Eighty percent of employees now use AI tools at work, up from 53% two years ago. But buried in that adoption surge is an uncomfortable number: only 3% of those users are actually hitting the productivity levels that justify the hype.
That finding comes from ActivTrak's 2026 State of the Workplace report, which analyzed 443 million work hours across 1,111 companies and 163,638 employees from January 2023 through December 2025. The AI-specific analysis tracked 10,584 users across 376 companies, comparing behavior 180 days before and after AI adoption.
The headline numbers look great on a slide deck. The details tell a messier story.
More AI, More Work - Not Less
Here is the finding that should worry every manager banking on AI to reduce workload: after employees adopted AI tools, not a single activity category decreased in time spent. Email time jumped 104%. Chat and messaging surged 145%. Business management tasks grew 94%.
AI is not replacing work. It is generating more of it. Every AI-drafted email creates a reply chain. Every AI-generated document needs review. Every AI suggestion spawns a conversation about whether to accept it. The report calls this the "amplification effect," and the data backs it up clearly.
The 7-10% Rule
The most actionable finding in the report: employees who spend 7-10% of their work hours using AI tools hit 95% productivity rates, the highest of any usage bracket. That translates to roughly 35-50 minutes per day in AI tools for a standard workday.
The problem is that 57% of AI users spend less than 1% of their work hours in AI tools. They have access, they have signed up, but they are barely using it. Meanwhile, the average organization now runs seven different AI tools, up from two in 2023. ChatGPT dominates usage at 27 times more activity than Cursor, the second most popular tool.
Companies are buying AI licenses faster than employees are learning to use them effectively.
The Disengagement Paradox
Burnout risk dropped 22%, falling to just 5% of employees. That sounds like a win until you see the other side: disengagement risk climbed to 23% of employees, a 21% single-year increase. Workers are not overwhelmed. They are under-challenged.
The report suggests AI may be hollowing out the interesting middle of many jobs - the problem-solving, the creative decisions, the judgment calls - while leaving behind the rote coordination work that nobody finds fulfilling. Employees are working shorter days (8 hours 44 minutes, down from 8 hours 53 minutes) and getting more done in those hours (productive time up 5%, with 19 extra productive minutes daily). But they are also losing the ability to focus deeply - average focused work sessions dropped 9% to just 13 minutes and 7 seconds.
One more data point that jumped out: weekend productive hours increased 46% on Saturdays and 58% on Sundays. Workers are starting their Saturdays at 7:11 a.m., nearly 90 minutes earlier than before. AI may be making the workweek more efficient, but it appears to be bleeding into the weekend too.
The real takeaway from this report is not about adoption rates - that battle is already won. The gap is between having AI tools and knowing how to use them well. With 80% adoption but only 3% of users in the optimal zone, most organizations have an AI training problem, not an AI buying problem.