Knowledge workers have real concerns about AI’s impact on the world. They also think it’s making their own work better, and aren’t willing to give it up. New research from Atlassian’s Teamwork Lab finds these aren’t contradictory positions; they’re coming from the same people.
Key findings
- AI is less divisive at work than the headlines suggest. 74% of knowledge workers say AI positively impacts their work, despite 64% also having concerns about its broader impact on society.
- Workers would give up personal AI before work AI. If they had to choose, knowledge workers are twice as likely to drop AI in their personal lives as at work — a preference that holds across every generation and job function.
- Just one in five workers think they’d struggle without AI. Yet half would fight to keep it. Only 19% of AI users say it would be hard to give up at work, and 57% think they’d perform just as well without it. But when it comes to losing access, 50% say they’d push back if their employer cut their AI tools.
- When we took AI away for a day, workers reached for it anyway. In an internal Atlassian experiment, 82% felt the urge to use AI on their assigned no-AI day — and nearly 1 in 4 reached for it without thinking. On their “use AI to the max” day, priority work felt 44% easier and stress was 22% lower.
AI is controversial at home, but table stakes at work
AI is one of the most contested topics in public life, and heavy users share those concerns. Yet that controversy is largely skipping over knowledge work: the same workers who have real societal questions about AI find it genuinely valuable in their own jobs.
Inside the office, AI is only becoming more integrated into work. Our latest Pulse survey, fielded in late June and early July, found a 90% topline adoption rate among knowledge workers — including a seven-percentage-point jump in heavy users* — and 74% with positive sentiment about the technology’s role in their work. That’s a notable contrast with signs of a consumer pullback in people’s personal use of AI, where some are consciously limiting it or feeling the novelty fade.
“Many of the same people raising real concerns about AI are the ones relying on it every day,” observed Molly Sands, future-of-work expert and head of the Teamwork Lab. “The striking part is what happens when you ask them to give it up: they won’t. They’ve decided, at the desk level, that AI earns its place.”
*Heavy users are those who describe AI use at work as their “new normal with redesigned workflows” or their “team’s standard with AI embedded in shared workflows.”
AI matters more at work than at home
AI’s true place in knowledge work is revealed less by what people say about it than in how they defend it. Asked to give up AI at work or at home, workers are twice as likely to keep it at work — a preference that holds across seniorities, generations, and job functions.
AI is so critical to knowledge workers today that half of users say they would resist efforts to revoke or curtail their access, with nearly 10% going as far as to look for another job.
Deeper use, deeper buy-in
Most workers would describe AI as helpful rather than indispensable. Just 19% say it would be hard to maintain their performance without it, and 57% think they’d perform just as well.
And yet, half say they’d push back if their employer cut their access, and that pull gets stronger the deeper the use goes. “Team standard” users — those whose AI use is embedded in shared, collaborative workflows — are more than 10x as likely as occasional users to say going back would be hard, and users of AI that consistently delivers high-quality results are 3x more likely to fight back.
The tell isn’t what workers say, but what they reach for
To see this dynamic up close, Teamwork Lab ran an AI Play/Pause experiment with a group of Atlassian talent acquisition specialists: one workday pushing AI use to its maximum, and one workday foregoing it entirely. Because these teams were regular users of Rovo — connected to the Teamwork Graph’s rich context about their work — we could isolate what actually changes when access to helpful tools and information goes away:
- 82% of participants felt the urge to use AI on the Pause day, and 23% reached for it without thinking
- Priority work felt 44% easier and stress was 22% lower on the AI Play day
But the Pause day was uncomfortable: 47% found tasks harder than usual, and 46% said their non-AI skills felt rusty. This speaks to a common concern that AI could erode human capability. The real question from the Pause day results is whether the urge to reach for AI reflects value or dependence. For managers, that distinction matters: the same behavior can signal a tool that’s earning its keep or a habit that isn’t.
That’s exactly what the AI Play/Pause experiment forced people to confront. After the experiment, 92% of participants said they felt more confident about when not to use AI. As one professional noted, “It reminded me to trust my own expertise more. I’ve been reaching for AI as a starting point even for things I’m already good at.”
The real risk is reliance without judgment — and judgment is something teams can build. That’s what these workers are doing: they’ve made AI part of how they work, and now they’re mapping where its role ends and their own begins.
–> Use the AI Play/Pause Day Play to assess your own team’s AI discernment.
“The fear that AI erodes capability assumes people are passive recipients of it,” said Sands. “What we actually see is workers who use AI well are building better judgment about its role — figuring out where it genuinely helps and where it doesn’t.”
Methodology
This research draws on two complementary data sources:
Pulse survey. Between June 26 and July 7, 2026, we conducted a double-blind survey of 1,001 U.S. knowledge workers on their AI adoption, usage patterns, perceptions, and attitudes. Respondents spanned industries, functions, seniority levels, and generations. The survey measured how frequently workers use AI, the depth of their integration (from ad-hoc use to team-wide standard), and how they assess the role and value of AI in their work and personal lives.
Controlled experiment. On June 9 and 10, 2026, we conducted a two-day experiment with approximately 123 Atlassian talent acquisition professionals. On the first day, participants were instructed to use AI for everything they can in their work, within defined guardrails. On the second day, participants were instructed to refrain from using AI in their work. At the end of each day, participants reflected on their experience by recording a brief Loom video and completing a brief survey. Participants also completed a post-experiment team reflection and survey.
