AI Users Work Up to 90 Minutes Longer Than Colleagues, ActivTrak Study Finds

Employees who use AI at work log days up to 90 minutes longer than colleagues who don't, according to ActivTrak research based on observed digital activity from 9,140 employees across 467 companies. The finding cuts against the common assumption that generative AI automatically shortens the working day, and it lands in the middle of an ongoing debate among technology leaders about whether the tools actually save time or just shuffle effort around.

The study, published November 14, 2026, examined knowledge workers' activity from April to September. ActivTrak compared workday length, application switching and focus time across four groups: heavy, regular, light and non-users of generative AI tools. It also tracked how employees accessed AI — through browser-based services or through tools embedded in the software they already use.

The stakes are straightforward. Companies are spending heavily on AI adoption, and this data suggests deployment method may matter as much as adoption itself.

How much longer are AI users' days?

The numbers are specific. Workers classified as AI users had workday spans of 8.4 to 9.1 hours, against 7.6 hours for non-users. Heavy users also worked more beyond scheduled hours, adding 18 minutes a day after hours, even though their overall share of after-hours activity looked similar to other groups.

The study's group definitions were precise:

• Heavy users: more than 45 minutes of AI use per day, in at least five sessions
• Regular users: 10 to 45 minutes per day
• Light users: less than 10 minutes per day
• Non-users: no AI use

After-hours work had its own definition. On weekdays, it meant productive screen time outside scheduled start and end times. On weekends, the researchers counted any stretch of less than two hours of productive screen time as after-hours work, reasoning that longer periods could reflect a normal weekend schedule rather than extra effort.

The takeaway: in this dataset, AI use and longer days went together, not apart.

What happened to focus time?

Heavier AI use correlated with more fragmentation. Heavy users averaged 1.1 application switches per hour, compared with 0.95 for non-users. Their uninterrupted focus sessions averaged 37.8 minutes, versus 48.2 minutes for light users — roughly ten minutes of concentrated work lost per session.

Meeting time bucked the trend. Formal meetings took up 3% of work time for heavy AI users, compared with 5% for non-users. So while AI users sat in fewer scheduled meetings, their solo work time was more chopped up.

The takeaway: AI users traded meeting blocks for a constant low-level drain on attention.

What is the 'swivel-chair tax'?

The sharpest contrasts appeared when ActivTrak broke the data down by how people reached their AI tools. Employees using AI embedded directly in their main work applications averaged 0.6 app switches an hour. Those relying on browser-based AI tools recorded between 1.7 and 4.2 switches an hour.

That gap is the operational cost of moving between a core task and a separate AI window. ActivTrak described it as a possible "swivel-chair tax," noting that browser-based AI users switched applications more than four times as often as non-users.

The role-level data reinforced the pattern:

• Software Engineers, more likely to use embedded AI tools, averaged 0.6 switches an hour and spent 78.5% of active time in core coding applications.
• Operations Managers, more likely to rely on browser-based AI, spent 75.5% of active time in core applications and averaged 1.1 switches an hour — rising to 4.1 during periods of heavy coordination.

The takeaway: where the AI lives — inside the workflow or in a separate browser tab — appears to shape how much it disrupts concentration.

Why measure activity instead of asking questions?

Most workplace AI research relies on surveys, which depend on self-reporting and attitudes. ActivTrak took a different route: the analysis was based on observed digital work patterns rather than employee surveys, aiming to measure how AI use relates to day-to-day habits rather than how people feel about the technology.

That design choice matters. Self-reported productivity gains are notoriously optimistic; observed switching and screen-time data is harder to argue with. It also means the study captures behavior people might not notice themselves, such as the slow accumulation of after-hours minutes.

The takeaway: watching what workers actually do, rather than what they say, produced a less flattering picture of AI's day-to-day effects.

What did ActivTrak's leadership say?

Gabriela Mauch, Chief Customer Officer and Head of the ActivTrak Productivity Lab, said the findings exposed a gap between using AI and integrating it into routine work. "AI may help employees complete tasks faster, but our research shows it contributes to longer workdays, more application switching and shorter focus sessions," Mauch said.

Her advice to leadership was pointed. "For leaders, the goal can't simply be AI usage; they need to treat AI usage as a means to measurable gains in productivity, efficiency and business performance," she added.

What does this mean for AI deployment?

The analysis feeds a wider debate: does AI save time in practice, or does it redistribute effort across the working day? The ActivTrak data suggests a concrete mechanism for the second outcome. Task-level gains can be offset in two ways — employees repeatedly moving between systems, and AI use extending work beyond scheduled hours.

For technology leaders, the study's most actionable finding is about deployment method. A tool built into an existing workflow appears less likely to interrupt concentration than one accessed through a browser tab or separate interface. Engineers with embedded AI averaged a fraction of the switching seen among browser-tool users.

The study covers 9,140 employees at 467 companies over six months, observed rather than surveyed — a substantial behavioral dataset by workplace-research standards. It doesn't establish that AI causes longer days; correlation and causation remain distinct. But it gives leaders a measurable question to ask before the next AI rollout: not just who is using the tools, but where the tools sit in the workday.