The promises around AI productivity have been everywhere for the past year. Agents will eliminate toil. Work will happen “magically.” The future of work will be reimagined. But what does that actually look like when you deploy it inside an enterprise?
We had the chance to dig into exactly that question during a panel at the 2025 IA Summit in Seattle, sitting down with three leaders who aren’t just talking about agentic productivity—they’re shipping it. Anu Bharadwaj, president of Atlassian, Rob Seaman, interim CEO for Slack at Salesforce, and David Shim, co-founder & CEO of Read AI, brought real experiences and data from hundreds of thousands of users who are living with AI agents in their daily workflows.
What emerged wasn’t just another set of productivity demos. It was a clear picture of how agents are actually being deployed, what works, what doesn’t, and what the experience looks like when you move beyond the hype to real-world implementation. If you’re building in this space, here’s what you need to know about where agents are today and where they’re headed.
Beyond the Assistant: Team-Native AI
The conversation started with a simple but profound shift in perspective. Productivity, Anu argued, isn’t just about helping individuals move faster; it’s about how teams work together with agents as peers, as genuine team members rather than sophisticated assistants.
“When we think about productivity, we think about how we make teams productive, not just individuals,” she said. At Atlassian, users of their agent platform Rovo report productivity increases either through time saved or the ability to do things that they weren’t able to do before, while 78% of users say Rovo gives them more accurate search results than other tools.
But the percentage isn’t the real takeaway, the shift in mental model is.
“You can deploy it in a work context such that working across humans and AI agents becomes as natural as working across a team of humans,” she said. This isn’t about automation or assistance. It’s about creating hybrid teams where AI agents handle workflows that span systems and people.
Rob reinforced this with a compelling data point: The average person has Slack open about 10 hours a day and actively uses it about two hours a day. “If you think of these agents, they’re like a teammate… they should be right there with your colleagues,” he said. The implication is profound: Agents aren’t tools you visit, they’re colleagues you work alongside.
The takeaway: The next wave of productivity won’t come from individual efficiency tools. It will come from systems that treat agents as genuine team members, coordinating, reasoning, and collaborating across people and software.
The Orchestration & Data Imperative
The conversation quickly surfaced a critical challenge: How do you avoid creating sprawl in the agent world like the sprawl that has happened across all the SaaS products that are used in an enterprise? The insight on how to avoid sprawl was about data. Agents are only as good as the information they can access and share, and the data they need to be effective is spread across that sprawl of SaaS products today and multiple agents tomorrow.
David offered the sharpest market prediction, drawing a direct parallel to advertising technology consolidation. “There will be less agents over time, and there’ll be less companies that are managing or orchestrating those agents,” he said. His analogy to The Trade Desk (worth $50 billion for orchestrating advertising across platforms) points to a fundamental truth: The value isn’t in having more agents, it’s in connecting data across systems.
The technical challenge is federation at scale. As Rob highlighted: “How do we effectively share information in a security aware way and share metadata about our agents so that they are aware of each other and can call each other?” Without shared context, agents remain isolated and limited.
At Atlassian, this data integration imperative is already driving user behavior. With Rovo’s 50 data connectors and 3.5 million users, “the outcomes that people achieve are going to be so much better when they have multiple connectors put together,” Anu said. The pattern is clear: Value comes from breaking down data silos, not building better individual agents. “From a customer’s perspective, they just want to unlock the outcome,” she said, and that requires reasoning across all their systems.
For builders: Data integration isn’t a nice-to-have feature, it’s the foundation. Plan for federation from day one. The companies winning aren’t just building better agents—they’re building the connective tissue that makes agent ecosystems possible.
Jobs Nobody Wants (And Why That Matters)
David offered perhaps one of the most pragmatic definitions of agentic value: “We’re taking care of the jobs that nobody wants.”
Read AI, now adding 50K new customers a day for its comprehensive productivity AI platform, started by solving the simple but universal pain of meeting notes, automating summaries, and action items. But the deeper opportunity has been in creating organizational intelligence and proactive AI assistance.
He described a global customer with 200 product managers conducting 1,000 customer interviews a week. “That data hadn’t been shared across the team,” he explained. “Nobody had the time to look at Japan versus the Netherlands versus the US and combine all that data together.”
The real value isn’t replacing the interviews. It’s scaling insights that would otherwise get lost. This is the difference between automating tasks and creating organizational intelligence. The biggest opportunities aren’t in replacing skilled work. They’re in using agents to scale insights that currently get trapped in silos.
Speed as Survival Requirement
The speed of change and innovation in this space was a consistent theme throughout the discussion.
Traditional startup and enterprise planning cycles have collapsed under the weight of AI advancement. As Anu shared: “Things are changing too fast, so let’s wait to respond. That’s not a luxury any of us building technology have… reasoning is getting so much better just over a matter of months … Not years. It’s not about quarterly roadmaps anymore.”
Rob put it bluntly: “Plans at this point are kind of pointless.” He said Slack is using a “prototype the path mentality” to allow them to code as quickly as possible.”
But, the flip side of this acceleration is fragility. “If you have a product and it starts to get some traction, somebody else can come in and add another feature because they can no-code a core part of it, and they add additional novelty… all of a sudden people will start to move over,” David added.
The takeaway: Speed isn’t a differentiator—it’s table stakes. The companies that can iterate, ship, and adapt weekly—not quarterly—will define the next era of work.
Startups vs. Incumbents: Who Wins?
The final question of the session was simple: In the race to agentic work, who has the advantage? Incumbents with data and distribution, or startups with focus and speed?
David didn’t hesitate: “Being nimble is the key. If you build something valuable, the market will come to you and that’s happening at a scale we’ve never seen before.”
Anu offered a balanced warning: “Speed is necessary but not sufficient. Security, privacy, and governance are real moats. Once something goes wrong, it’s hard, especially for a startup, to recover.”
The takeaway: The future of intelligent work won’t be won by size. It will be won by execution. Startups will lead with speed; incumbents will win by pairing velocity with trust.
The Bottom Line
The takeaway from this conversation isn’t just that agents are working. It’s that they’re working in fundamentally different ways than most people expected. The real value isn’t in replacing humans with AI, but in creating hybrid teams where agents handle the organizational intelligence work that was previously impossible at scale.
The winners will be those who build for team productivity, not just individual efficiency. They’ll create agents that play well with others while maintaining enterprise-grade security and governance. And they’ll move fast enough to keep up with a pace of change that’s compressing traditional development cycles from years to months.
The era of agentic work has begun, and the teams that learn fastest will lead it.