of town because then you have to add local transport. So it’s the connections between these different data points, which really is almost like the secret sauce of the meeting, which makes a meeting successful.

And I think the way we look at local experts is that there are people in every city around the world who are focusing on organizing professional meetings in those cities, and they know exactly. If tomorrow I want to do an executive meeting in Seattle, they will tell me, well, there are really only these three hotels which will work for you and if you take this hotel, then you should take that restaurant and you should take that activity, etc. And I think that’s the fantastic knowledge we are bringing into the technology to actually help you to really make your meeting successful.

Steve: That’s incredible. So you end up getting a far better experience in each location than you could if it was centrally managed through an organization that may or may not know the details of that particular city or that particular town?

Dennis: Yeah. Just maybe to give you a concrete example, we are planning our next ALTI meeting. As I said, we are fully remote, we’re planning our next ALTI meeting for next year, and we start running it through Troop Travel, and we came up with a short list of destinations we want to actually get proposals in. So we shared them with one of our internal meeting planners, and then, she actually immediately said we were looking for example at Zanzibar, there’s one option in Tanzania, and then she said, “Hey, Zanzibar will not necessarily work for your meeting because there’s really only one hotel which can host let’s say a group of 60 people. When you do your next meeting, you might be around 80 people, so you really don’t have accommodation there. And secondly, the internet is very slow there.” So there is this knowledge, these gold nuggets, so to say, of knowledge that are crucial when you plan those meetings because it could really completely change the outcome of your meeting. =

Steve: So look, let’s shift gears a little bit here. So you’re effectively creating a new market or technology for a segment that was largely served through manual processes. One of the things that we’re seeing across other companies is how AI can be used to really aid in the customer service or customer experience part of the process. In the case of Troop Travel, post-booking or the in-trip experience, AI can be used to help. Maybe you can share some examples of how Troop is incorporating AI into that experience.

Leonard: Steve, I think it’s a very interesting position that, again, because we know the data from the planning side, we know the attendees are going to attend, we know what flights they’re going to be on, so I think there we see opportunities to use AI to monitor either the flight or the bookings or other flights, delays, etc., and it can be used there in both ways or two sides of the planning process. One is for the attendees. We could notify them of potential delays and other suggestions like flight options to make sure that the meeting can still go on. On the other side — the meeting planner — I think there are some interesting use cases there as well where they know about all the attendees, but they are not going to monitor it all the time, and I think there’s a service both ways in the offering that we’ve got to use AI around that type of servicing.

Steve: In fact, if you are a large global TMC that’s partnering with Troop Travel, your cost structure to service that group travel experience is, in effect, a lot lower than in today’s model.

Leonard: Yeah, exactly. I think the interesting thing there is because we are integrating a lot of data providers, it could either be flights or hotels or risk that comes up, the duty of care. I think there are so many scenarios where we could make this process more effective and, as you say, reduce cost and the human effort to service that.

Steve: Fantastic. Let’s maybe shift gears for just a second and let’s talk about the team. Dennis, you mentioned that Troop Travel is a completely distributed organization. How do you see the company continuing to evolve and frankly, how do you think about the team and what you need, and how you’ll be able to bring those individuals in in what is a very different model of building a company?

Dennis: Yeah, so over the last, well, 16-17 months, we have grown the team from literally 6 people to 60 now. So we did massive hiring efforts and it was very important for us to really find the right people and they have the right mindset and specifically, the right core values. And I think there are a couple of things which have worked for us. One thing I think is actually quite beneficial for us is that while we are basically a Spanish-South African startup, as I said I’m in Spain, Leonard in South Africa, and I think these kinds of opportunities we have here are not very common in our markets, while in the US we have more, let’s say maybe companies similar like us with this opportunity. So it’s quite interesting for us to capture talent because it is a very unique opportunity. People understand the value of it, and we are able to really get great talent here.

And at the very beginning, I think, we, not on let’s say on the technical side but let’s say on the business side, we were adding more generalists because there are so many different things you have to work on and you have to have this entrepreneurial mindset basically because you have to build it. You have to build the sales motion, we have to get the first customers, we have to figure out how do we manage those customers, etc. And that obviously has changed as we were growing and now, especially after our latest funding round, we really focus on bringing now those experts of those different areas into the company. Now, we have reached a certain level and now we bring these people in who have built sales motions, grown into millions of ARR, built marketing functions, built customer success to really then focus on those areas.

And the biggest thing I think for us is our core values. So a key value of us is trust, and I think that’s crucial. So I don’t want to monitor someone, how many hours are you working because I believe that you have an interest in doing the best for the company because it’s beneficial for you as well. And I think values and culture for us are really, really important, and that’s why we have a very strict recruitment process where we go really into those culture interviews and where people as well sometimes are not able to pass to the next level.

Steve: So your recruiting process is really optimized and designed around a distributed and global workforce. Is that different, do you think, than more traditional settings?

Dennis: I think so. I really think so. Again, not everyone likes to work fully remote and we understand that, obviously, in-person interactions are crucial, but luckily we have our own technology to help us figure out how can we come together. And that’s why once a year, we bring the full company together, but then even throughout the year, they have their smaller meetups. So it’s really a combination of things. It’s the moment when you have this in-person connection, that’s where the magic is happening, but then you go back home, and there’s a lot of things you can obviously do fully remotely, and it allows you actually as well to bring your life and work closer together. I feel it’s almost like we can’t look at it separately anymore. We shouldn’t think about, okay, this is my life, and this is work. No, it’s actually this is your life, and you should spend your time on things you’re excited about. So we are bringing people on board who are really excited about solving this problem we’re trying to solve.

Steve: Yeah, I love that. I love that. Hey, maybe let’s hit a couple more questions, and then we’ll go ahead and wrap up. You’ve obviously raised a few rounds of funding. For folks who are just starting out, as they think about capital raises, what advice would you give them?

Dennis: I always remember when actually we met the first time in person, Steve and I asked you, “Steve, why did you actually invest in us?” And I’ll never forget what you said and you said, “I invested in you guys because I see that you’re obsessed with solving the problem and I know that whenever there’s a challenge, you’re going to find a way to overcome that.” And that’s for me really, I think, the basis, to be honest, of any business. Build a business because you want to change or you want to solve a problem, not because you want to become rich or whatsoever. All these things will come if you solve a problem. That will be the side effects. But if you are not excited about what you’re building, what you’re solving, then you will see another opportunity and then you will follow that opportunity, but in the end, you will not be able to build what you want to build. So I think that’s for me one of the key things and then so as a starting point, the basis.

Secondly, then, work with your customers as closely as possible — it worked for us at the very beginning. We didn’t have any customers, but there were so many people in this industry who were helping us out and just talking to us and giving us input because they were excited about innovation. And all this feedback is very, very interesting, very important. Still today, for us it’s crucial. And our customers, again, are large corporations and they tell us that they laugh, that we are listening to what they say, and that actually when we have now our quarterly business reviews, we are bringing our developers into those meetings to actually listen to the comments of our customers, and people really appreciate that.

Leonard: So I think maybe just to add something to raising money from investors, I think Rolf’s investment really kicked off this journey for us, but I think really the way we raised money from a few different investors and I think for us what is really just unique with the investment from a donor is the active involvement from particularly Steve on the donor side. And I think for us as founders, that’s really incredible value add to the whole journey. But I think, in general, the decisions we make and the context — we’ve got the network … I think that raising money from investors, it’s not only the money that you should look for, it’s also the type of investor or the particular investor group that you raise from is very important in this journey.

Steve: Fantastic, fantastic. When you think about Troop Travel, when you think about it 10 years from now, what is it? What does it look like? What’s a measure of success?

Dennis: I think almost every company in the world has an empty spot available that we should fill because companies like us, small companies, we need our own technology to manage our meetings. So I would love to see that whenever someone thinks about organizing a meeting, they think about Troop Travel.

Leonard: Yeah, I agree with what Dennis said there. Maybe just to add, I think from a product perspective, I think what would be great for us is that, during this value chain, we’ve got the meeting planner, the attendee, the local expert, the data providers, the partners, that all of these players see value in what Troop Travel is and want to work with the Troop product and not be forced, for instance, like a Fortune 500 tell their local experts to work with it. You want all the different players in this ecosystem to love working with the product and to want to use it. And then, yeah, I think what Dennis said is the other side of it.

Steve: I’ll just add one thing to both your comments. So, Dennis, you’re a hundred percent right. What really drew me to both of you gentlemen was your obsession to solve this problem, and having built a company myself in the past, it’s that maniacal focus on solving the problem better than anyone else in the world can solve it. That really takes you through to the realization of that opportunity. That said, the other thing that I was truly amazed by is that this is an area that really hasn’t been addressed, the simplification of how you plan a meeting, how you book it, how the expenses are accounted for, the analytics, and then managing the itinerary all the way through the process and notifying everybody about where things are, where other travelers are, and when they’ll be there and coordinating across an entire group experience. It’s amazing to me that this problem hasn’t been solved.

And so we saw an opportunity, we at Madrona saw an opportunity to partner with two incredible founders to go build a multi-billion dollar company to solve this problem, and to be fair, it’s not so much that hey, we want to go build a multi-billion dollar company, which of course we certainly do, but solving a problem that becomes the fabric of how business is conducted, that’s compelling.

So we’re really honored to be on this journey with both of you and with the 58 other members of the Troop Travel team today. Gentlemen, thank you very much for being a part of Founded & Funded, and I look forward to seeing you in the near future at our next meeting location.

Leonard: Thank you, Steve.

Dennis: Thank you, Steve.

Coral: Thank you for listening to this episode of Founded & Funded. If you’re interested in learning more about Troop Travel, visit trooptravel.com. Thank you again for listening and tune in a couple of weeks for our next episode of Founded & Funded.

Today, Madrona Managing Director S. Somasegar talks with Statsig founder and CEO Vijaye Raji, who spent 20 years at Microsoft and Facebook before launching Statsig as a way to bring the powerful tools he was used to using inside Facebook to all builders.

Statsig experimentation platform and automated AB testing helps companies make product decisions in a data-driven way at scale, which means shipping the right products faster. Vijaye and Soma go way back and take a little trip down memory lane, talking about Microsoft’s Small Basic. But then they dive into what it’s like transitioning from big tech to a founder, the importance of data-driven decision-making, integrating AI into those decisions, identifying the ideal customer profile, and learning how to sell your product. Something that did not come naturally to Vijaye. He and Soma talk about all of this and so much more.

The 2023 Intelligent Applications Summit is happening on October 10th and 11th. If you’re interested, request an invite here.

This transcript was automatically generated and edited for clarity.

Soma: Hello, everybody. I’m Soma, one of the managing directors here at Madrona, and today, I’m very excited to have Vijaye Raji, the founder and CEO of Statsig, here with me. Before I get started, I want to sort of go down the memory lane a little bit. I remember meeting Vijaye for the first time when both of us were at Microsoft. He was working on this tool called Small Basic, and he was really passionate and focused on figuring out like, Hey, how do I make programming a whole lot easier and simpler to get started for the next generation of developers? And since then, throughout whatever Vijaye has done, there’s always been a focus on how do I make developers and development teams successful and effective, and how do I make their jobs easier, and make them more agile in the process. Vijaye, you were at Microsoft for a while and then you went to Meta and now at Statsig. Can you shed light a little bit on the journey that you’ve taken so far leading up to Statsig?

Vijaye Raji: Yeah, absolutely, Soma. And thanks for having me on your podcast. Really excited to be here. You touched on Microsoft’s Small Basic, which is something near and dear to my heart, primarily because I was so passionate about building a tool for kids and early programmers to get their programming teeth cut. And particularly so right now because my 10-year-old son is also learning programming right now, and he’s using Small Basic to program, and it’s really awesome because on the weekends, we sit down and we jam together, come up with games, and then he codes it up and then shows it off. So really, really relevant. Thanks for bringing that up.

Yeah, so we met, I think, sometime in 2005 or so. I was working on Small Basic and then we were talking about how you wrote the first blog post of releasing Small Basic from Microsoft. I remember that. Thank you for doing that. I’ve always been passionate about the developer space and how we can get more productive tools out to the developers. And then, in 2011, I left Microsoft when I was working on Windows, actually developer tools on Windows 8. I joined Facebook, which at that time was still called a startup when it was only about 1300 people. Then, as an engineer, I continued to build various different products and, throughout the process, was also enamored with the tools that Facebook had invested in that helped all the engineers and the product builders inside Facebook to build and release products really fast and precisely.

And those are all the formative years. And I think about what I learned there, what perspective I gained there, and what I took away from that all led me to leave Facebook in 2021 to start Statsig, which is basically a culmination of the tools that I was so excited about that I wanted to bring and build for everyone outside. And so that’s the journey of back in the day, as you remember, Microsoft, Small Basic all the way down to Statsig.

Soma: Thanks Vijaye for sharing that. We at Madrona have been talking about intelligent applications for many, many, many years now. In fact, I would go so forth to say that we are strong believers in saying that every application that gets built today is an intelligent application. And by intelligent application, what I mean is it takes into account the data that the application has access to and uses AI and ML in some fundamental ways to be able to build a continuous learning system that, in turn, helps you deliver a better service to your customers.

So if I look at the world through that lens, I look at Statsig and say, Hey, Statsig is a classic example of what I would call an intelligent application enabler that allows product teams to experiment with different features and ensure that they create best-in-class experiences for their end users. But before we talk about Statsig and what Statsig is doing, particularly in the world of generative AI today, which you can’t have a conversation anymore without talking about generative AI in some way, shape, or form. Tell me the origin story of Statsig. How did you come up with this idea and how did you get together with your founding team?

Vijaye Raji: So if you trace back software development, all the way back to the ’80s and the ’90s, a lot of what the development process that led to software being shipped came from hardware. So you would have this waterfall model where you’ll go talk to your customers, come up with a set of requirements, and then that goes into a design document. The design document gets reviewed and then engineers pick up that design document, turn it into an architecture document, and then write code. And then, the coding is done during milestones, which will then be packaged and sent off to the QA team. The QA team will run the test and then validate it, it’ll get released, it’ll go to, I don’t know, back in the day, best Buy Circuit City and sit in boxes, and people will go buy that and take the CD and insert it.

And that’s how software development happened. Then V2 will happen, V3 will happen. And then over time, with the internet, a lot of that process got faster and faster. So people got into this build measure learn loop where the continuous deployment of new features enabled a lot of this fast iteration. And along with that came the data aspect of it. So, where you were instrumenting your product, your understanding of what users were using, which products or features they did not care about, and how much they used. And then it subsequently led to this really fast iteration that also led to this experimentation-based software development, which is, I think, the latest trend, where every product that you build, every feature that you release to your users, you want to understand the impact of that. Without understanding the impact of it, you don’t always get your product intuition right.

And that is an interesting progression over the last couple of decades. And so, I was in the middle of all of that when I went from Microsoft, which at that time in Windows was primarily following a waterfall model. When I went to Facebook, in my first six weeks of boot camp, I got to release a bunch of features that actually got to reach 600 million people at that time within days. And that was fascinating. First of all, I have to say that when I first got that experience, it was a little scary. I thought something was going to break all the time, and just knowing all my software engineering, this is not how software should be built. But over time, I started to understand and appreciate that kind of fast iteration and when I dug deeper, the tools that were underlying all of that software development process were fundamental to keeping everything going in the right direction.

They were fundamental in capturing data. They were also instrumental in understanding the impact of every single feature. Now what was interesting was that the tools shape the culture that came down. So the downstream culture of product development where everyone felt empowered to look at the data and make product decisions, both good and bad. If the numbers were not looking good, they decided to cut the feature. And there was no personal attachment. And so, the downstream culture was one of the very objective data-oriented decision-making, which was distributed among all teams.

That was fascinating. That was really, really powerful. And so the journey of Statsig is basically that so how do we bring that kind of a product-building culture to everyone? When I looked outside from Facebook, not every company has the luxury of building these tools or the team sizes or the engineers to go build all of these tools by themselves. So that was the opportunity that I thought would be useful to go and solve for the world. The mission really is like how do we improve software development using data-driven decision-making and bringing that to the masses.

So in 2021, I left Facebook. At that time, I was the head of Seattle, head of the entertainment division at Facebook, and I felt so strongly that this was an important problem to go solve, and so did seven other people; they all came with me on day one, we started building, and since then we’ve been building. It’s been two and a half years. It’s been a fascinating journey. I’ve learned a lot. So happy to share.

Soma: That’s fantastic, Vijaye. Hearing you tell that story reminded me of why I was so excited to be a part of the developer division at Microsoft for many years. Because when you end up building products for people like yourselves, there is a special excitement around that. And I always wondered for the last many years, as companies and developers have been building what I call analytic tools and AI tools for the rest of the world, I was wondering when are developers going to start building these tools for themselves. And I look at Statsig and Statsig is doing just that, right? Building tools for people like us who are building next-generation products, and that’s a whole lot of fun. But Vijaye, as you mentioned earlier, you were at Facebook for 10 years and at Microsoft for 10 years prior to that. Having worked at two large companies, two large technology companies, because though when you started Facebook, it might have still been called a startup, by the time you left, it was a fairly large company. But what was it that made you decide to leave a large company environment and move to start an 8-person startup? And related to that, what experiences, whether it’s the experience that you got at Microsoft or the experience that you got at Facebook, do you feel prepared you to launch your own company?

Vijaye Raji: I have to say that anyone who is starting a startup from scratch has to be a little bit crazy. Some amount of irrational decision-making has to be there because, look, if you are completely rational, every formula, every expected outcome that you calculate should indicate that you go join a large company. The only reason why you would start a company from scratch is something you feel so strongly about solving a particular problem in the world so that you’re willing to leave behind all the comforts and go do this thing.

So for me, that happened about three or four years before I started Statsig. For me, the journey was more around what are the necessary conditions for me to be able to be successful in a startup environment. What kind of skills would I need? Am I good at hiring people, coaching people, and growing people? Am I good at understanding business, the profit and loss for a business, how we sell, and how to market the product? Am I good at finding product market fit for a new idea? Am I good at getting it to market and actually having people understand the problems that the product is solving?

So those are all the kinds of criteria in my mind, I was like, okay, well, how do I learn all of that? And so the last three years that I spent at Facebook was basically figuring out or putting myself in positions where I would learn a bunch of that stuff. And then finally comes the point of, okay, who would want to leave great jobs and want to pursue this journey? A great entrepreneur would have to have a set of followers, people who are crazy enough to follow an entrepreneur who’s about to jump off a cliff. And so that’s the image that you want to think about. And if you can convince a set of folks to come and join you, that is already a validation. And those are the kinds of stuff that I prepared myself for.

Now, there’s another aspect of it — I am married, and we have two kids, a 10-year-old and a 7-year-old. Family is every single bit involved in your startup journey as you are. And so a lot of the commitment has to also come from the family side. So if I think about my wife and my kids — every day, we talk about the startup. I feel like they’re also bought into the idea of this startup — Statsig. Without their help, I wouldn’t be here.

Soma: So Vijaye, I know that you’ve been asked this question a ton by other people, whether it’s from Facebook or from other companies, but if I have to take a step back and ask you, Hey, what advice do you have for someone who’s thinking of maybe making a similar transition to what you did from a larger company environment to move off and say, Hey, I’m going to launch my own startup. Are there any pulls of wisdom or key learnings that you want to share?

Vijaye Raji: So I’ve only done this one time, and so take this with a huge grain of salt because there are a lot of biases here. The one thing that has helped me quite a bit is once I understood that my next journey is going to be a startup, it really comes down to what kind of skills can I pick up. What kind of connections can I create now? What kind of relationships could I build right now? And then what kind of team can I bring on day one to the startup? And if you walk backward from it, then it becomes very clear what all should be done in order for the startup to just get up and going as quickly as you can and start building.

So most companies won’t let you write code for your startup as you’re continuing to work in the other company, but you can always think about the problem. You can talk to other people about the problem. So there’s a lot of validation that you can do even before you start the startup, even before you embark on the journey. And so, that should all inform you on what is day 1, day 90, and day 180 going to look like.

And then the other one that most people don’t really talk about is your financial safety net. Because when you’re going into a startup, you’re kind of committing several years of your life to the startup, and during those periods, there’s going to be a lot of instability and a lot of ups and downs, primarily on the financial side. And if you have a family, you want to make sure that you build a good financial nest that supports you during that journey. And so those are all necessary components as you think about starting a startup. For me, it took me three years or so to ready myself to leave behind a comfortable job and get into Statsig. And sometimes, it could take longer.

Soma: In looking back at your origin story for Statsig, I think you were lucky enough to start with a founding team of eight people, all of whom you had known from before as you all had worked together in one way, shape, or form at Facebook. But some people have the luxury, some people don’t have the luxury, but every entrepreneur thinks about, Hey, I want to make sure that the founding team is the best team that I can put together, and I need to start thinking about culture from day one or all that fun stuff.

From your vantage point, how did you decide, hey, what your founding team should be and then, more importantly, what kind of a culture you want to set? On the one hand, you all had what I call the fortune to work together in a company like Meta with a particular culture, but here you are starting off on your own, and I’m sure you had some ideas of what you wanted the culture to be, and what you want to prioritize and what you wanted to be below the line. Talk to us a little bit about how you found the right people to instill the right set of culture from day one.

Vijaye Raji: I think culture is a very, very important aspect of a company, especially as you’re starting from scratch, because there’s only a few things that you can hold strongly as the company grows and as you add more and more people. So you want to hold onto the things that you really care about. And those could come in the form of what is the environment that you want to build, that you want to continue working on 10 years from now? And if you put that in that kind of lens, then that becomes really clear. What I’m doing is building a place that becomes exciting for me to come in every single day and work every day.

So early on, we set out to pick a set of values, and when we said we were going to pick values, we decided we’re going to pick a handful of values. Those are the ones that we really care about that we don’t want to compromise ever on. You don’t want to have 15 or 20 values because nobody can even remember that many. And so we picked four, and even when it came to the set of values, we decided these have to be trade-offs. These cannot be truisms like motherhood and apple pie. So you can’t say things like, oh, be honest or be transparent or be happy or be friendly because those things are all given — those are all table stakes. A value should have a downside and a trade-off. Every day you should use them in debates and conversations and decision-making processes, thereby doing the right trade-offs, and that sets the tone for the whole company.

I’ll give you one example. We have one value that says, “No sacred cats.” No sacred cats is actually a way to say, do not take anything for a given unless it’s logical, unless it’s reasonable. If you have a set of processes or traditions that people are following, and if nobody can explain why, then it’s okay to change it. But the trade-off here is there’s a lot of overhead in changing things, but the outcome is that you will never end up in a position where things are illogical. Nobody would ever come in, like why the heck are you doing this thing where it doesn’t make any sense? So that’s the trade-off. And so those are the kinds of values we ended up picking early on.

And I remember having exercises with the early teams of seven or eight people where we all sat down together, we all agreed, and these are the values that we’re going to adhere to the company, and we have continued to maintain that.

Now over time, that becomes your culture, that becomes your identity, and when you hire more people, they opt into the culture, and so you want to make sure that something that is attracting people or attracting good people. So yeah, I think those are all important stuff. Like I said, you want to pick a handful of things that you can hold onto because every time the company grows, culture evolves, culture shifts, and if you try to control too much, it doesn’t work.

Soma: If I take a look at the progress that you made, first of all, Vijaye, you guys have made some great progress in the last two and a half years that you’ve been around. More than anything else now, I continue to be impressed with the velocity of product development. It’s one of these things where you’re building a set of tools to enable the rest of the world to do that. And by virtue of that, being able to be in a position where I’m leading by example is a fantastic place to be, and that warms my heart when I think about Statsig. But in the last couple of years, you’ve sort of gone through a phase of learning about what does it mean to go identify the right customer, or people call it, ICP or the ideal customer profile. You now have what I call small companies and large enterprises as your customers. You’ve got companies that are delivering consumer-facing services. You’ve got companies that are delivering B2B services or enterprise-facing services. So you’ve got a wide array of customers today.

Can you share a little bit of your perspective on how has your thinking evolved in terms of who’s your ideal customer and anything that you’ve learned in the process of deciding, Hey, this one works well, that one doesn’t work well? How do you define yourself in terms of going after customers and landing customers for Statsig?

Vijaye Raji: This is a really, really good question because it was a journey. It wasn’t intuitive or it wasn’t very clear on day one. I come from an engineering background, and I had lots of exposure to product roles, but I had very little exposure to sales or marketing. So I don’t have intuition for sales or marketing, something that I had to learn during Statsig in the early days and something that I actually made a lot of mistakes in the early days of trying to sell a product and then learning how not to sell. So when it comes to product building, now you said the velocity, we take pride in the fact that we help other companies build fast, and it is important for us to actually be an example for that. If we cannot even build fast, then we can’t claim that we can help other people build fast.

So a part of that is what are the set of conditions that would make engineers, product managers and designers and all the creative people blossom and flourish? And so removing all kinds of friction, removing overhead, removing ability, one of the key elements is transparency, radical transparency. Everything is available for everyone. All the code is transparent, all the data points are transparent, so people don’t have to get stuck in the process in order to move fast. So creative people should be allowed to move as fast as they can. And so we’ve established an environment where product building can happen really fast. Now what happens after you build the product, you have to go and convince a set of customers to try and use the product. That took a long time.

So I think the first eight months, we couldn’t even convince people to use our product for free. How do we sell this product? We know there’s value in the product. We believe very strongly that this is an important platform that others should use, but we were not having very much success convincing people to try and use it. One day, we saw an ex-Facebook engineer start using our product and started sending a lot of events, and those events came from Singapore, Malaysia, and Indonesia. At first, we thought, oh, somebody’s DDoSing us from various different countries by sending lots and lots of events. And then later on, we realized nobody cares about Statsig, so they probably are not DDoSing. And then we figured out it’s the next Facebook engineer that is using the product.

That was when it hit us like, okay, there’s a set of people that left Facebook that all had access to the tools inside Facebook. They are probably all misusing those tools, and that should be the people that we should be talking to. Then we started talking to, I think Headspace was our first customer. We talked to an ex-Facebook product manager at Headspace, and it resonated very, very strongly, much stronger than what we were trying to do before.

And so that was our first ICP, really not the big companies or the small companies and not the stage of the company or what series they’re funded. It was really like the ex-Facebook folks that missed the tools that they had access to. And that’s how we figured our way, almost like in a stumbled out on our ICP. And then since then, we’ve had plenty of traction, and that was, for me, a pivotal moment in understanding the value of our own product.

Soma: That’s super, Vijaye. Two and a half years since you started the company, since you became an entrepreneur since you sort of went from zero to wherever you are today, I’m sure not everything has been fine and everything going well. Can you sort of highlight one or two unexpected challenges that you face? The things that you did not expect that you said, oh my God, what is this new thing? And more importantly, how did you deal with them?

Vijaye Raji: Every single day is a learning. There are hundreds of things that happen during the day, and for me, there are a bunch of new things that I learn and take away. If I were to think about the big ones, understanding how sales teams operate was key for me to be able to unlock. So one of the earliest hires we made, Sam, who came from Segment, who leads our sales — he’s the head of sales for us. He knows sales like I don’t. I learn a lot from Sam every single day. The way he thinks about problems. The way he thinks about customer’s problems. This is one of the things where founders make a mistake — customers don’t care about your vision. They care about their problems.

And so when I sit down and talk about it, I get so excited about sharing Statsig’s vision to this poor customer who’s like, I have a problem to solve and I’m not here to listen to your vision. And then you hear how Sam talks to the customer and there’s so much to be learned from there. And so that was a pretty large area, I would say, a huge area of figuring out, slowly making some mistakes, and then learning from the people around me who are so much better than me at what they do.

Then the next one was marketing. That’s another one of those things. So our tool is used generally by data scientists and engineers, and those are the two groups of people that don’t want to be sold to, don’t want to be marketed to, and how do you make sure that they understand the value of Statsig? And so the way you approach that is through communities or content where you actually make our product useful for them and then have them pick it up, learn what they can do with it, versus actually being out there and shouting.

And so those are nuances that I had to learn about marketing. And then here’s another one. It’s so naive at this point when I think about this, when everyone says marketing, they combine about 10 different types of marketing in one bucket. That’s another one of those revelations like, oh, there’s content marketing, performance marketing, demand gen marketing, product marketing, events marketing. Whenever I bucket everything into marketing, it actually turns out, there are like 10 different ways of marketing stuff. Even that aspect I didn’t know, and I had to learn. And so once I hire the marketing folks around me, now they’re doing a great job and I learn from them every single day. So yeah, it’s like every day is a day of learning and it’s been great.

Soma: That’s awesome, Vijaye. You’ve been working in the technology industry for a couple of decades now. As you know, we are in the midst of a huge platform wave. People call it AI, people call it generative AI, and there seems to be an announcement about a new generative model, a new AI application, or a new piece of innovation, literally every other day. It’s almost like a tsunami is over us in terms of the rate of innovation as far as generative AI goes. With the speed the market is evolving, how do you decide or how do you know what is the right set of features that you want to prioritize and focus on? And then, more importantly, what would you tell your customers? How should they prioritize?

Vijaye Raji: One of the things that Statsig was fortunate about is we built a set of tools, the primitives of experimentation, that was applicable for AI tools. So we have some of the biggest AI companies that use Statsig for optimizing their user experience that are based on AI. So the application of AI, when you present the experience to your users, you want to understand if it’s actually improving and if it is improving by how much that impacts quantification. And so, Statsig became a tool for AI producers and AI consumers to validate some of the AI applications. So when they bring a new model, when they bring new prompts, or when they tweak the parameters like temperature or frequency penalty, they use Statsig to validate whether this is positive or detrimental to their product. And so we found ourselves in that space and we started doubling down on it.

So we have partnered with companies like LangChain, where we provide tooling for everyone using AI for their product to just start running experiments. In the world of experiments for AI/ML, there are offline experiments and then there are online experiments. In offline experiments, you take a set of training data and you verify and you validate, okay, I’ve gotten it to a pretty good place. Now I want to put it in production. But production is an entirely different story. How the model operates at scale, and how it actually improves your business-level metrics is something that you need to measure, that you need to validate, and that’s where Statsig’s abilities are really shining.

And then the second aspect of that is like, okay, how can Statsig help using AI within our own product to maybe even bring a lot of insights to the customer? So, for example, one of the things that the Statsig experimentation platform provides is the new feature that you launched yesterday — maybe it’s a registration feature that is improving your conversion metrics by 4%. With AI/Ml, it’ll be easy for us to actually even dive deeper and tell you, okay, on average, you’re seeing a 4% improvement in your conversion. However, on Android phones that are in non-English speaking countries, you’re seeing a 20% drop. And that’s a question that you wouldn’t even ask in the beginning.

So, if AI is able to identify these anomalies and patterns and bring it back to you and provide insights, you can go fix that problem or the bug that is affecting all the Android, non-English speaking phones. Then after that, your 4% improvement can actually become a 5% improvement, which is a huge, huge win. And that’s the area that we’re kind of starting to investigate and invest in.

Soma: When I think about Statsig, Vijaye, I think about Statsig as sort of a product experimentation platform, a product observability platform, but really something that is going to be helping you build the right set of features and the right set of capabilities irrespective of whether you are building a generative AI application, you’re building a large language model, or you’re building what I call an intelligent application kind thing. So I think about Statsig as, Hey, you are an enabler for generative AI. Is there a good way to think about it, or do you think about how you position Statsig in a world of generative AI differently than that?

Vijaye Raji: So anybody who is using AI or ML today will need a set of tools to understand the impact of that. And so Statsig being an enabler of, in general, better products, better features, or even knowing which of the features or the products that we launched to the customers is doing well. So that, to me, is an enabler of better product building. So one of the things that at Facebook, at least we understood was unless you’re a Steve Jobs, you’re going to be about 33% correct in your product intuitions. It’s kind of like a humbling fact and brings a lot of humility in, like when we have product intuitions, it’s been validated 33% of the time you’re going to be absolutely right and about 33% of the time you’re neutral, and then the rest of the time you’re actually hurting the product. You may not think that is the case, but the data shows that.

But just by knowing which 33% of the products are actually hurting your company’s business metrics, and by turning those off, you’re going to get a bump up, but being able to know that is how you’re going to first start by measuring and then understanding and then actually removing your personal attachment to those features. Those are all the ways in which you end up building better and better features, and Statsig is one of those enablers of building better features.

Soma: In my conversations with you, particularly about generative AI and talking about large language morals and the like, at least it feels to me that hey, you have a little bit of a contrarian view on what many people think of as a hallucination problem that exists with generative models and large language models. Can you share your thoughts on that or your perspective on that? Because I think it’s a fascinating line of thinking.

Vijaye Raji: Yeah, absolutely. At first, it started out as, like, let me take a contrarian view, and then I started doing some research and then I started actually believing in it. When people talk about hallucinations as a big problem, I think it’s a little bit overblown and it’s actually not that bad of a problem. In fact, if you think about it, it’s a creative way that whenever there’s stuff that is made up when you give two inputs and then there’s stuff that is filled, the gap is filled. That’s what we do all the time. Humans do this all the time, and our brain does this all the time. When we do it, we call it creativity. So when AI does it, we call it hallucination, which is, over time, it’s going to get better and better.

If I already think about how AI is generating art, how AI is generating music, those are non-deterministic. So every time you run the same model, you don’t get the same output, you don’t get the same image, you don’t get the same music. It’s a little bit different. I think it’s a form of creativity, and a lot of times what happens is these creative ways, sometimes based on if we think about evolutionary practices — over time, these mutations start to get better results. And so if those are able to be absorbed back into the system and the system gets overall better, I think there’s a potential upside to all of these hallucinations. And so that’s my take on it. We’ll see where things go. Hallucinations are not that bad, and over time, they will get better. That’s my feeling.

Soma: No, I agree that they’re going to get better over time. I know that we are coming up on time here, Vijaye, but before we wrap up, I want to ask you about your own usage of generative AI, whether it is for personal reasons or for professional reasons. Do you use generative AI on a day-to-day basis? And if so, what do you use or how do you use that?

Vijaye Raji: I use it all the time now. In fact, if there are no temporal data points, I prefer using AI to ask questions, and over going and searching on Google or Bing. Just last week, one of the things that I was doing is I was trying to remember the name of this electrical thing and I couldn’t express what that is. And so I went to OpenAI ChatGPT and started typing like, Hey, look, I’m looking at this thing. It’s an aluminum box. It’s rectangular in shape, and it’s got conduits coming into it. I see electrical wires. I don’t know what it’s called. What is it called? And it said, “Oh, it’s a junction box.” Oh, thank you. And then I went to Home Depot. I searched for Junction Box and I found it and I bought it. So I think it’s really cool. It’s just crept into everyday usage. I love it.

Soma: Vijaye, thank you so much for spending the time with us today on this podcast session. I do want to take this opportunity to congratulate you again on all the success that you’ve seen so far, particularly the progress that you and the Statsig team have made in the last couple of years. And looking forward to seeing what is possible and potential with what you’re building. Thank you so much.

Vijaye Raji: Thanks, Soma. Thanks for having me here.

Coral: Thank you for listening to this week’s episode of Founded & Funded. If you want to learn more about Statsig, visit Statsig.com. If you’re interested in attending our intelligent applications summit, visit ia40.com/summit to request an invite. Thank you again for listening to Founded & Funded, please rate and review us wherever you get your podcasts, and tune in in a couple of weeks for a conversation with Troop Travel’s Co-founders

In this week’s episode of Founded & Funded, Madrona Managing Director Matt McIlwain hosts NYSE Group President Lynn Martin ahead of our Intelligent Applications Summit on October 10th and 11th. Lynn will be speaking at the Summit, but we thought it would be great to have her on the show to talk more about her background, the NYSE, capital markets, IPO trends, and, of course, the role data, AI, and large language models are playing in the markets and in companies broadly! This conversation couldn’t come at a better time now that we’re seeing a number of tech companies teeing up IPO plans.

Lynn thinks we’ll be back to a more “normal” IPO environment in 2024 – but you have to listen to get her full take!

The 2023 Intelligent Applications Summit is happening on October 10th and 11th. If you’re interested, request an invite here.

This transcript was automatically generated and edited for clarity.

Matt: I’m Matt McIlwain, and I’m a managing director here at the Madrona Venture Group. I’m delighted to have Lynn Martin, who’s the president of the NYSE Group, which includes the New York Stock Exchange, and she’s also Chair of ICE Fixed Income and Data Services. Lynn, you started your career coding, you were a computer programmer back at IBM out of college, and now you’re leading the world’s largest global stock exchange and a bunch of related businesses. Before we dive into things like the capital markets and the businesses that you run, and of course, the topic of AI that’s on everybody’s mind, I’d love to jump into how you get from programming at IBM to the financial markets?

Lynn: So first, Matt, thanks so much for having me. It’s always good to spend time with you — great question as to how I wound up here. As you pointed out, I was a programmer. I was trained as a programmer, I have an undergraduate degree in computer science, and it was around the time of the dot-com revolution, and markets, in particular, were starting to have technology integrated into them.

I was at IBM, and I was really looking for something that spoke to my passions. My passions at the time were financial markets, financial market infrastructure, and the math that underpinned financial markets. It’s what drove me to get my master’s degree in statistics. I happened upon an opportunity in 2001 with a derivatives exchange, which was based in London called London Financial Futures Exchange. They had a tremendous amount of people in their organization that could speak about the products that they had listed, so really European short-term interest rate debt and equity index futures, but they didn’t have anyone who could talk to the new breed of trader, which was people who were writing to the API as markets started to go electronic.

I interviewed with them, and what really interested them about me was the fact that I was a programmer. I had conversational knowledge of some of the models and some of the products, but I could talk to a programmer who was writing to this thing called an API at the time.

Matt: Yeah, before APIs were cool.

Lynn: Exactly.

Matt: Yes.

Lynn: At the time, the whole management team of that exchange was like, “What’s an API?” So the right place and the right time, and the interesting skillset that led me to financial markets back in 2001.

Matt: No, that’s fantastic. To fast-forward a little bit, you ended up at NYSE Group and have had successive opportunities and broadening roles. I don’t think most people really have an appreciation for the breadth of the NYSE Group and then the related relationship with ICE. Can you paint that broader picture, and then we can talk specifically about the areas that you have responsibility for?

Lynn: Yeah, I’ll fast-forward through a little bit of history. In 2012, I was with NYSE. I was CEO of a futures exchange that NYSE was a 51% owner of, and 49% was owned by a variety of institutions on the street, and a company called ICE came along and acquired the New York Stock Exchange. ICE was an electronic group of exchanges, a company that was very much driven by the implementation of cutting-edge technology and the application of that technology to financial markets. That acquisition was completed in 2013, and I found myself in a very interesting position in that I was in a directly overlapping business unit.

I wound up taking a position being offered and then taking a position to run one of their clearing houses. I got to know the founder of ICE, a gentleman by the name of Jeff Sprecher, who is still our chair and CEO of ICE, who really got to know my background both from an academic standpoint and a business standpoint. He’s the type of guy who’s always thinking about two years, three years, five years down the road, and he’s been a great mentor to me in that regard.

At the time, he was thinking about, “Okay, a really important output of financial markets and equally an important input to the most liquid financial markets is data.” And he was already thinking at that time, “There’s something around data that I want this ICE group to be a part of.” It led us to form ICE Data Services in 2015, which I was named president of, and it rapidly grew into ICE’s largest single business unit through a variety of acquisitions. The way you should think about that group is pretty much what I just said. It’s all down to the premise that the most liquid markets have a tremendous amount of data as outputs. In order to make a market — and a market can be broadly defined — liquid and actionable, you need data to form what value is going to be.

That can apply to everything from the U.S. equity markets, it can apply to fixed income markets, it can apply to a lot less traditional markets such as where ICE Group is now entering the next generation of its evolution, the mortgage market, the U.S. mortgage market, and how we use data to make a more informed decision or allow for a more informed decision in the U.S. mortgage market.

Matt: You chair this whole group, the ICE Fixed Income and Data Services. Can you give us one tangible example of the more popular data services you provide today and how that works? We’ll come back later to talk about how AI is changing or giving you opportunities to enhance some of these areas.

Lynn: AI plays an incredibly important role in the Fixed Income and Data Services vertical that we have at ICE. You can take AI and large language models from an efficiency standpoint. How do you do more with less bodies? How do you cover a broader universe, as I like to think of in the fixed-income markets? In the fixed income markets, that manifests itself in us providing valuations for 2.7 million securities globally and also providing terms and conditions data, the underlying aspects of a bond where when it matures, what its coupon rate is, all those sorts of good things on 35 million securities globally.

We do that through the implementation of natural language processing, large language models, and all different forms of artificial intelligence. Still, importantly, we have a strong human overlay that knows if what those models are producing is good information or bad information. You don’t want the erroneous information to continue to perpetuate through a system because then that winds up polluting a system.

Matt: You go from a virtuous to a vicious cycle, and these capital markets are dynamic. Before we dive more into some ways you use ML and AI, what is the state of the capital markets today? I’m thinking about the public and the fixed-income markets generally, and how do you think about that market space, the context beyond all the services that you provide?

IPO Trends

Lynn: They’re operating as you would expect them to operate. So 2020, and 2021, if you think about the equity markets, you had a tremendous amount of IPOs coming out because the system was flush with cash, and people were taking advantage of high valuations. And then what you saw in 2022, early 2022, you saw volatility start to creep in, volatility instigated by multiple factors, the war in Ukraine, rising interest rates, monetary policy, how do we pay back some of that money that’s been injected into the system? That caused the IPO markets, the people who were tapping the U.S. equity markets, to grind to a halt, as we saw in 2022.

As we moved into 2023, you’re now starting to see the effects of the Fed implementation of rising interest rates start to work its way through the system. So effectively, you’re starting to see inflation come down, and you’re starting to see the desired effects that the Fed had with its monetary policy start to take hold. The way we’ve started to see that manifest itself in our markets is number one: the volatility has come way down. The key barometer that I always look at is the VIX. The VIX has been 13, 15, somewhere around there, definitely below 20 for a sustained period of time.

That’s really important for a company that’s thinking, “Okay, is it time for me to tap the public markets from a capital perspective?” What you’ve now seen is, at least in Q2, we saw the IPO markets start to reopen. So you saw Kenvue, which was a spin-out of Johnson and Johnson, raise a $3.8 billion price at the high end of the range upsize. You saw CAVA, you saw Savers, you saw Fidelis, you saw Kodiak more towards the end of Q2. What encouraged me about everything we saw in Q2 is you saw both U.S. domestic deals get done, you saw international deals get done, and you saw multiple sectors come to market, and that’s the first time that you’ve really seen a good healthy cross-section of activity tap the public markets since 2021, certainly.

Matt: Well, I’m sensing some cautious optimism in the return of the IPO market. It’s been a year and a half since Samsara listed on the NYSE back in December of 2021 — since you’ve really had a technology IPO and the sense of how we think about it in the tech venture business, how are you seeing the end of this back half of this year in terms of technology IPOs and then also the trade-off between the specific situation that some of these IPOs might be at a lower price in the public markets than the last private round?

Lynn: I think the investors have gotten their heads around the fact that the valuations are going to be different from 2021. 2021 was a bit of an anomaly in terms of the valuations. Very optimistic that we’re going to start to see deals come out in the second half, particularly in technology. Ultimately someone’s got to test the waters, though, so there’ll be a couple of deals that will probably come out in the second half of 2023, which are going to do that. And we’re working with a variety of companies in our backlog who are now saying, “Yeah, we’re going to go in probably early Q4 timeframe and there may actually be one or two that go prior to that.”

Very optimistic that those deals are going to come out, those deals are going to get done, and then that’s going to really set the stage for 2024, which is when I think we’re going to be back to a more normal IPO environment.

Matt: That’s encouraging. With that in mind, we work with private companies at all different stages. Some of them are getting close to being IPO-ready. What kind of advice do you and your teams give to these kinds of private, rapidly growing companies about what it means to actually be IPO-ready?

Lynn: Yeah, and I use the term IPO-ready a lot. The bit that 2021 showed was that 2021 was giving value to growth at all costs. It was for every dollar of growth, the market was rewarding you. The companies that are coming out now have balanced growth as well as profitability and how to not sacrifice growth but also come out with some discipline on the expense side. Ultimately, an investor’s going to ask you about expenses during your earnings call, they’re going to ask you about growth during an earnings call. I think being already of that mindset that you’re balancing growth with profitability is a good philosophy to have going into your IPO because if you’re a company that’s a year old, five years old, 10 years old, 100 years old, you’re going to get asked those questions on your earnings call.

Matt: It’s a tricky balance, and things got a little out of kilter in the 2020 back half of 2021. I agree that we’re coming back to more of a balanced outlook there. Maybe just turn just a little bit, you’ve got this whole area of fixed income you’re responsible for. Just give us a view of the fixed-income markets. Treasury prices have been volatile. I think they’ve stayed more volatile in general here and are trending downward. The rates on the ten years have popped up a good bit again. What’s your outlook on the fixed income side, and how does that influence the timing and this return to the IPO market?

Lynn: It’s all ultimately tied together. When the equity markets are a bit more volatile, there could be a flight to quality, and it could be more of the treasuries that people invest in. We’ve seen a lot of volatility not just in treasuries, we’ve seen it also in the muni markets, the U.S. municipal markets. Also, investment grade and high-yield corporate debt have been bouncing around a bit as well. The treasury market, though, has been incredibly volatile. Typically when the treasury market is volatile, the muni markets are volatile too because those two trade at a spread to each other. I think to the extent that there is continuing to be treasury market volatility, you’ll see it manifest itself across those two asset classes in particular.

Matt: There’s just so much input data and output data in all these different capital markets and related data services areas. Tell us about the groups that you are responsible for and how you guide and direct them around what more can we do with this data? What can we do with AI in general, generative AI in particular?

Lynn: One of the things that I’ve always been focused on the data side is observable data. Good data is gold, and bad data, when it gets into a system, is really challenging to expunge, so I tend to focus on how we harness the good data and build the right models off of that good data to extrapolate other pieces of information. One of the areas that we’ve been super excited about is how we can take data and how we can take these large language models and add efficiency to the trader’s workflow.

One of the big mantras I have is to always stay close to your customers and talk to your customers about what their pain points are, so one of the pieces of functionality that we’re incredibly proud of that we built almost a decade ago in our data services business is our ICE Chat platform. That chat platform connects the entirety of the energy markets ecosystem, everything from a drilling company to a producer, to an investor, and even to people who carry the cargo from different ports for delivery. Within that chat platform, we had developed a proprietary large language model, which knows if I’m talking to you about what day of the week today is, or it knows if I’m talking to you about the price of crude oil.

If it senses that we’re talking about the price of crude oil, it’ll pop in some analytics alongside it, including a fair value for what it detects we’re talking about. Then if it further detects information, it underlines the price and quantity that we’re exchanging information about potentially transacting and allows a seamless submission to our exchanges and our clearing houses. It’s adding a tremendous amount of efficiency to a trader’s workflow, and we’ve actually seen an uptick in volume on our benchmark energy markets on the ICE side of the business during the first half of this year by 60% using this mechanism.

I think that’s because people are becoming more and more conversant in large language models, how to use them, how to have them add efficiencies to their workflows. We just see it manifests itself in our energy markets and then, more recently, in our utility markets.

Matt: Now, it’s very interesting. It’s sort of a capital markets chatbot copilot, as it were. We’re having a conversation, we’re trying to understand things, and you are providing relevant, real-time information that facilitates the exchange of ideas, which ultimately can facilitate the exchange of financial transactions.

Lynn: Yep, that’s absolutely right. And it all comes back to our north star, which is data adds transparency to markets, and information adds transparency. You think about what all of these large language models are doing, they’re adding transparency, they’re giving you pieces of information. That’s why the underlying data that underpins them is so critically important to have correct.

Matt: Yes, I agree. It seems like much the whole world agrees now, whereas even nine months ago, I think the dots had not been connected for people, but of course, we had ChatGPT, and that was kind of a moment we’ll all look back on. We recently did some analysis here at Madrona, working with some AlphaSense data about just how many people are talking about generative AI and specifically looked at Q2 earnings, we’re not quite through the Q2 earnings season, but we’re getting close, and what was amazing was that we looked at both software as a service companies and then other or non-SaaS, and there was about a hundred of these software-as-a-service companies. They mentioned generative AI or AI on average seven times per company, with over 700 mentions on their earnings calls just this quarter.

And what was maybe an even more amazing statistic was that the non-GenAI companies, in this case, over 300 of them had mentioned on average five times in their earnings call generative AI or AI-related topics. And this is companies, Warby Parker, Lululemon, Warner Bros. I mean, you can go down the list. A whole bunch of amazing companies that you all, of course, have helped and trade on the New York Stock Exchange. What are you hearing from CEOs and CFOs from tech and non-tech companies in this GenAI area?

Lynn: Well, I agree with you. AI seems to be the letters of the year. I mean, I grew up in a generation where we had letter of the day, these are the letters of the year. From every company I talk to, they talk about the importance of data. And unsurprisingly, it’s a topic that I love to talk about. Now, there is so much data floating around in this world. It is really hard as a human to process all of the information that you have at your fingertips. I mean, petabytes of data at your fingertips on any one given day. You need things like AI, you need things like large language models to help parse through the data in an efficient fashion that’s going to give you additional insights, it’s going to give you good insights on some of the names you just mentioned on what your customers are doing, what your customers are saying about your products, and what your customers are doing in terms of purchase behavior.

That’s going to help drive your own investments, it’s going to help you drive where your next investment dollar is going to be. In the case of Warby Parker, it’s going to be on a pair of sunglasses. They recently rolled out sunglasses that flip from two different colors. What are people saying about that? Has that been a hit? If it’s been a hit, should we make more of those types of products? So it really helps the investment flywheel in a company accelerate because you’re getting that piece of information a lot quicker than you normally would.

Matt: I think that’s well said, and it makes me think back to the FinTech market specifically and 10, 12 years ago, just the rise of cloud computing and computing at massive scale. I think, as you’ve mentioned, part of what ICE took advantage of, I think architecturally as a business, how are you seeing within FinTech the potential for leveraging GenAI and LLMs? And of course, I know you’re already doing things, but how are you all strategically thinking about where this might go and evolve your business over time?

Lynn: We’re looking at ways within our own business across ICE to make processes much more efficient that are ripe for efficiency plays. One business, I mentioned it earlier, that ICE is keenly focused on is our growing mortgage technology business and how to make the process of homeownership in the U.S. much more streamlined and, therefore, more efficient for the consumer. And how do we put good data in the consumer’s hands so they’re making a more informed purchase or getting a more informed rate on an interest rate?

If you look at the tremendous amount of data that’s available within that vertical in particular, we think there’s a tremendous amount of insights that can be unleashed to make the process more efficient in terms of buying a home, securing a mortgage, refinancing a mortgage, even the securitization process. It’s an area that we’re keenly focused on at the moment.

Matt: Given the leading-edge work that you all at the NYSE Group are doing, as well as the fact that you’re really looked to from all these public companies or companies that are considering going public, are they also asking you for your perspective and advice in these areas? And how so?

Lynn: Yeah, and we like to be a good partner to our firms. I always describe companies the first time we meet, either early on the startup side or we’re pitching them to go public on us. We’re in it to be your business partner. And as part of being your business partner, we’re here to share our experiences. We’re here to share our learnings, we’re here to share our tools. I think it’s important to zoom out on issues because there might be a micro-type issue that someone is struggling with around data privacy, for example.

But if you zoom out just a little bit, and while you may think a consumer business doesn’t have anything to learn from a financial services business, they wind up having a lot to learn from each other because of customer information, for example, and the way people handle customer privacy information. There are a lot of learnings to be shared across our companies, and we’re sort of the clearing house, if you will, for that type of information. I sometimes feel like my job is to help issue spot and to spot macro trends for our companies and make connections within our community to allow companies to have those conversations on topics that they may think they have nothing in common on, but they actually do.

Matt: It brings to mind your own experience now having been president of the NYSE group, it’ll be approaching two years, what has been a bigger learning or two for you in terms of this leadership role and something that you’d want to share with others that are in similar leadership roles? I’m curious what you’ve been learning over the last couple of years.

Lynn: The thing that excites me the most about this job is just the platform that NYSE has and the role that we play in financial markets. Certainly, thought leadership and financial markets globally, but the role we can play is to be a convener of people to have really important in-depth conversations on a variety of topics. The topic of this year is clearly the two letters we’ve been talking about, AI, and that’s an area, but AI is one thing, and if you double-click on AI, you wind up with a variety of sub-issues, data privacy being one, what the application of AI to different industries is too? It’s such a meaty topic that I think, given the role that we have in financial markets, we have a very good opportunity to bring people together to discuss these topics.

Matt: It’s a topic that we’re going to be unpacking more, and we’re just delighted that you’re going to be coming out to join us for the Intelligent Applications Summit on October 11th in Seattle. It’s going to be a very special event again this year, and we’re looking forward to you sharing your perspective and learnings on what you’re doing in the AI area, as well as what’s happening in the capital markets.

Thinking back again to where we started in this conversation, the wonderful rise from being a computer science major and a programmer and working at IBM to being president of the NYSE Group and all these other responsibilities with ICE, tell us what advice you would give to college students today and college students that are trying to start off on the right foot with their career journey.

Lynn: I was fortunate in that I had the opportunity to start in a company, IBM, which gave me the ability to interact with multiple different companies. I started in a consulting role at IBM, and that gave me the opportunity to see a variety of different industries. The one thing that I think was incredibly important in my first few years was that I fell in love with a specific area, so the advice I always give to college students is you’ve got to love the field you’re in. You have to wake up every day wanting to learn more about it. I wake up every day and I come to 11 Wall Street, and I learn something new because that’s just what you should be doing. The world is constantly changing, constantly evolving, and if you have the opportunity to join an organization where you could love the subject matter and continuously learn, you’re in the right place.

Matt: It’s such good advice. Having that feeling that you wake up every day wanting to learn more, being at a place where you love the subject matter, you enjoy the people, you respect the people, completely inspiring to me as well, and I’ve been out of college for a few years.

Lynn: So have I.

Matt: I can’t wait to spend more time together here and it’s incredibly impressive what you’ve done in your rise through NYSE and then this broader role and what you and your teams are doing on their front foot to innovate in areas like AI and generative AI in particular. So Lynn, thank you very much for joining us on the podcast, and look forward to seeing you soon.

Lynn: Thanks for having me.

Coral: Thank you for listening to this week’s episode of Founded & Funded. We’d love you to rate and review us wherever you get your podcasts. If you’re interested in learning more about the New York Stock Exchange and what they can do for companies, visit their site at www.nyse.com. And if you’re interested in attending our IA Summit and hearing more about IPO trends from Lynn Martin, visit ia40.com/summit. Thank you again for listening, and tune in a couple of weeks for our next episode of Founded & Funded with Statsig Founder and CEO Vijaye Raji.

This week, Vivek Ramaswami talks with Chroma Co-founder and CEO Jeff Huber. Chroma is an open-source vector database, and vector databases have taken the AI world by storm recently, with Chroma quickly emerging as one of the leading offerings in the space. Jeff’s background is in machine learning, and for him and his co-founder, starting Chroma was a classic “there has to be a better way moment when it comes to going from demo to production”. Vivek and Jeff talk about why on earth we need another database, the new wave of software engineering happening right now, community-led growth, the competitive landscape, and so much more.

The 2023 Intelligent Applications Summit is happening on October 10th and 11th. If you’re interested, request an invite here.

This transcript was automatically generated and edited for clarity.

Vivek: I’m Vivek Ramaswami. I’m a partner at Madrona, and today, I am joined by Jeff Huber. Jeff is the co-founder and CEO of Chroma, an AI-native open-source embeddings database, which is also more commonly known as vector databases. Jeff, thanks so much for joining us today.

Jeff: Yeah, thanks for having me. Looking forward to the conversation.

Vivek: Jeff, you’ve got a history in the tech and startup world that predates your founding story at Chroma and I would love to hear about that. Maybe you can just share a little bit of background on yourself. How did you get into this world? How did you get to the founding of Chroma?

Jeff: Yeah, maybe the quick life story version of this is I actually was born in the Peninsula, grew up in North Carolina, and then moved back out to the Bay Area about 11 years ago to work in early-stage startups. It’s really fun to build things from scratch. A number of at-bats there. The most previous was a company called Standard Cyborg. We went through Y Combinator in the winter ’15 batch, which at the time felt like being late to YC, and now there’s winter ’23 or whatever, or summer ’23. So now winter ’15 feels like a long time ago. Standard Cyborg had many journeys, I think a few of the journeys that are maybe relevant to our conversation today. We had built out an open-source, online, dense reconstruction pipeline for the iPhone. The iPhone 10 came out in 2018, and we thought this was really interesting. There was a new sensor that was going to be in everybody’s pockets, and that was going to enable accurate 3D scanning to be in every person’s pocket for the first time.

Apple just gave you the raw depth data, so we built up this whole SLAM pipeline around that, open-sourced it, and then we were working with a bunch of companies to develop both conventional computer vision-based approaches as well as machine learning-based approaches to analyze that data at scale. We were doing things like virtual helmet fitting, powering a smart glasses company’s virtual try-on and fitting solution, working with a cycling company around shoe sizing and fitting, all kinds of interesting sizing and fitting applications.

We were doing stuff with ML. It was, like we said, the Wild West. It still is today, but at the time, TensorFlow 1.5 had just come out. PyTorch wasn’t really a big deal yet. We were doing point cloud segmentation, going way beyond object detection images. The thing that we really felt the pain of quite intimately was how hard it is to take anything that has machine learning in it, to take it from demo to production. That was the pain point that my co-founder and I, Anton, connected over. Was, wow, machine learning is really powerful and likely has the opportunity to change the way that we build software and change the world more broadly, but getting from demo to production is incredibly hard. That’s actually part of the reason we started Chroma — to solve that problem.

Vivek: Super interesting. Well, first of all, you were building in machine learning far before it became cool in this current hype cycle that we’re seeing and with all the interest. What were the challenges of building in that period relative to today?

Jeff: Yeah. Obviously, large language models are a bit of a paradigm shift. Before, if you wanted to build a machine learning model to do a thing, let’s say, for example, draw a bounding box around all the cars in a picture — a classic computer vision use case. You had to train a model to do exactly that from scratch in many cases or maybe you use a base architecture and you fine-tune on top of it. And cars is common enough. You could pick up a model probably fairly easily and get something decent, but let’s say that you wanted to draw bounding boxes around chairs, maybe that data doesn’t exist. It’s a huge effort to collect the data. It’s a huge effort to label the data. It’s a huge effort to train anything at all. And you can put in all of that effort before you even get your first taste of how good this thing could be.

Then you have to put in a month or more of work probably for each iteration cycle. You use the thing that you built, it works, we kind of have a sense for where it’s weak, and let’s do some things, label some more data to go try to make it better in that way, and then you wait two to four weeks to get your next answer. It’s a really long feedback loop. You basically have very little information as to why the thing is doing the thing that it’s doing, and really honestly, you have educated guesses for how to make it better, which makes for an incredibly infuriating developer cycle. Take somebody who has also built out React applications and you can change the style sheet and dynamically updates instantly. You can get that feedback loop down to milliseconds. Feedback loops of months or more is really painful in software development.

So yeah, those are a bunch of challenges with traditional deep learning. I think with language models and this news zeitgeist of people calling it generative AI. The interesting change that happened is that there are now general-purpose models, and you can do things with just base models that before you’d have to put a ton of effort into. Now application developers can pick up GPT-3 API and pick up some embeddings and in a weekend, have an interesting thing, and that is a very new thing. Before, building even the small demos was only the purview of machine learning engineers, data scientists at best and was sort of untouchable for your average developer, not average in terms of skillset, just average in terms of backend development, frontend development, full stack development, infra. And now those engineers can build things, and they are building things. And that’s a pretty incredible and exciting thing.

Vivek: Yeah, the resources and tools available to you as a builder in this space now are obviously significantly greater than they were not even just five years ago, even a year or two ago, and that’s terrific for the devs and the engineers in this space. I definitely want to get into how you think about generative AI, and even before getting into Chroma itself. Let’s just start at a higher level. We’re all hearing about AI, we’re hearing about vector DBs, we’re hearing all of these things. Just in terms of this market, why does AI memory matter at all? When we talk about things like AI memory retrieval, just paint a picture of why that matters, how that connects to this generative AI boom that we’re seeing today, and all the applications like ChatGPT and others that so many of us are familiar with?

Jeff: One way to think about what’s happening right now in AI is a new wave of software engineering is starting. Historically, the only way to get a computer to do a thing was to write explicit instructions in the form of code, and hopefully, the computer wrote those instructions, and hopefully, you didn’t write any bugs. If you followed those guidelines, you didn’t write any bugs, that program’s going to run 9999.999999 times out of 100. It’s always going to work. That is not the case with AI. At least today, we don’t yet have the tools necessary to get that level of reliability. That being said, traditional software is also bounded because there’s only a very narrow set of use cases where you can so tightly define all of the inputs such that traditional software development works. Imagine the space of all possible programs that can exist; code written by humans likely only covers 1% of that map of all possible programs that can exist.

It’s a classic example in xkcd. I believe it’s a manager goes to his employee and says, “Hey, can you add this new column to store a last name?” And the person says, “Yeah, it would take three minutes.” And then he says, “Hey, can you also add a thing that says if they upload this photo, whether it’s a cat or a bird,” and the person says, “Yeah, that would take three years.” That was an xkcd comic from a while back. I think that has changed and has gotten a lot easier, but it remains true that it is difficult to bring things in AI to production. So there’s, I think, a lot of exciting work going on. Obviously, retrieval is one method. So zoom out again. We want to be able to program language models, we want to be able to get them to do what we want them to do every time. And you basically have two options, and they’re not mutually exclusive.

One option is to change the data that the model has seen in its training, and you usually do that today through fine-tuning. The other method is to change the data the model has access to at inference time, and that is through the context window, and that is through, generally, retrieval. Instead of programming with code, we program with data. And that’s what this new zeitgeist in programming will be about — how do we guide the execution path of language models, this new generation of software, and how do we guide it using data? And again, retrieval is a really, really, really useful tool. I think that most production use cases today need to use retrieval, and I think most production use cases in the future will also need to use retrieval. I think fine-tuning today is probably a fairly low percentage of use cases, maybe like 10% need fine-tuning, though I expect that number to go up over time.

Vivek: That makes sense. And so retrieval is important, and I think there’s this new paradigm of companies that are being built. Especially if you are leveraging all of the AI tools that we have today, retrieval is more and more important. At the same time, one could argue that there’s a number of databases that already exist. And you just go on DB-Engines, and it lists 400 plus. Why do we need a new type of database? Why are we hearing so much about new databases and new vector DBs that are popping up, not just yours but others? What’s your sense of why we have this novel need?

Jeff: I think each generation of software, whether you look at web, mobile, web 1.0, they have different ergonomics and they have different needs that developers have, and that leads to new kinds of databases. What happened with web and mobile, developers both wanted to have faster development speed, so they wanted to get away from tightly rigid schemas and migrations, and then they also wanted to be able to support much larger scale. There’s the classic meme version of this, which is MongoDB is Web Scale, which is pretty funny to go back and watch, but there are some truths to that: different tools are better at different things. And you could put racing tires on a Honda Accord, but it does not make a race car.

In the current phase that we’re in, my recommendation to application developers is to not necessarily try to take an existing technology that’s added supposedly vector search on top of an existing database and try to make it work for your application because application developers need to squeeze all of the juice and get all of the alpha out of the latest and greatest technology. And you can’t be six months behind or a year behind because you’re too far behind. Furthermore, you really need to have things like really great recall. Again, production is very hard in this world, and so doing things like getting much worse recall but in an existing database is not a trade-off that I think developers should make.

So there’s a bunch more to it, which I can speak to. Everything from integrations, ergonomics, dimensions around scalability. There’s a lot to do with even just the low-level data structures that certain storage algorithms inside of traditional relational databases, for example, are good for and certain things that they’re not good for. I think, in some ways, the proof is in the pudding. Developers are reaching for, in many cases, these purpose-built solutions because they solve their problems. And ultimately, that’s what developers want. Developers want to build things, and developers do not want to spend lots of time taking a Honda Accord and retrofitting it to be an F1. They just want to go. They want to build something. And if both tools are open-source and both tools are free, there’s kind of an obvious answer there.

Vivek: I think you named a number of dimensions in which, to your point, you can’t retrofit a very different kind of car with different types of tires. It needs to be purpose-built for what the engineer and what the developer is trying to do.

Jeff: Adding to that, we’re just so early in all of this stuff. We’re just so early, and I think it’s comical how people are adding vector search to their existing databases. Oftentimes, they are pretty bad versions of vector search. There’s a ton of hair on them not advertised, but they add it and they throw up the marketing page, and the VP of marketing is happy, and so is the CEO. Maybe the public market is happy too. We’re so early.

It is not even obvious that cosine similarity search will be the predominant tool 18 months from now. It’s not obvious that points, which vectors are, and fancy rulers, which search is — points and rulers. I do not think that’s a sufficient tool set to build a production application. A lot more has to exist and has to exist at the database level. I think that’s the other point here is that just we’re so early to what memory for AI means and looks like, and it’s pretty likely that a learned representation ends up being the best anyways. You actually want to have neural nets running against your data and at this very low level to power retrieval. And that’s how you get the best memory. And if that’s how you get the best memory, that’s what you’ll want to use.

Vivek: Yeah, you made the point. It’s sort of like how every app is trying to figure out their generative AI strategy or their generative strategy, and they’re just creating something that sits on top of OpenAI or GPT-3 or 4. And now it’s, we can add this vector capability, but time will tell how much that really moves the needle or not.

Jeff: For the record, for the database companies listening to this, keep doing it, please, because you’re educating the market and doing me a huge favor, so we appreciate it.

Vivek: That’s great. Well, let’s get to Chroma then. So you kind of laid out really nicely what the market looks like and the needs for developers today in the kind of databases that they need and especially as they’re building these kinds of products. So what was the founding story to get to Chroma itself? Where did you see that there was a unique need for you and your co-founders that led you to this ah-ha moment?

Jeff: I’ll say upfront, in many ways, I think we’ve gotten really lucky, and I think the community ultimately is why the project has grown. We released the project on Valentine’s Day of this year. It was the day after when it was supposed to release, so we had a bug we had to fix. The growth we’ve seen from those early days to this point in terms of the number of users and people talking about it online has been entirely organic. For most of this year, we’ve been a very small team. We still are, to this day, a very small team. That will remain true for probably some time, honestly, into the future. With a small team, you can only do so many things. And the thing that we chose to focus on was trying to serve our users to developers. They use Chroma. Trying to serve them as best as possible, take their feedback seriously, give them help on Discord as fast as possible.

The community helped to make it better in terms of giving us feedback, opening up pull requests, and opening up issues. I think we’ve been lucky thus far or our path to this point. We started the company, again, because we felt like the gap between demo and production and building anything with machine learning or AI was too hard. We suspected that embeddings mattered, that there’s something about latent space and embeddings, which, pun intended, had some latent alpha. There’s something there that has not been fully explored yet. When you go on an exploration, you can have a hunch that there’s a promised land at the other end of the journey. You don’t know. You’re not certain. And so, the thing that you want to do is run really clean experiments. They’re tightly bounded for time. You’re very, hopefully, opinionated about whether something’s working or not. And don’t trick yourself. Don’t trick yourself into thinking, “oh, this is good enough, we’re getting good enough reaction to this, let’s spend another three years of our lives.”

I think that we as former founders had a high bar for what we wanted to see in terms of what existed versus having to push it for years and years. There’s a longer story there, but basically, we had built out a bunch of retrieval systems ourselves to serve some analytics we were doing for other use cases in embedding space. We found that we didn’t feel like any of the solutions that existed really addressed us as developers. We felt like they were really hard to use, really complicated, really strange deployment models, really strange pricing models, just didn’t fit. It just didn’t make sense. One of those classics, there had to be a better way, and started talking to some more people, some more users that were building stuff back in December when LangChain was a very early project. OpenAI embeddings had just gotten 100X cheaper or something like that with ada-002 embeddings.

And yeah, just started training a bunch of users and found that they shared the sentiment that something that was easy to use and powerful didn’t exist. And so we’re like, that’s interesting. Let’s go see where that goes. Let’s follow that trail and see where that goes. And that led us to, again, launching the project in open-source, and now cumulatively, the Python project has crossed 1.5 million downloads since launch and 600,000 plus of those in the last 30 days.

Vivek: Let’s talk about the community. It’s important and integral in the Chroma story and the momentum you’re seeing now and where you’re going. You talk about how you got some users, and people start to find out about you. There’s some organic growth. I’m sure a lot of the founders listening to this podcast, especially the ones who are building open-source products, are like, how do I even get the first few people in? How do I seed the initial community and start to attract more people? Is there something that you and your founders did to help build that up, or would you say it’s truly organic? I’m sure there’s some elements of both.

Jeff: The general sentiment is if you build it, they will not come. I have certainly always believed that because people underrate distribution, overrate technology day in and day out. I’ve done that before. It feels like that didn’t apply here for some reason. I’m not exactly sure why. It feels like we kind of did build it and they came. So I don’t know if there’s really much to learn.

I think one thing that we did do unintentionally is at the end of December, I had reached out to a bunch of people on Twitter who had tweeted about LangChain and embeddings and all this stuff just to get their feedback on what we were building and see if they shared our opinion. What I realized now, is that we not only were doing user testing and user feedback, but we were also talking to the loud users who had large presences, generally speaking, or were active on Twitter and thinking about the edges of technology. And so in a weird way, we ended up, I think, getting a lot of the early adopters and the early vocal people to be aware of what we were building and why we were building it and in many cases using what we were building. And again, that was very unintentional, but that didn’t hurt.

Vivek: It’s interesting. It’s funny when we say, “back in December” as if that’s years and years ago. It was eight months ago, but it feels like that because of how much the entire ecosystem has grown since. You talk about early days of LangChain and the number of users that they have now and what that community looks like and what Chroma is growing into. Do you find, and I think would be interesting to hear your take on, is building a community in this era and this cycle in AI very different from building a community a few years ago? One of the reasons I ask is I think we all see the number of GitHub Stars. To get to 50,000 stars is just sort of achieved much more frequently now than it was before, and you can kind of go down the list of metrics that matter and what you think about this open-source community. How do you think about building a community today and growing that and sustaining that community because a lot of this is going to be novel relative to maybe the open-source products of yesteryear?

Jeff: Yeah. I don’t know that any of the best practices have changed. I think you need to care about the people who are engaging with your thing. I think you need to give them good support. I think you need to try to get them on Zoom calls, and get to know their names, and get to know what they’re working on, and figure out how you can make what they’re working on better if you can. There’s just a lot of boilerplate community stuff, which isn’t particularly glamorous, but it is the way. There’s a difference, maybe with AI. In some ways, it’s a good thing. In some ways, it’s a bad thing. So in some ways, it’s a good thing that there’s just a lot of eyeballs on this stuff, and people are paying attention to it. Top of the funnel problem is a little more solved than it is in other areas where you’re trying to get people to pay attention.

I’ve heard stories of the early days of HashiCorp with the whole infrastructure as code movement. And really, it took years of going to conferences and tons of evangelism to get people to understand what this thing is and why they should like it. It’s not quite like that. It feels like people, at a minimum, want to like it, and their CEO is saying that they should have an AI strategy. There’s the zeitgeist. The downside of that is that you get a lot more tire-kicking type applications, or not every user who joins our Discord is going to stay engaged. The Discord is bigger than probably it would be if we were building infrastructure as code in 2017 or whenever. HashiCorp got started earlier than that. But the number of engaged users is also percentage-based, probably lower than theirs was at the time. I don’t know, and it’s kind of a mixed bag. It’s kind of weird and interesting to have that top-of-the-funnel problem halfway solved. That’s sort of a strange thing, I think.

Vivek: Yeah, I like the top-of-the-funnel analogy you draw because you’ve got a lot of people interested and willing to try something. Now the question is, how sticky can you make it? So you’ve got this big top-of-the-funnel, and then you’re going to have some percentage that comes down, and how do you keep them on? So any lessons that you have learned from that process? Which is saying, we’ve got all these people, a lot of initial tire kicking interest, how do you get them to continue using?

Jeff: Yeah, I think we’re in some ways not that focused on that as a metric. Clearly, if you get someone to use it and then they stop using it, it could be because what you’ve made sucks, in which case you should care. It could also mean because, I don’t know, they had a spare extra couple hours on a Saturday and their kids at soccer practice, and they hack something out for fun. And they’re going back to their job on Monday, and they churned. But is that a bad thing? I think, no. I think we’re very much in the age of experimentation with all this stuff. In some ways, what we care about is maximizing the number of things that are built. Even if a lot of those things don’t necessarily stick immediately, we take a very long-term view of this.

And so what matters right now is building the best thing possible that serves the community and making sure people don’t get stuck and trying to increase the chances that they get to production. They get us in reliable enough that they would want to use it. And then I think there’s some just returns to time. We’ve been going for five months. What will be true five years from now? What will be true 15 years from now? Just keep going. Also, if you become obsessed over things like churn and retention, you start doing a lot of bad behavior, which is anti-community. And so you start gathering people’s emails and sending out terrible drip content marketing just to try to keep them engaged. You do weird, bad stuff. And I never wanted to join a community that did that to me. And so I feel like there’s a degree to which you can overemphasize that.

Vivek: I think it’s a very good reminder of how early we are in all of this. Where my brain goes to funnel metrics, what does that look like over time? But really, this is experimentation. We’re just so early in this. To your point, ironically, I’m sure a lot of the classic marketing elements will do more to push people out than it will to bring people in and keep them engaged.

Jeff: Exactly. Broadly, people want authenticity, and the developers, first and foremost, just want authenticity.

Vivek: This is clearly one of the most exciting spaces when I talk about vector DBs in this broader category of AI we’re seeing right now, which also means that there’s a number of folks that are doing this. And there’s the number of competitors, and it seems more often than not, we hear of a new competitor coming up and getting funded and all of that. How do you think about operating in a space where a number of new companies have emerged and raised money? And I’ll ask this in two ways. I think one is there’s competing for users over time that’ll be customers, but the second part is competing for talent, which is not easy in any environment, but I think especially so in a highly competitive space like this. We’d love to get your thoughts on both.

Jeff: Yeah, in open-source, we don’t call them competition. We call them alternatives. I think, in some ways, the first rule of competition is don’t be distracted by the competition. If they were perfect, they would’ve already won. And you’d have no room to be good, and you’d have no room to win. I think we don’t spend a lot of time paying attention necessarily to what the alternatives are doing. We spend a lot of time focusing on our users and what they need. That’s sort of the reactive side. The proactive side is, again, going back to this drum I keep hitting, which is there’s a gap between demo and production. It is true about traditional deep learning. It is also true about language models and embeddings. It’s easy to build a sexy demo in a weekend, and people should. It is hard to build a robust, reliable production system today. Still very hard. And there’s a lot that needs to exist to get people there.

I think when you think about our obsession with developer experience, our rich and deep background in AI, our location in San Francisco, I think those will play out to be differentiating features and factors. I wish our competitors, our alternatives, well. I feel like we all had different theses about what’s happening here and what’s playing out. There’s maybe one point, which I will say, is for better or for worse, we are building a lot of this stuff in 2023. And in that sense, we’re “Late to the game” maybe compared to other people, but I think that we’ve made some different decisions specifically in the architecture of what we’ve built in our building. Those decisions, you probably couldn’t see a couple of years ago because nobody could see a couple of years ago. Those are a little bit hard. If you’ve ever made those decisions, they’re a little bit hard to unwind. Again, not necessarily important to get into detail. I think all that really matters for us is to serve the community we have well. We just got to do that.

Vivek: When you serve the community well, good things happen. And to your point, I think that 2023 we’re saying, “We can’t call it late because the next best time is now for all of this stuff.” It’s like things are happening right now, and this is when you want to jump in. We’d love to get your thoughts if we were to zoom out for a second; we’re kind of nearing what I would at least call the first year of this new generative AI boom that we’re in right now.

If I think of ChatGPT launching in November of 2022 as sort of the initial point, you can say we’re ending the first year, and we’re already starting to see from some applications there’s been a little bit of a decrease or at least a flattening of new users and a little bit of the muted version of that exponential growth we saw a year ago, especially on the application side. You have a really interesting vantage point as a founder building this space, more so on the infrastructure side than on the end-user application side. Give us your perspective. What’s happening here? Where are we in this game? Why might we be seeing some flattening in the growth, and what do you think is happening in all of this?

Jeff: I think it’s sort of natural. It’s like there’s a new technology. If you look at the curve of LK-99 reproductions, I was just thinking it’s all a very similar curve. It’s going to be this ramp-up really fast. Everyone’s like, what’s happening? Go, go, go. And then it’s going to taper off, and there’s going to be a bunch of people trying different variants of it to see if you can do something differently. I’m not an expert in this. I won’t pretend to be either. Eventually, maybe somebody finds a variant that works, and then it has this long, what is it called, the plateau of productivity and the adoption cycle, the hype cycle. It feels like Twitter and AI stuff tends to be these hype cycles are extremely compressed. It was pretty funny to watch.

Right before we raised vector databases, we’re like, “Just on the way up.” And then everybody on Twitter is like, vector databases are stupid. There was an immediate trough of disillusionment four weeks after we raised. And now it’s just a long road back up where people are like, “Oh, actually, long context windows, maybe you’re not a panacea with current architectures. Maybe it is useful to be able to control what goes into these context windows. We don’t want to just dump it all in there because these things could distract.” It’s just funny. These hype cycles are so silly. I think that the biggest risk here, period, to this entire space is if we can figure out a way as a community, as an industry, whatever you want to call it, to cross the chasm from a lot of demos that have been built to a lot of useful production applications, and that means many nines of reliability. Not one nine, 90%, which is even probably pretty good for a lot of LLM-based stuff today, but many of nines reliability.

If we can figure that out, then the amount of value that will be created, I think, through this kind of technology is easily in the trillions. It’s just really, really, really, really big. If we cannot, if it remains more, it’s great for some internal tools, it’s cool for some creative stuff, but we’re not going to put it in front of a user because it’s not good enough, or we can’t predict where it’s going to do well enough, so we’re just not going to put it in front of a user. We can’t trust it, ultimately, if we can’t get to that point. It’ll still be valuable, but it won’t be as valuable. It’ll be 1/10, maybe 1/100 as valuable. That’s something that we think a lot about, and so do the language model companies, and so do many of the other observability folks in the space. Everyone has the same goal, the same mission, which is how can we turn this thing from today, which is alchemy, and how can we turn it into engineering?

Vivek: Oh, that’s a great point, the alchemy to engineering point. I think just going back to what you said earlier is so true, which is the hype cycles were always fast in tech, but they have been increasingly compressed. I’m sitting here talking about nine months ago, there’s euphoria, and now the whole thing has been compressed so much. We’re in agreement with you, which is you’re going to have a natural tapering off of these things. I do think it encourages everyone in this space. There’s great experimentation happening, and then the next phase is going to be what’s useful? How do we get this in production and get someone to actually use it versus talking about all the cool marketing, hand-wavy use cases? I think when that happens, there’s no doubt that it’s going to be game-changing for every application, every piece of software, every workflow that we can think of.

Jeff: Exactly. I think more broadly, multimodal, large model-aided software development will more fully approach the space of all possible programs. That’s a good thing for humanity.

Vivek: Well, before we get to the lightning round, take us through where Chroma is going next. You’ve built this amazing community, and you’re getting an incredible amount of traction. In the last month, two months alone, you’re seeing the majority of your usage and growth come in, and so there’s this amazing compounding effect. What’s next for the company, and the community, and the business going forward?

Jeff: We want to do what’s best for the developers that use Chroma today and in the future. A few things need to happen to do that better than we do today. Number one, we need to have a good story around the distributed version of Chroma, which is mostly about scale. I think most data is not big data, but big data does exist and people want to have a path to scale, and I understand that. We’re working on that right now. It’s a distributed cloud native database version of Chroma, which will scale forever in the cloud. The next thing we’re working on is a hosted offering that many developers in their path of hacking out something on their local computer. Then they want to throw it up on a hosted offering to make it easy to share with their team or with their friends. There needs to be an answer to that, and it feels like none of the existing solutions are that answer. We’re really excited about that, and that’s coming down the pipe pretty soon.

And then the other big basket of work that I’d say is I want to help the community answer a lot of the questions that every developer faces in building applications. It’s very basic things. How should I chunk up my documents? Which embedding model is best for my data, both my data and my query pattern? How many nearest neighbors should I retrieve? 3? 5? 10? Are these nearest neighbors relevant or not? Did I get five of the same thing? Did I get five that are super far away? This is the most basic workflow that every single developer faces when building with these tools and technologies. And there are no good tools or strategies really that exist, at least that have been productized or popularized to help people cross the chasm.

So that’s sort of the first, I would say, task that our applied research team will focus on are those kinds of questions. Going deeper and going longer, again, our goal is to create program memory for AI. I alluded to earlier in the conversation that viewing cosine similarity search as the end of the road for advancements in programmable memory for AI would be extremely foolish. We’re not going to do that. We’re going continue to aggressively research and build more and more powerful approaches to do that well because it matters. We have to cross the chasm here.

Vivek: Really exciting roadmap and I’m excited to see where all this goes. Let’s dive into the lightning round here. Let’s start with the first one. Aside from your own company, what startup or company in general are you most excited about in the AI space, and why them?

Jeff: I’d say Meta. Not a startup, clearly, but I think they’re interestingly counter-positioned to a lot of the other labs. I think that the open-source motion has been extremely bold. And I think that open-source matters here for a few reasons. Number one, it really matters because a lot of the a dvanced retrieval and a lot of the advanced application that you’ll want to build need access to a level of the model where the weights could leak if you’re using a closed-source model. Again, in that goal of going to production, open-source models matter because you want to do that level of surgery, and in order to do that level of surgery, as we currently know and understand, it would be a leaking situation for a closed-source model.

Vivek: They are doing amazing things here. Outside of AI, what do you think is going to be the next greatest source of technological disruption in innovation over the next five years?

Jeff: Does it have to be a contrarian view?

Vivek: Any. Spicy takes or non-spicy takes. Whatever you want.

Jeff: I’ve written about this before. I feel like, in some ways, the two fundamental primitives of advancement are energy and intelligence. You need to have energy to take stuff out of the ground, energy to move it around, and energy to repackage it into other things. You need to have intelligence about what should we take out of the ground, where should we move it and how should we combine it together. And so broadly, technologies that lower the cost and hopefully increase the sustainability of both energy and intelligence are sort of in the direction of flourishing. So obviously, I think AI is marching along the intelligence sector, whether you believe that we’ll get to artificial super intelligence or not, which I’m not sure that I do. And then I think on the energy side, there’s a lot of interesting things happening already around net positive fusion, new geothermal techniques, the list goes on and on. You probably know more about this than I do, but I think that’s pretty exciting too.

Vivek: I was going to say that I think this might be the first episode we have that we talk about LK-99s because it’s probably the first one after this.

Jeff: This might be the first and only episode you talk about LK-99 — now that it’s been disproved, so I guess I’ll take that honor.

Vivek: Exactly. Yeah. We caught the moment in time really, really well. That’s awesome. Okay, last one. What is the most important lesson that you’ve learned over your startup journey but specifically related to talent? You’ve been a co-founder, you’ve been a founder before, you’ve been in this game for a long time. You know that talent is difficult to acquire and retain. Any lessons that you have that you can share with the audience about?

Jeff: I think I’ve really appreciated the idea of go slow to go fast. How that pertains to hiring is I think founders should be extremely picky about the technical bar for engineers. Obviously, or the broad talent bar for other roles, cultural fit, and stuff like EQ. How much of a grownup is this person versus not? And a lot of people are not. I think that that consistently gets underrated. I think what you want to do is build this very small, tight-knit team that shares the same vision and is working in the same direction with a ton of trust. That’s so rare. If you want to make magic, you need to do that. Magic might not still happen, but that’s the best chance that you have to make magic.

And to do that, you have to think in a bit of a weird way about what you’re doing, just dare I say, historical sense. It’s a weird mentality to put on. I think most people don’t want to allow themselves to be that choosy, and they believe I’ll just play the hand I’m dealt. It’s better to hire somebody now than hire somebody three months from now. But if you look at lots of success stories. I think Stripe — it took them over a year to hire their first engineer. I think that even after that, they were still under 10 for a couple of years after that. But that culture, that early DNA and culture of Stripe, even to this day, is paying huge dividends.

Another way to say that is if you believe that what you’re doing has compounding returns across time, you should optimize for the long term. And that’s for both personal health but also things like team composition. Do the thing that you want to be true 10 years from now. I think that’s hard because most founders, including myself, are not very patient. I always want to go faster than we’re going, but I think it’s right. We’ll see if it’s right for us, but I think it’s right.

Vivek: I think what you say is so true, which is coming from the fact that you’re a second-time founder, you’ve been through this before, allows you to take a much longer view. I would think that it’s probably easier for you to step back and take a look at the bigger picture versus eight years ago, your previous startup and that original founding journey. It gives you that perspective, which I think is really helpful for everyone to hear. Jeff, thank you so much for your time. I really appreciate everything, and best of luck with everything at Chroma.

Jeff: Thanks so much. This has been great. Appreciate it.

Coral: Thank you for listening to this week’s episode of Founded & Funded. Please rate and review us wherever you get your podcasts. If you’re interested in learning more about Chroma, visit trychroma.com. That’s T-R-Y-C-H-R-O-M-A.com. If you’re interested in these types of conversations, visit ia40.com/summit to learn more about our IA Summit and request an invite. Thanks again for listening, and tune in in a couple of weeks for our next episode of Founded & Funded with NYSE Group President Lynn Martin.

Today, partner Jon Turow and Cohere CTO and Co-founder Ivan Zhang dive into the world of foundation models for the enterprise. Cohere was founded in 2019, a time when even the most passionate believers did not realize how soon the world would be struck by the capabilities of foundation models. The company announced a $270 million raise in June, which put it in an over $2 billion valuation. And Ivan’s co-founder Aiden Gomez was a co-author of the seminal paper, Attention is All You Need.

Notwithstanding the big numbers, Cohere keeps itself scrappy and hungry, having still only raised a fraction of what others in the space have to date. And in this week’s episode, Ivan shares about his decision not to finish school and instead gain practical experience working in startups and publishing research to basically prove that he could without a degree. He dishes about applying that same renegade spirit to hiring, shares his thoughts about the importance of feedback loops with customers, as well as the differences between fine-tuning and teaching a model how to retrieve the right knowledge, and so much more.

The 2023 Intelligent Applications Summit is happening on October 10th and 11th. If you’re interested, request an invite here.

This transcript was automatically generated and edited for clarity.

Jon: Ivan Zhang, it’s so great to have you here today. Really excited to talk about you and your journey with Cohere.

Ivan: Thanks, Jon.

Jon: Now you folks got started in 2019, you and Aiden and Nick Frost. Ivan, so before we go back in time, let’s start today in 2023. What is Cohere?

Ivan: Cohere is an AI company based in Toronto building foundational models. We’re at about 200 employees now. We have offices in Toronto, SF, and London. Our mission is to accelerate the next leap in productivity for knowledge workers. That manifests in two ways. So one, we provide a platform where you can access foundational models to build features like summarization, search, and chatbots. And another way is to actually use these models to make employees more productive.

Jon: Let’s rewind all the way back to 2019 or a little bit further. Tell me about your journey into deep learning, and then what was the path from that in Cohere?

Ivan: I like to describe my journey into even tech itself as being a bit of an underdog. A bit more background about how I even got into that position in the first place. So at the time, I knew that I was a builder, and that’s how I learned best. And I wasn’t much of a sit-in-a-classroom-and-absorb-a-lot-of-information kind of guy. I needed to tinker. I needed to get my hands on the technology to learn. So when the opportunity came up to drop out of school to work at my friend’s startup, I obviously did it as a backend and infrastructure engineer. I wanted to expand my skills and learn a bit more. And that’s when I met Aiden Gomez, one of my co-founders now, who was interested in starting an indie research group. We wanted to be independent and do research basically for fun and, in a way, prove that we can do it. And yeah, I got pretty inspired by that, and I thought it would be pretty badass to publish papers as a dropout.

And so we just started working together. And after a few years, we felt like we were ready to start a company. We learned our working styles, and we both got more experience just working and learned how things were done properly. Myself, I was exposed to more and more founders just from the space and seeing how the sausage is made. I felt comfortable with the idea of starting a company.

And so in 2019, I pitched Aiden like, “Hey, why don’t we start something new together?” We tried a few ideas at the time, and they were all AI companies, and it was very difficult. Our first idea was a platform where you upload your neural net, and we’ll compress it and make it more efficient. But I mean, the thing is, nobody was using deep learning in 2019, so there wasn’t much of a market there. But what he had seen within Google was that he invented this thing called transformers, and internally it just proliferated. Every single product team was adopting this architecture for solving language problems, and the improvement gains they were seeing were crazy. Absolutely unbelievable.

Just this one tiny change outpaced 20 years of just heuristical engineering. And so we thought it was really cool, and we saw the potential of this technology like, hey, we can actually use this as a way to help computers interact with humans. So that was one thing. So we saw that, okay, computers being able to understand language was basically solved by birth style models at the time. And then we also saw, oh, GPT-2 came out, and that was sort of a hint at scaling these things up was important, increasing the capacity and also like, wow, these things are very efficient, so we can actually feed a ton of data into it.

And the architectural change of making it decoder only now these things can write. And so the two key pieces were there to actually build a system where it can read and write language. And we thought that was quite exciting, and we decided to quit our jobs and bring Nick along as well to build this company. And at the time, we had no idea what the product was going to be. We were just so excited about the idea of making computers understand language and talk to us. And that’s how I got into deep learning. And being an outsider doing deep learning almost gave me so much energy to invest all my after-work time every day, working till 3:00 AM just managing experiments and making the code base a little bit better. It was quite an exciting project.

Jon: Were a lot of people doing research outside universities and corporate labs at the time?

Ivan: So at the time, no. And I think a lot of listeners might know this group called EleutherAI who started a bit after us. We were For AI — and how we approached it was a bit more closed off. We would be open to applicants, but we would be very selective in who’s actually doing research and who we’re getting compute to. I think Eleuther had a different strategy where they were way more open about who was collaborating, and they were also doing open research. But at the time, basically, nobody was doing research outside of labs. And one shout-out I want to give was GCP’s Research Cloud program made this possible. They gave researchers like us free access to TPUs. In fact, we were one of the first users, first power users, of TPUs outside of Google. And that’s how we ran basically all our experiments.

Jon: How do you think the renegade spirit that you had in those early days with Aiden reflects in the way Cohere operates today?

Ivan: We hire differently. We look for people like us who are from very different backgrounds, interested in the field and want to make a big impact. We’ve brought For AI into Cohere as a way to keep the effort going and provide people with unconventional backgrounds a path into research. And what we learned with For AI is that, hey, you don’t need this perfect background of doing all this research at Fair or DeepMind or Google to make impactful work happen. Some of the papers we published were with people doing research for the first time, and they brought some interesting ideas from whatever field they were coming from, and that made a big impact. So we took that philosophy in how we built the early team at Cohere. We didn’t look for the marquee brand name 10-years-at-Google brain, sort of talent. We found folks who are very clearly builders, and they’re interested in the field. We gave them a chance, and a lot of them have paid off.

So I think even though Cohere is our first official startup, Aiden and I learned a lot from that For AI experience. We’re also way more risk-tolerant, I think. The culture is also very playful in that sense. So we like to do a lot of exploring on the technical front. We do word it that way. We’re playing with the technology to find breakthroughs, and we’re very practical. When you’re running a research lab with basically no money, you have to be very practical about what you can and can’t do. And so we’ve spent some time in our engineering to make sure our tools and platforms are working well for our researchers.

Jon: So when you started Cohere, was it clear you were going to provide foundation models as a service?

Ivan: No, at the time, we weren’t sure what to do with a thing that can talk back to you. It’s such a bizarre experience. I think anyone who tried GPT-2 at the time has shared the same experience. It’s cool, but we’re not sure what we could do with it. So, we try to build a product as our first project at Cohere, which was an autocomplete that sat in every text box, and we thought we were just going to throw ads on it and start making money. But I think we were very naive at the time on how difficult the front end would be for something like that. Because even at the time, the text boxes were coded in all these weird ways. They weren’t just native HTML text boxes anymore. There were weird react components, and there were spell components.

So there’s all these edge cases. It just wasn’t our competitive advantage to focus on. So we decided to just rip out the front end and provide an API service. And we knew at some point, well, we’d find some use cases for the generative model. But we thought the embeddings and fine-tuning were also important features to have for developers trying to build something for production. At the time, 99% of NLP use cases required word embeddings, and we knew to take any model to production required fine-tuning some way to customize the product for your problem. So after a few months, we pivoted the Chrome extension into this API platform with generations, embeddings, and fine-tuning.

Jon: If I kind of zoom forward today, what we have as Cohere today is a collection of really high-quality foundation models as a service behind APIs and a particular traction with enterprises, a particular focus on that. Can you talk about what are the most urgent kinds of things that enterprises are doing with foundation models today?

Ivan: Yeah, so at Cohere, we apply an enterprise lens in all we do. Whether that’s the product roadmap, resource allocation or go-to-market. And so they’re very particular product problems for the enterprise, and that’s not endpoint specific. There’s fine tune control of how scalable you want your models to be. Making the precise trade-off between throw put and cost, endpoints that actually make it easy for you to adopt this use case. So stuff like better search summarization, building assistance for your customer support workforce. Stuff like platform management that’s like a non-starter for enterprises if you don’t have that. Being cloud agnostic, with a comprehensive set of deployment options. Whether that’s using the API deploying into your VPC being where the cloud AI products are like SageMaker or serving people on-prem. Beyond just the raw endpoints themselves, we think about the whole enterprise experience and the precise problems they have.

Jon: So, what are some of the misconceptions that CIOs have when they start engaging in these products?

Ivan: I think the biggest one that’s made its new cycles is hallucination. It’s totally reasonable to have an issue with these models that are basically making things up from your prompt. I think it’s actually a feature, not a bug. The model is doing its best to complete your request. The solution to hallucination isn’t to, oh, well, we’ll just fine-tune the model with all our company documents. Well, no, no, no. That’s not what you do with your employees either. You don’t make your employees memorize all your internal processes and documents. You actually just give them a tool, your internal document store, they do a search against that tool, find the relevant information, and then they produce an email, an answer, a report, whatever it is. We have to think about models in a very similar way. We don’t need these things to memorize. Instead, we need to teach them how to use our tools to retrieve the right information. So I think retrieval augmented generation is a big deal, and that’s something we’re serving enterprises with.

Jon: When we touch on retrieval augmented generation, RAG, and we talk about things like in-context learning, that’s often juxtaposed against things like fine-tuning, which you also mentioned. How do you guide your customers about when each of those would be more appropriate?

Ivan: I think the nice thing about this version of AI that we know with transformers and LLMs is that it’s actually quite intuitive to how humans learn and what the limits are applying our intuitions about how humans operate and learn. I would say you need to fine-tune if you need to teach the model a new capability. Like a new process, a new function. Whereas if you’re finding that, oh, it’s not accurate in its knowledge, if you just want to give it more knowledge, that’s a RAG problem. We need to teach the model how to retrieve the right knowledge. So basically, fine-tuning for new functions and give it a database for everything else, like knowledge.

Jon: Let’s talk about the limits of feeding data into a model. In the context of a bunch of announcements that have been made recently from large data platforms, there’s kind of a debate in our industry over whether it’s appropriate to feed lots of raw data into an ever-expanding context window or whether we should instead be building up a semantic model inside an LLM and feeding that back to run a deterministic code in some traditional environment.

Ivan: I’d like to clarify that’s not one or the other. I think, obviously, there is a balance based on the product that you’re trying to solve. As for the limits of these models today, I definitely don’t think these models are at the capacity of solving RAG perfectly — or even learning how to use tools. We have more scaling to do from what we’re seeing internally, 52 billion parameters or 10 billion parameters. Those models solve a specific set of problems, but the customers are demanding more and more complex and sophisticated use cases that we actually do need to scale up these models. So I think there’s more scaling work to do to raise the limits, yeah.

Jon: What do you see as the role of proprietary or enterprise, home rolled models in an ensemble together with hosted models from something like Cohere?

Ivan: I think there will always be model builders within enterprise because it’s such an interesting problem to work on. However, I think most users will just go with an out-of-the-box solution. It’s very hard to hire MLEs and keep them motivated. So if enterprises have the option to just pull an API or pull a SageMaker deployment to solve the problem that they actually want to solve, they’re going to do that. So I think it’s not unlike open source software where people pay for services on top, people pay for a managed offering because they don’t want to be experts in managing Postgres. They just want a database to store information. And I think we’re going to see a similar thing with foundational models. People don’t want to build up the internal expertise to pre-train a 200 billion parameter model. They just want to solve the language problem in their product.

Jon: When I think about this from an implementation standpoint, what are the most important kinds of data sources that enterprises want to prioritize?

Ivan: Yeah, so I think it’s less important about what the source is. I think the most important thing is the data feedback discipline. How well does their feedback loop work within their product? Are they able to get feedback from their users relatively quickly and easily and then feed that back to their ML team? I think that gives me more of a signal on whether that company would be successful with LLMs or not.

Jon: Does that have implications for who should be doing the work within the organization, whether it’s a data engineering team or a product service team? Who ends up doing the integration work within the enterprises?

Ivan: Typically, it’s a product owner who’s entrepreneurial enough to try this new technology, and they see the opportunity because they realize, hey, I have this feedback loop that gives me great RLHF data to further enhance my models. The great thing about these models is that they’re so flexible in whatever language format you give it’s quite easy to fine-tune these models.

Jon: What we see in a lot of waves of disruptive new technology is individual developers and teams excited to adopt the technology bottom-up without a lot of top-down controls. But senior executives are rightly concerned about the sensitivity of the data that’s flowing through these models. So what should enterprises know about the security of using foundation models in the enterprise?

Ivan: Yeah, I think with bottoms-up adoption, it’s about education. I mean, they should know that they’re potentially sending a ton of sensitive information to a third party like Microsoft or OpenAI. But the solution isn’t to ban this technology. There’s so much productivity to be gained that we should instead find an alternative. And that’s Cohere’s bread and butter is that we provide a private LLM that can deploy into your VPC on-prem, so that you don’t have these risks of sending us any data. We don’t actually see any of our customer’s data if they’re deploying with those options. So if you can’t ban it, well yeah, you have to provide a better alternative. Right.

Jon: So I’ll do a couple of lightning round questions, Ivan, that we can use to wrap this up. So one, what has been the biggest surprise to you about the developments of the past six months in our field?

Ivan: I’m surprised that the iPhone moment came so early. My timeline for this tech was three to four more years out, but unexpected things happen.

Jon: What would you expect to see happen in the next six months? What are some of your predictions of developments in our field?

Ivan: I think most work in front of a computer will be fundamentally different. I think the barrier of leveraging computers to do your work will be way lower. You don’t have to learn a special language to make computers do what you want. You don’t have to learn a special graphical interface to navigate programs. I think in a couple of years, if you know how to describe what you want to do, the computer will do exactly that, and that’s beyond just generating emails, but also taking a series of actions and doing most of your work.

Jon: Well, Ivan Zhang of Cohere. This has been just so much fun. Thank you so much for spending time on the discussion.

Ivan: Yeah, thanks for having me, Jon. This was very, very fun.

Coral: Thank you for listening to this week’s episode of Founded and Funded. If you’re interested in learning more about Cohere, visit cohere.com. If you’re interested in attending our IA summit, visit ia40.com/summit to request an invite. Thank you again for listening, rate and review the show wherever you get your podcasts, and tune in in a couple of weeks for our next episode of Founded and Funded with Chroma Co-founder and CEO Jeff Huber.

This week we have the pleasure of having former Snowflake CEO Bob Muglia on the show again. Bob is an active investor and sits on the boards of many next-generation data platform companies, and more recently, he launched his first book — “The Datapreneurs.” Given the long history between Bob and Madrona Managing Partner Soma, we had to have Bob joins us again to talk about his new book and dive into the world of data and AI. These two old friends discuss what exactly a datapreneur is and the Arc of Data Innovation concept Bob wrote about in his book. They also talk through how companies can add value with AI through copilots and agents and what white spaces and opportunities there are for entrepreneurs right now — especially when it comes to semantic models.

The 2023 Intelligent Applications Summit is happening on October 10th and 11th. If you’re interested, request an invite here.

This transcript was automatically generated and edited for clarity.

Soma: Hello, everyone. I’m Soma, a managing director here at Madrona. Today I’m really, really excited to have Bob Muglia, a datapreneur himself with a large body of data platform work to his credit across Microsoft, where he was one of the topmost senior executives and then, most recently, he was the former CEO of Snowflake. Bob is also an active investor and sits on the boards of many next-generation data platform and tools companies. Before we launch into our conversation today, Bob, I do want to take this opportunity to congratulate you on publishing your first book, Datapreneurs.

Bob: I appreciate it. It was more work than I anticipated it was going to be.

Soma: But Bob, given your experience and accomplishments in the world of data over the last 35 plus years, I’m not at all surprised to see the focus of your book, which explores the people and critical pivots and technology history that catapulted us into the modern age of computing and AI. Bob, I thought we’ll just jump quickly into a set of questions that I have for you today to start off the conversation. First, how do you define datapreneurs, Bob, and more importantly, how do you see their contribution in building the data economy of today and in the future?

Bob: Well, a datapreneur is simply the concatenation of a data entrepreneur together, and that’s where the term comes from. This realization, I worked on the book with my co-author, Steve Hamm, and when we first started writing the book in 2021, we really didn’t have a specific objective in mind. We didn’t start by saying we’re going to wind up writing a book. I knew I’d had some things to say and communicate, and we were trying to decide the best vehicle to do this, and after Steve and I talked for a little bit, it was obvious there was a narrative that could be had and turned into a book, and so we started outlining the chapters and things.

In that I realized, in that process of those conversations, I realized that throughout my career, even though I was working at a really large company like Microsoft for a good part of it, I was actually working with very entrepreneurial people and teams building largely new products for Microsoft and for the industry and certainly doing things that were very much revolutionary for the industry. I realized that I’d been working with entrepreneurs all along, even though they were at this big company. So that’s where the idea came from and the recognition that the technology that we experience and live with every day is really the culmination of the work of thousands and thousands of people at many, many companies.

It certainly includes the work of some of these great datapreneurs that I highlight in the book, and their accomplishments have really led us to where we are today, so I wanted to highlight that and describe why some of these things are so important — and give us some people and some background on the technology so they have some foundation because obviously, it’s impacting all of us as we see AI being really the topic du jour of 2023 and clearly going to have a huge impact on all of us.

Soma: That’s great, Bob. Having read the book, I can tell you that you referenced a number of datapreneurs that you worked with in the past and ones that you’re working with currently in a variety of, sort of different companies. But first and foremost, Bob, when I think about you, I think of you as a datapreneur, given what you’ve accomplished with data at Microsoft and then more recently at Snowflake and the impact you’re having today with a variety of startups. I see you first and foremost as a datapreneur. But you also worked with many other datapreneurs over the span of your career. How would you categorize the role the datapreneurs have played in your career?

Bob:Well, they’re the source of all inspiration in some senses. I’m not the deep technologist that builds the product and the code. Mostly I was highlighting the people who did that who were actually in there building core parts of the technology and creating some of the revolutionary ideas that have led us to where we are today. My role, as you know so well, since we worked together for many years, my role was really what Microsoft would call a program manager, the modern term is more of a product manager. But in the Microsoft definition of it, it was really more about the product and building the product and defining the product for the customer.

So I’m used to specifying things, talking to customers, understanding their requirements, and then passing those requirements on to the technologists and the architects inside the engineering teams that build things. That’s sort of always been a core part of my role. As a manager and as an executive, a lot of that comes down to leadership principles, running organizations, instilling values into teams, things like that. I see my role as being very different than a lot of these brilliant people that are actually coming up with these incredible ideas. I couldn’t do that, but I have the ability, hopefully, to help provide them with some guidance to help as people are building the products.

Soma: Absolutely, Bob, absolutely. I want to go back to a little bit the Microsoft timeframe, Bob now. If I remember right, you came pretty early into Microsoft. At that time, Microsoft was working on Windows and OS/2 and LAN Manager and starting to work on Windows NT and maybe even the early days of-

Bob: Actually, it was slightly after that. I mean, it really got started in the latter part, right?

Soma: Yeah, yeah.

Bob: But yes. That’s the very beginning, so.

Soma: From your perspective, if you look back in time, when did you feel like, hey, data is going to be a key, key part of the future of the world? When was that moment in time where you felt, aha, there is something here that is going to fundamentally change how the world is going to be operating?

Bob: Well, no matter how much I tried to get away from it, I kept coming back to data or data kept coming back to me in a way. My first job, the technical job while I was still in college, was working for a company in Ann Arbor called Condor Computer, which had a relational and honest to god relational database. It had a joint command. It was not SQL, it was the very earliest days. SQL was just emerging at that point. This was in the late 1970s, early 1980s. It ran on a tiny microcomputer with these massive 8 inch floppies that stored almost nothing and literally 16K of memory, literally 16K of memory. That was my first experience, and that was building applications for companies.

In college, I was focused on communications, and my first role, my job out of college was at ROM Corporation, which built a telecommunication system for business, an internal switch called a PBX. Our term was a CBX. That was ROM’s product. I worked on the team that configured those products. So again, I had a data-focused job, and I was building data-oriented solutions when I was at ROM. It was really that experience at ROM that led to my moving to Microsoft, largely driven by my wife’s desire, our collective desire to move up to the Seattle area, we were in the Bay Area at the time, and her really finding Microsoft as being an incredible company that was emerging back then.

I wound up joining Microsoft as the first technical person on SQL Server. In a way, that really cemented the focus on data throughout the rest of my career. Because SQL Server, the PC version of SQL Server that Microsoft offer, really did revolutionize business for smaller companies. A large part of what we did in the 1990s was bring out both the server products as well as the database products and tools products. Visual Basic was a big part of that. So you had Windows Server and SQL Server and Visual Basic is the front end for a lot of those early applications. That’s what automated the dentist offices of the world is that software. Some of it’s still running probably, which is probably not a good thing, but that really changed the way people worked with information.

Soma: That is cool. That’s cool. I’m going back and forth in time here, but I want to sort of revisit what happened in the 2017 timeframe. That is when we connected up upon, hey, Snowflake is coming along, is starting to make some revenue, has got some customers, and starting to realize the power of what Snowflake could be kind of thing. First of all, thank you for giving us the opportunity to invest in Snowflake at that time and be on the journey with you.

But one thing that I remember from one of our earlier conversations, Bob, is I was thinking about, Hey, how much should we invest? Because remember, we are an early-stage investment firm, but I remember you telling me, “Hey, Soma, you’re going to make money in this deal. I can’t tell you how much, but you’re going to make some money.” I was thinking, “Hey, how much money am I going to make?” Because the valuation was a little higher than what we were normally seeing at the time.

Bob: Seemed rich at the time.

Soma: Yeah, it seemed rich at the time. In hindsight, I would say, “Hey, you just foreshadowed what the world was going to move toward kind of thing.” But I did not imagine, and I’ll be the first to tell you, I did not imagine what kind of a trajectory Snowflake could have at that time. In fact, so much so that when Snowflake became a public company or went through the IPO process, ended up being the biggest software IPO in history. How did you see the trajectory, and how did you feel about Snowflake becoming the largest IPO kind of thing? But more importantly, what does that signal in terms of the importance of the data cloud platform to the future of the computing world?

Bob: It was really great working with you in the early days. That was really when we were looking at opening an office in Bellevue and bringing in talent in the Seattle area. That was driven by our desire to move to Azure as a platform, as a second platform. We were on AWS already, and we wanted to add Azure as another choice for customers. The realization that there was nobody in the Snowflake team that could work with Azure down in San Mateo, California, we needed some people up north to help that. Obviously, this has worked out very well. I knew the company was going to do very well. It certainly exceeded my expectations in that regard, and the IPO was extremely successful.

I would say it was really successful from this perspective, but for public investors, I have a little bit of concern because the stock went way high, and it’s now still trading below that amount, which is definitely hard for public investors. But it’s been incredibly successful, and I knew it would do well. How well, I didn’t know. The reality of it was the reason it’s been successful is because it solved a problem that was not solved before, and that was how do you scale your databases for analytics to be able to consume and work with all the data you wanted to at the same time and make that information available to all your end users.

Up to that point, the technology before Snowflake really didn’t enable that, and Snowflake was a breakthrough product in the sense that it really was the first to enable a general-purpose SQL database that allowed for unlimited scale. I think that has been amazingly valuable because people are working with data and information. It’s only getting more valuable over time as we begin to find more and more uses for data, and, certainly, the emergence of these large language models and artificial intelligence that is obviously of great interest right now, that raises the importance of data even further. So I think all of these things are important.

Now, Snowflake, as I’ve said a number of times, I mean, I see Snowflake as being in three somewhat different businesses. They’re in the sort of more traditional data analytics, data warehouse business. They’re in the data sharing business and have now established a very strong position in helping companies access data that is outside of their organization and make sure the data can be appropriately shared within the organization. Then more recently, they’ve been really focusing on an application platform, a coherent application platform, the so-called data cloud that provides a set of coherent services for people that are building these next-generation intelligent applications, AI applications, data applications, call them what you want, and then enabling those to run anywhere that Snowflake runs.

They’ve emerged as an important player in the application market, and that’s a fairly new space for them, but it’s going to become increasingly important as we begin to build application services, these intelligent application services, that take action on our behalf. The fundamental distinction is, is that a typical business application acts on behalf of a direct request from an end user, from a person. These class of applications take actions based on information that’s coming into them, and they can do things on their own. So it’s a whole different way of working than creating different business process, but I think it’s going to become a very big part of what people are doing in the next five years.

Soma: The thing that was interesting to me, Bob, as I thought about Snowflake and the Snowflake’s journey kind of thing, you literally have five what I call data cloud platform vendors of some scale and consequence in the world, right? You got the three hyperscalers, the cloud infrastructure guys in Microsoft, Amazon, and Google, and then you got Snowflake and Databricks. I don’t know that I would’ve predicted even, say eight years ago, that there’s going to be five cloud data platform at scale vendors kind of thing, right?

So it’s just fantastic to see how the hyperscalers obviously, that’s not a surprise that they are sort of the meaningful players here, but the fact that Snowflake and Databricks have been able to come to where they are today, it’s just fantastic to see the innovation and how broad the space is and the opportunities ahead in terms of how this can fundamentally change what I call another world of computing and by extension all applications and everything else that people do on that.

I want to come to your book now for a second. There was a great thing that you talked about in your book called the arc of data innovation. To me, now, I was excited when I sort of read through that because it sort of helped me think about, “Hey, how do we visualize that technology progress while making predictions, knowledge predictions, or what could happen in the future kind of thing?” Can you talk a little bit about how you came up with that and how that concept came to life in the book?

Bob: It was always, in a sense, the central concept of the book. It had been present from the earliest versions of it in some form. The idea is that there has been an acceleration of progress over time as technology has continued to improve. That’s why I drew it as an arc because the line represents the speed of progress, and that’s increased as the decades have gone by. The idea of it was to identify both the key data types, the key types of data that were introduced. People don’t always think about it. There’s structured data, there’s text data, there’s semi-structured data, and then there’s what people often call unstructured, which I would call complex data, and that’s video and audio. It really has structure associated with it. It’s just that the structure is so complicated that we tend to think of it as unstructured.

Those sources are now real sources of data that they never were before in the past. That idea of that progressive evolution was always present. The interesting thing is that the finish of it changed while I was writing the book because when I started writing the book, the apex of it was the data economy and the idea that data is a central part of what we do in our business. What I recognized was that as I was writing the book and I saw the explosion of what’s happening in the AI space, I realized that the horizons that I had thought in my head for when some of this intelligent technology was going to become available and useful and really achieve some major milestones like artificial general intelligence, and potentially super intelligence, those were going to happen in a much shorter horizon than I expected.

If you asked me in 2021, when would we have AGI, I would’ve said 2100, and now I will say 2030, thereabouts. That’s a pretty dramatic change. Frankly, it was a breathtaking realization for me because I had always believed that we would develop this intelligent software, that that would come, but I just didn’t think it was going to happen in my lifetime, and now I realize it is. The implications to all of us are just really profound. So the arc of data innovation really now goes to super intelligence and potentially a technological singularity, which is just really a continued acceleration of progress largely faster than human speed. I think it’s going to happen. It’s certainly something that, in general, I think it’ll be very positive, but I know it’s also a bit scary at the same time. I think it is not 50 years away, I think it’s probably a lot less than that.

Soma: Yeah, I think this notion of predicting when you’re going to see major inflection points in technology, I think it’s anybody’s guess kind of thing, but it’s one of those things where sometimes we overestimate, and sometimes we underestimate. I actually remember when I was still at Microsoft back in the 2012, 2013 timeframe, if you talk to any of the traditional automobile companies, “Hey, when are you going to have a truly self-driving car?” They were all talking about 2030, 2035, 2040. Then for a while there, now with the self-driving progress that companies like Tesla made, everybody thought, “Hey, in the next three years, it’s going to happen.” The reality is it’s somewhere in between, right? But the rate of innovation is the one that I think I’m more strict there about because that sort of signals what is possible and whether it is actually 2021 or 2025 or 2030. We cannot debate that.

Bob: Well, I mean, having played with Teslas and things. From what I could see, the technology was still a few years away. I had thought, and this is just getting reinforced in the past set of months, that the 2030s would be the era of robotics and that we’re going to see another explosion of innovation in autonomous robotic devices that are part of our lives in a variety of ways, ultimately resulting in humanoid robots that have intelligence and can do things on our behalf. The interesting thing about a car was the autonomous car, which was I think about an Uber ride, and every Uber ride that I’ve ever been in ends with a brief conversation with a driver about where to drop you off.

I was like, “Well, how are you going to do that with a machine?” Now I realize you can just tell it. Now you’ll be able to tell the robotic car where to drop you off. So I mean, some of the problems that seemed very intractable, now that English is a language that computers can understand, I mean, a whole bunch of problems that were there before disappear and become at least potential to solve.

Soma: That is very true. That’s a part of NLP or natural language processing. I guess you’re right. Hey, you can talk to a machine. Now pretty soon, I think I know you can already touch a device with touchscreens. You can now speak to a device. The interaction — the more you can interact with the device or with a machine or a system, similar to how you interact with each other, the more natural it’s going to be, and the more people can seamlessly think about computing in an ambient environment as opposed to, hey, I need to walk up to a machine, or I need to walk up to a device, as opposed to, I’m going through my life, I’m doing my job, I’m sort of having fun, and I can leverage technology in my natural workflow, so to speak.

I think that’s an exciting world that I’m looking forward to. Like you said, for a while there, it wasn’t clear whether it’s going to happen in our lifetime or not, but I think the rate of innovation, there is a fond hope that could happen well within our lifetime. So we’ll see how it goes.

Bob: Even for old people like us, even for old people like us, but for the youngsters in there, I think they’re going to see a whole lot. For people who are in the earlier stages of their career, they’re going to see a massive amount of change. I mean, think about the fact that these devices have intelligence in them. We’ve always been able to put rules, and we’ve always been able to program rules in a computer, but now you can actually take and have intelligence associated with decision-making. It’s really extraordinary. It’s an extraordinary innovation.

Soma: That’s true. Bob, you can’t talk to a company today without them telling you why they’re an AI company. Every company says that.

Bob: You can’t, you can’t. That’s for sure.

Soma: But here is the thing that, and we sort of have this taxonomy, at least a mental model that says, “Hey, there are going to be AI-native companies, and then there are going to be AI-enhanced companies.” Which means most companies that exist today have to think about how to incorporate AI in fundamentally core parts of what they are doing as a business or as a company kind of thing. The challenge, though, is all of these companies have an existing what I call data architecture, how they think about data, how they deal with data, how they process data, how they take advantage of data kind of thing. As AI becomes more and more central and core to every business and every company, how should an existing company, where should they even start about, hey, I want to be a AI-first company, but I do have a way of doing things today with data. What should I do? Any sort of thoughts on what you tell companies?

Bob: One of the really interesting things about this technology is that it really can be additive to existing applications and add value. I think Microsoft demonstrated that with this co-pilot approach, which I think will be the predominant, at least initial, way in which AI will be incorporated into existing applications, which is these agents that sort of support you in your efforts to do things. I think most companies who have an existing product can take and modify that product to put in one of these co-pilots that can build off of the knowledge basis that it has, and fundamentally, essentially just be a much more effective help system to help you work and navigate through the product you have.

If you don’t have an existing product but you have a business that you’re running, one of the interesting things that now there are opportunities to build applications that couldn’t be built before. Here I always just say, “What is the domain expertise that you bring?” Because it’s now possible to effectively take that domain expertise and leverage the intelligence in these large language models to bottle effectively the expertise you have and put it inside the application so that the knowledge and behavior and processes that you created can now be run and executed by the machine versus by people that previously performed these actions.

So there’s opportunities across the board to do that if you’re in an existing business, whether you have an existing application where you can obviously take the knowledge base and everything you did and directly incorporate it or whether you’re building something new. Then I’d say, just to continue along the lines of the different types of apps, you say there’s just new kinds of applications that you couldn’t build before that now, for the first time, can be done.

Soma: That is true. I fundamentally believe that AI is going to permeate every industry, every company, every walk of human life kind of thing, okay? But having said that, if you think about the different verticals or the different industries, do you think some are more naturally attuned to having the most innovation and positive impact by AI? Or do you think it’s just going to be across the board?

Bob: I think it’s pretty horizontal. I mean, like anything, there’ll be uneven distribution of it. Certainly, different industries will have different speeds of adoption somewhat based on the regulatory challenges they face. Frankly, in some industries, take the medical industry for example, we have to make sure that the technology is matured to the point where it’s working correctly. I mean, if AI hallucinates and writes a poem incorrectly or something, or it puts something in a poem, it’s one thing, but if it misreads a doctor’s notes and that results in a patient getting the wrong drug, that would be very bad.

So you want to make sure you have very high degrees of accuracy in certain scenarios. So that will take a little bit more time, but every domain will be impacted. It will certainly impact a lot of people in their careers and jobs because the more mundane sort of repetitive tasks, the tasks that tend to be repeated, those are the ones that are historically been the providence of people. Those are the sorts of tasks that will wind up being automated first.

Soma: One of the things that I really enjoy about your book is in addition to talking about data, data technology, the modern data stack, what’s happening in the world of data kind of thing, you go through what I call case studies or specific examples of companies and the kinds of technologies that they’re building. You talk about many, many different companies developing solutions for what I broadly refer to as the modern data stack. But having said that, do you think there are still white spaces or big problems that need to be solved? Particularly if you’re talking to the next generation of startup founders, what would you tell them about, Hey, what opportunities exist in the modern data stack? Or do you think that’s all being solved, and we should just move on?

Bob: Well, it’s not all being solved. Let me just start by saying that. The modern data stack has made great progress. As you mentioned earlier, there are really five different vendors providing solutions. It’s good to see those solutions maturing over time. They all come from different places. They all start from a different place, but they’re all kind of building the same product now. They’re all heading towards similar products that they’re creating. Although there’s strengths and weaknesses of the vendors, like I say, largely associated with their history.

There are definitely open spots in the modern data stack that still need to be filled, largely around compliance and management. Those are areas where I think there’s a lot of challenges still. Certainly, access control and managing coherent data access control continues to be a significant problem that I don’t feel is well solved. I think one of the biggest areas of future potential is building and creating what people talk about is semantic models for different things. In particular, I think it’s interesting to think about the semantic model for a business. If you look at any company’s business product, any organization’s business, and their process associated with executing their business, it’s the knowledge of how that work is scattered in different spots.

It exists somewhat in the applications they run. It exists somewhat in the analysis things that people do using tools like Tableau and Power BI. It exists in the heads of a bunch of people. It exists in documentation and probably maybe more than anything in Slack messages. It’s all over the place, and it’s not very well defined. In order for these language models, these agents, to behave in a way that is consistent with what we want, they need to understand the semantic model of the business. So I think we have to be much more explicit about that because if we want these machines to do things for us, we’re going to have to explain what we want them to do. Today, where would they look? I mean, it’s all over the place.

So I think centralizing that is going to be a fairly major opportunity. I’ve made the statement, I think this is an accurate statement, which is one profession, which has emerged as a really interesting profession in the last few years, which is data engineering. I think that profession’s going to change over the next few years to become business engineering. We’ll begin thinking not just of the way the data is modeled, but of the overall business model and ultimately will derive the data models from it. That’s where this concept of knowledge graphs come in. I believe these things can be very coherent with the large language models.

Soma: Bob, one of the things that you’ve done in the last couple years, at least now, is be an active investor, as I mentioned before, in the next generation of data companies in one way, shape, or form. If you put on your investor hat for a minute, how do you evaluate which company you want to invest in and be a part of kind of thing? What trades do you look for as you make that decision?

Bob: The traditional thing is the best return on your investment, right? That’s probably the traditional focus of most venture capitalists. That’s really not high on my list. I mean, I want something to be successful, obviously, but actually focusing on the actual return is not what I’m caring about as much. I care more about the technology and the impact that the technology is going to have on the industry and the world as a whole. I try and focus my time on companies that I think are doing things that have the potential to have a material change in the way the industry works.

One of the reasons working with infrastructure technologies, it provides the opportunity to do that because they tend to be very horizontal versus just applications which are focused on a given vertical or a given industry. So most of my focus is on horizontals, largely because I’ve always had kind of an infrastructure focus. So in a way, my criteria is about the importance of the technology, the relationship with the entrepreneur and the CEO, and the potential impact for this in the world. That would be where I really put my key focus. To me, the return is a more secondary thing. In fact, one of the questions people always ask is what’s your check size. I’m like, “Really, I follow what you need and what the other investors do.” To me, the important thing is supporting the organization.

Soma: I actually agree with what you said, Bob, about what you look for kind of thing. Because as much as the venture capital industry is about making sure that we have the right returns on our investment, not just for us but also for our LPs kind of thing, to me, focusing on the team and the impact automatically will lead you to great reports of goal along the way kind of thing, right? So I-

Bob: I think they’re very related. I mean, they’re very related together, but it’s just a question of is it a spreadsheet-driven decision or is it more of an impact-driven decision. It’s the latter.

Soma: Got it. Bob, before we wrap up, I want to ask you one final question. If you were to start a company today, what kind of an AI company would you start? I’m assuming it’ll be something to do with data and AI, so I’m presuming that but what would you want to do in today’s landscape?

Bob: I will do this serendipitously through the entrepreneurs that I work with, and I am sort of doing that in a way, but it’s really what I described about that business semantic model, that to me is building high-level semantic models of everything is of interest. I think about this one sometimes, but I haven’t braved myself to go into it yet, which is the semantic models associated with legal contracts and laws because I think there’s a discipline called computational law where you can apply computational approaches to legal problems and really a contract is a program. It’s a program that is interpreted by lawyers and executed by people.

I don’t think that’s the way things are going to work for very long, because, in a fairly short time period, machines will also interpret these contracts, and much of the execution of the tasks will be done by the machines. So I think we’re seeing a transition of high-level semantic concepts into executable form that can be operationalized within the business. So it would be along those lines that I would put my focus right now. Whether the domain is management of the IT infrastructure, which is an interesting domain, or whether it’s a different domain like law, all of those things are potential applications.

Soma: Bob, as always, it was a fun conversation. Thank you for taking the time to be here with us today and really enjoyed the conversation. Thank you.

Bob: Thanks so much. It’s always great talking to you.

Coral: Thank you for listening to this week’s episode of Founded and Funded. If you’re interested in reading Datapreneurs, visit, www.thedataepreneurs.com. If you’re interested in attending our IA summit, visit ia40.com/summit. Thank you again for listening, and tune in in a couple of weeks for our next episode with Cohere’s Ivan Zhang.