Often, yes. ActivTrak's October 29, 2026 analysis of 9,140 knowledge workers found AI users' workdays span 8.4 to 9.1 hours against 7.6 for non-users, and heavy users' focus sessions average 37.8 minutes against 48.2 for light users. Browser-based AI meant 4.2 app switches an hour; embedded AI, 0.6. Kahneman's point that task switching costs effort says: bring AI into your tools.

The newest number on AI and attention does not come from a survey. It comes from software that records what people actually click, which makes it harder to argue with and easier to misread.

On October 29, 2026 the ActivTrak Productivity Lab released an analysis of 9,140 knowledge workers across 467 enterprises, built from digital activity logged between April 1 and September 25, 2026 rather than from self-reports. It sorted people by use: heavy AI users spend more than 45 minutes and five or more sessions a day in AI tools, light users under 10 minutes, regular users somewhere in between. AI users' workday spans ran from 8.4 to 9.1 hours against 7.6 hours for non-users, which is where the headline's "up to 90 minutes" comes from. Heavy users switched applications 1.1 times an hour, 15.2% more often than non-users at 0.95, and their uninterrupted focus sessions averaged 37.8 minutes against 48.2 minutes for light users, a 22% gap. They also spent less of the day in formal meetings: 3% of work time, against 5% for non-users.

The finding I would act on sits further down the release. Employees using AI embedded inside their main work application averaged 0.6 app switches an hour. Those using browser-based AI tools averaged between 1.7 and 4.2, and at the top of that range they switched more than four times as often as people who use no AI at all. ActivTrak calls this a "swivel-chair tax". Its own role comparison makes the point: software engineers, who tend to use AI inside their editors, average 0.6 switches an hour and spend 78.5% of their active time in core coding apps, while operations managers, who lean on browser tools, average 1.1 switches an hour, rising to 4.1 during heavy coordination.

This fits ActivTrak's larger 2026 State of the Workplace report, built on more than 443 million hours of activity from 2023 to 2025. There, AI users' daily focus time fell 9% while non-users' barely moved, the average focused session shrank from 14 minutes 23 seconds to 13 minutes 7 seconds, and the employees with the highest productivity spent 7 to 10% of their work hours in AI tools. Only 3% of users sat in that band.

Daniel Kahneman's Thinking, Fast and Slow explains why a switch count deserves more attention than it gets. System 2, the slow and deliberate thinking that strategy, pricing and careful writing depend on, is effortful and allocated on demand, and in his chapter on attention and effort Kahneman notes that switching from one task to another is itself effortful, especially under time pressure. Every round trip from a document to a chat tab and back spends some of that budget before the real work starts. A depleted System 2 is also a lazy one: it tends to endorse whatever fluent answer is on the screen. A founder who ends a nine-hour day of tab-hopping has not only lost an evening. The late decisions are the ones made with the least scrutiny.

The one-window rule (three steps for the next two weeks)

  1. Measure your span, not your hours. For ten working days, note the time of your first and last work action. ActivTrak's gap is in span. If yours has grown since you adopted AI, that is the number to bring down, not your hours of "real" work.
  2. Move routine AI work inside the tool where the work lives. Drafting in the document, analysis in the spreadsheet, code in the editor, using the assistant built into that tool (Claude Pro, for example, includes Claude in Chrome and Microsoft 365). The 0.6 versus 1.7 to 4.2 split is the clearest lever in the data.
  3. Batch the browser chat. Keep the standalone chat closed during a focus block and open it at set points, with your questions written down first. As a rough budget, ActivTrak's most productive band, 7 to 10% of an eight-hour day, works out at about 34 to 48 minutes in AI tools.

Now the limits, because they are real. This is correlation: heavy AI users may simply hold the longest, most coordination-heavy jobs, and ActivTrak does not claim AI caused the longer days. The sample is enterprise employees on monitored machines, not founders running their own calendars. ActivTrak sells the software that measures all of this, so read it as an informed vendor rather than a neutral lab. "Up to 90 minutes" is the top of a range; the low end, 8.4 against 7.6 hours, is 48 minutes. And the 7-to-10% band comes from an older dataset than the October analysis. What survives all of those caveats is the embedded-versus-browser comparison, because it compares how people use AI rather than whether they do, and that is the part you control tomorrow morning.

For the workload side of the same pattern, see what Jabra's October 2026 study found about AI giving people more work.

Sources: ActivTrak Productivity Lab: AI Users Have Workdays Up to 90 Minutes Longer Than Non-Users, PR Newswire, October 29, 2026; ActivTrak Productivity Lab: AI Is Accelerating Work, Not Replacing It, ActivTrak, 2026; AI Doesn't Reduce Work, It Intensifies It, Harvard Business Review, February 9, 2026; Plans and pricing, Anthropic, checked October 30, 2026.

Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · What University Will Not Teach You