AI users have longer workdays, up to 90 minutes longer than everyone else in the business, and the reason has less to do with how good the tool is than where it sits on their desktop.
ActivTrak, an Austin, Texas workforce analytics vendor, measured digital work activity from 9,140 knowledge workers across 467 enterprises between 1 April and 25 September 2026. Its Productivity Lab found that AI users logged workday spans of 8.4 to 9.1 hours, against 7.6 hours for people who do not use the tools. ActivTrak sells workforce analytics software, so the study is vendor research and deserves the scrutiny that comes with it. The figures come from behavioural screen data rather than a self-reported survey, which is why they are worth setting out here. ActivTrak is explicit that the findings show an association between AI use and longer days, not proof that AI causes them.
AI Users Have Longer Workdays When the Tool Lives in a Tab
The study’s sharper number sits inside the first one. Employees using AI built into their main work application, a coding assistant inside an IDE, for example, averaged 0.6 application switches an hour. Employees opening AI in a separate browser tab switched 1.7 to 4.2 times an hour, against 0.95 for people who do not use AI at all. At the top of that range, that is more than four times the non-user rate. ActivTrak calls the cost of that movement between windows a “swivel-chair tax.”
Role data makes the split concrete. Software engineers, who use embedded AI more often, spent 78.5 per cent of their active time inside core coding applications and switched 0.6 times an hour. Operations managers, who lean on browser-based AI tools, spent 75.5 per cent of their time in core applications but switched 1.1 times an hour on average, rising to 4.1 during heavy coordination periods.
The longer day shows up after hours too. Heavy AI users spent roughly the same share of their day working outside scheduled hours as non-users did, but because their overall days ran longer, that share translated into about 18 extra minutes of after-hours screen time daily. A tool sold on saving time was, for this group, quietly extending it.
Focus Pays the Price First
ActivTrak defines a heavy AI user as someone logging more than 45 minutes of use across five or more sessions a day. That group held uninterrupted focus for 37.8 minutes on average. Light users, under ten minutes a day, held it for 48.2 minutes, a gap of 22 per cent. Formal meetings fell too, from 5 per cent of the working day for non-users to 3 per cent for heavy AI users.
Gabriela Mauch, chief customer officer and head of the ActivTrak Productivity Lab, put the finding plainly: “AI may help employees complete tasks faster, but our research shows it contributes to longer workdays, more application switching and shorter focus sessions.” Her second point is the one a PMO can act on: “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.”
The Decision Leaders Actually Control
Microsoft, Google, Salesforce and Adobe are all racing to embed AI assistants directly into the applications people already have open. That race matters more than it looks. ActivTrak’s own data ties the integration point, rather than the AI model itself, to how much of the day gets eaten by switching between windows, and it stops short of saying placement causes the full 90-minute gap. The switching numbers behind that link come from direct measurement, which is a firmer footing than the overall workday figures can claim.
A leader rolling out a new AI capability rarely asks where the tool will sit inside someone’s existing workflow. The licence gets approved, the browser extension gets installed, and the first app-switching cost lands within the week.
Before the next AI tool reaches someone’s desktop, check two things: whether it sits inside software the person already uses for most of their day, and whether anyone is tracking focus-session length alongside adoption numbers. Usage counts tell you the tool got turned on. Switching counts tell you what it cost to use it. A PMO that reports adoption percentages to its steering committee without either figure is reporting half the story.
LUTT has argued before that AI rollouts are won or lost one layer below the steering committee, because middle management decides how a tool actually gets used day to day. ActivTrak’s data adds the half that argument left out: where a tool sits on the screen is tied to how long the working day runs.
The next AI licence request is also a workflow design decision, whether or not it arrives labelled as one.
