A new study from Atlassian’s Teamwork Lab highlights the tensions at the heart of AI in the workplace. Although adoption has been nearly universal, acceptance has been much slower.
Everyone does it, but the honest ones are punished.
Secrecy around AI in the workplace is officially being lifted. 94% of knowledge workers report using AI in the last month (two-thirds use it regularly), but they don’t hide it. Nearly 80% are completely transparent with managers and co-workers.
Still, researchers have found that disclosing the use of AI comes with significant professional costs. Even when evaluating the same work, colleagues rate AI users as: 10 times lazier and, 24 percentage points less likely to recommend This shows that although secrecy around the use of AI is waning, the stigma remains deep.
These insights come from a controlled experiment with 961 U.S. knowledge workers, conducted alongside a pulse survey of 1,006 people on AI adoption and attitudes. In the experiment, workers rated identical works. The main variable was whether they were told that an AI helped create it. When the use of AI was revealed, evaluators found its creators to be grossly lazy, unprincipled, and unworthy of recommendation.
Taken together, these data reveal that employees are adopting AI faster than their culture can absorb it.. Traditional workplace norms, like how much to share in a meeting, when to CC your boss, and how to take credit gracefully, were established through years of observation and instillation. But AI doesn’t have that runway. Data shows that employees are improvising new etiquette in real time, and guessing wrong can have real costs.
“This isn’t a technology problem; it’s a social and cultural problem,” says Molly Sands, director of the Teamwork Lab. “Companies are telling their employees to use AI, but employees are punishing each other for being honest about AI. There will be no AI strategy until leaders change the culture. AI is contradictory.”
AI penalties are real, but not inevitable. Here’s how to shrink it:
A framework that AI uses “for the team” rather than “for itself”
Compared to employees who said using AI “saves me time,” those who viewed using AI to “support my team and clients” were 11 points more likely to put in effort (45% → 56%) and 8 points more likely to be recommended (43% → 51%). However, both remained far below workers who did not mention AI at all.
Culture is the real differentiator
In companies that actively celebrate the use of AI, colleagues and leaders model it, wins are highlighted, the stigma of laziness all but disappears, and AI users are actually rated as more efficient than their non-disclosing colleagues.
Previous research, including Atlassian’s own research, consistently shows that cultures celebrating AI are concentrated in pockets: the technologically advanced industries and sectors where AI tools have been embedded the longest. In other words, the same behavior that earns trust in one workplace may arouse suspicion in another.
“The disparities between workplaces are greater than many people realize,” Sands says. “In my world of working in the tech industry, I have to explain myself if I’m not using AI at all. But these findings make it clear that my experience is an exception. Most workers live in environments where that same openness is a drawback.”
takeaway
Business leaders have spent years solving the AI adoption problem, but the battle has already been won: 94% of knowledge workers are using the technology. A new and more difficult problem is cultural influence. It’s an environment that penalizes employees for the integrity they claim to value.
It’s not just a moral issue. It’s a strategic risk. If workers learn that being honest about AI will get them labeled as lazy, a reasonable response is to stop talking about AI. And organizations whose use of AI goes underground lose the ability to learn from AI, extend what works, and figure out what doesn’t.
Sands says workers can’t solve this problem on their own. “Saying you’re using AI for your team may soften the stigma, but it won’t erase it. The solution is culture: leaders who use AI visibly, teams that treat AI as a shared tool rather than a shortcut, and organizations that reward outcomes rather than optics.”
methodology
This study utilizes two complementary data sources:
controlled experiment. We conducted the experiment with 961 full-time knowledge workers in the United States from March 30 to April 7, 2026. All participants were asked to evaluate the same work product (an email summarizing a business proposal) from the same virtual colleague. The only variable was a short note explaining how the work was completed. Either there was no mention of AI (control) or there were AI disclosures with varying levels of detail and motivations. Participants then rated their coworkers on several dimensions, including perceived effort, competence, laziness, and how likely they would recommend the coworker to senior leadership for high-profile projects. We also asked participants to describe their team’s norms regarding AI, from “forbidden” to “celebrated,” allowing us to test whether workplace culture mitigates AI penalties.
Pulse investigation. From April 24 to May 6, 2026, we conducted a double-blind survey of 1,006 U.S. knowledge workers about AI adoption, usage patterns, disclosure behaviors, and attitudes. Respondents ranged across industries, functions, seniority levels, and generations. The survey measured how often employees use AI, the depth of AI integration (from ad-hoc use to team-wide standards), and how openly they share their use of AI with managers and colleagues.
