In March, Madrona celebrated our 30th anniversary and also hosted our annual limited partner meeting. As part of this, we hosted a panel with three leaders building Applied AI companies, unpacking how AI is being adopted across functional teams and what’s working beyond the hype.
We were joined by:
Linda Lian, the Co-Founder and CEO of Common Room, which is building a customer intelligence platform for go-to-market teams; Gaurav Oberoi, the Co-Founder and CEO of Lexion, a legal AI company that was acquired by Docusign last year; and Anna Veronika Dorogush, the Founder and CEO of Recraft, an AI design company building a model and workflows for professional graphic designers.
The conversation centered around three core themes:
- Why Now? What is happening now that is leading to greater AI adoption
- Is AI driving real business results and ROI?
- How are business models evolving to support this shift?
What’s clear is that AI is no longer a side project, but a foundational part of how teams operate. The playbook is still being written. The teams that will win are the ones that deliver value quickly, build trust through measurable outcomes, and stay flexible in how they price, package, and scale.
Below is a recap of our great conversation with Linda, Gaurav, and Anna!
Why Now? AI is No Longer Optional
The key question we opened with was: Why is now the moment for AI in enterprise workflows? The TL;DR: pressure is high, and excuses are low to not adopt these solutions.
In sales, Linda shared that pipeline generation and efficiency have officially become board-level concerns, and AI is one of the few levers left to pull. “Time is kind of up,” she said. “We’re seeing growth-stage public companies that used to expect double-digit growth now dipping into the single digits. Efficiency isn’t just a theme – it’s a mandate.”
Common Room sells to many Enterprise customers (e.g., MongoDB, Snowflake). Linda shared how teams in the acquisitions are being asked to 2X the business with just 30% headcount growth, and those new hires are more junior. She shared how difficult it is to hit these growth targets without AI, commenting, “that kind of squeeze forces teams to leapfrog. AI isn’t a luxury in that environment – it’s infrastructure.”
In legal, Gaurav pointed out that adoption is finally happening because the tools are genuinely good now. “When we started Lexion, I’d hear the same story again and again – it takes five days to find a contract before a big meeting. That’s a simple search problem, but legal was so far behind. Now, AI can actually deliver a first pass at contract review that’s good enough to save real time and money.”
For design, Anna painted a picture of bottom-up meets top-down sales motions. Sometimes, a couple of designers start experimenting with AI tools on their own, then leadership hears about it and reaches out for a broader demo. Other times, entire AI evaluation teams at large companies get involved. “It’s no longer just one creative trying a cool tool,” she said. “Companies are looking at this seriously.”
Measuring ROI: What Does “Working” Even Mean?
The second topic we dug into is understanding ROI. We are starting to move beyond the ‘experimental’ phase of AI, and teams are now really starting to measure how much ROI is improving organizations. So, how do teams actually know AI is helping?
Gaurav shared how legal teams tend to think about ROI in terms of either acceleration or cost savings. This could reduce outside counsel spend or help sales teams close deals faster. “When we show teams how many manual steps AI can remove,” Gaurav said, “you can see the lightbulb go off. This isn’t theoretical – it’s measurable.”
Design ROI, as you’d expect, can be more difficult to measure given the subjectivity in the field, but the impact is no less real. Anna shared a story about a freelance designer who used Recraft to create visuals for a client presentation in just two hours – work that would normally take a week. “She got paid €4,000 and went on vacation,” but “that’s what ROI looks like to an individual.”
Inside larger design teams, the savings show up in different ways, such as eliminating weeks-long outsource cycles or avoiding stock imagery entirely. “Sometimes it’s just about being able to get exactly the slice of pizza you want,” referencing one customer’s joy at now generating unique images on the fly.
In sales, it’s all about pipeline. Linda shared that one customer booked over $2 million in pipeline in the first four weeks of using AI-powered prospecting workflows. “We’re still in early days, but these quick wins matter,” she said. “The goal is always speed to value.”
New & Emerging Business Models: Pricing is Still a Mess, but We’re Early!
The last topic we explored was probably a candid conversation around business models. Everyone agreed: pricing AI products is still a work in progress!
For design tools like Recraft, seat-based pricing didn’t work; people didn’t want to pay a flat rate when they weren’t sure what results they’d get. Usage-based pricing (per image or in tiers) is more common now, but even that comes with tradeoffs. “The biggest issue is when users pay for multiple generations [of an image] but don’t get what they want,” Anna said. “It’s totally fair – they’re spending money but not solving their problem.”
Anna talked about the ideal world for pricing, which was charging per task completed, but as she admitted, it’s hard to define what “done” means because it’s all subjective. “We don’t yet have a great way to measure ‘solved.’”
Gaurav added a broader enterprise perspective, comparing his experience pricing at Lexion to what he now sees at DocuSign. At startup scale, they could charge a platform fee and not worry too much about compute costs. But at a company with 1.6 million customers, the math really starts to matter. “When customers start running heavy AI workloads on millions of documents, suddenly your cost structure can break,” he said. “You want to give them value, but it has to be sustainable. These things are weirdly at odds – you make a feature better, and it costs you more.”
Linda also mentioned the challenge of blending automation with personalization in GTM, both of which are expected from customers, but require very different backend cost models. Their pricing combines seats with a usage-based lever, but it’s constantly evolving.
Conclusion
It was an absolute treat to have Linda, Gaurav, and Anna speak so candidly about the progress that’s been made in Applied AI as well as where the industry is heading. It’s clear we have moved past the “initial euphoria” point to now where AI applications are squarely being used inside enterprises, scaling companies, and at the individual level. While there is no shortage of exciting AI tools out there, we also believe there will be tremendous ROI generated for customers as these tools become more embedded and
Thank you, Linda, Gaurav, and Anna, for joining us.