Atlassian Survey: 89% of Executives See Employees Speed Up With AI, Only 6% Can Show Real ROI

Everyone's Using AI. Almost Nobody Can Prove It's Working.

Atlassian's Teamwork Lab put a number on something a lot of executives have quietly suspected: their companies are drowning in AI activity with almost nothing to show for it at the organizational level.

Dr. Molly Sands, who heads the Teamwork Lab at Atlassian, laid out the numbers during a fireside chat with VentureBeat senior technology contributor Sam Witteveen at VB Transform 2026. Her team surveyed 12,000 knowledge workers globally and interviewed roughly 200 executives at Fortune 1000 companies.

The headline stat: 89% of executives said individual employees are speeding up thanks to AI. Only 6% could point to specific, clear examples of return on investment.

The 14% Who Figured It Out

Buried in the same data is a more useful number. Roughly 14% of teams had actually converted AI usage into measurable value, according to Sands. That means a single company can simultaneously have a handful of teams crushing it and a majority treading water or worse, all using the same tools.

Sands' team identified three things the high-performing teams had in common: context, workflows, and culture.

On context, the winning teams stopped keeping institutional knowledge in people's heads and started building what Atlassian calls a "context graph," capturing goals, decisions, and organizational knowledge in shared digital records. Products like Jira and Confluence, Sands said, connect work items, goals, and the people doing the work, so AI tools actually have something to draw on instead of guessing.

On workflows, the teams that pulled ahead redesigned entire end-to-end processes instead of just making individual tasks faster. Sands' point here is blunt: if you speed up individuals who are all pointed in slightly different directions, they don't produce more value together. They "very quickly start to crash into each other," as she put it.

On culture, the fastest-moving teams had leaders who explicitly told people it was fine to experiment and fine to fail. That's a real management choice, not a slogan. Most corporate cultures still punish visible failure, even the productive kind.

Constraints, Not More Tools, Drove the Learning

The counterintuitive finding is that the teams seeing the biggest gains didn't get there by giving people more freedom or more tools. They got there by imposing artificial limits on themselves.

Sands described teams breaking every task down into the smallest practical unit of work, a single story point, and teams committing to write zero code by hand for an entire week, forcing total reliance on AI tools to see what broke and what didn't.

"Most of it is not sustainable to do forever, but it is a really, really fast way to learn," Sands said.

A manager skeptical of AI hype could reasonably ask: if these constraint experiments aren't sustainable, what's the actual long-term operating model? Atlassian's research, at least in what's been shared publicly, is stronger on describing what fast learning looks like in a sprint than on proving it scales into permanent productivity gains across a full fiscal year. That's a legitimate open question.

The Real Bottleneck Isn't the Technology

Sands argued the biggest obstacle to AI ROI isn't the models themselves. It's that individual employees are left to figure out AI use on their own, with no coordination.

Every employee ends up building their own prompts, their own workflows, their own assumptions about what the AI tool is supposed to do. Multiply that across thousands of employees and you get exactly the pattern Atlassian's data shows: individual speed, organizational chaos.

This tracks with what plenty of IT leaders have said off the record for the past two years: shadow AI adoption, where employees quietly bring in their own tools and workarounds without IT's knowledge or coordination, is rampant. Atlassian's numbers put a hard figure on the downstream cost of that lack of coordination.

What Comes Next

Atlassian didn't publish a full breakdown of the 200 executives surveyed by industry, revenue size, or geography in the material presented at VB Transform 2026, so it's unclear whether the 6%-ROI figure holds steady across sectors or is skewed by a handful of laggard industries. That's a gap worth watching if Atlassian releases the full State of Teams Report with more granular data.

The bigger open question is whether "context graphs," workflow redesign, and constraint-based experimentation are things most companies can actually execute, or whether they require the kind of dedicated behavioral science team Atlassian itself runs. Not every Fortune 1000 company has a Teamwork Lab. Most don't.