ActivTrak Study Links AI Use to Workdays Up to 90 Minutes Longer

ActivTrak's October 2026 report found workday spans of 8.4 to 9.1 hours for AI users against 7.6 hours for non-users. Heavy users and people on browser-based tools show the most fragmented work. The comparison is between groups, so it cannot show that AI adds hours.

Employees who use generative AI at work have longer workdays than colleagues who do not, according to a report from ActivTrak's Productivity Lab released Oct. 29, 2026. The company reported workday spans of 8.4 to 9.1 hours for AI users, against 7.6 hours for non-users, a gap it describes as up to 90 minutes a day. The finding landed as a direct challenge to the promise that AI hands time back to workers.

The result is real as a measurement, but narrower than the headlines suggest. The study compares different groups of people at one point in time; it does not follow the same employees before and after they adopt AI. ActivTrak does not claim that AI causes the longer days, and the company sells workplace-monitoring software, a commercial interest that shapes how the findings are framed.

What ActivTrak measured, and in whom

According to ActivTrak's press release, the research covered more than 9,000 employees at 467 companies between April 1 and Sept. 25, 2026. Other outlets give the exact count as 9,140 knowledge workers across 467 enterprises. The sources found for this article do not specify where the employees are located, and the data comes from ActivTrak's own customer base rather than a random sample of workers.

ActivTrak sorted employees into tiers by how much they use generative AI tools. The release defines the groups this way:

• Heavy users: more than 45 minutes of AI use and at least 5 sessions a day.
• Regular users: 10 to 45 minutes a day.
• Light users: sparing use for quick tasks, under 10 minutes a day.
• Non-users: people who do not use generative AI tools.

The central metric is "workday span," the term ActivTrak's release uses. The full definition could not be retrieved, so it is unverified, but the figure is probably elapsed time across the day rather than time spent actively working. That distinction matters: a longer span does not necessarily mean more hours of effort.

ActivTrak's own write-up of the October study was not available for review. The details here come from the press release as republished by news-wire mirrors and from news coverage. Outlets such as TechRadar and Gizmodo draw on the same release, so their agreement does not amount to independent confirmation.

The headline numbers and what they mean

The 8.4 to 9.1 hour range for AI users is a range, not one number, and the "up to 90 minutes" claim is its top end. Subtracting the non-user figure of 7.6 hours gives a gap of roughly 50 minutes at the low end and 90 minutes at the high end. A secondary summary reports unrounded figures of 8.42 to 9.08 hours against 7.58, which fit that arithmetic, though that source is weak.

The sources do not say which groups sit at each end of the range. Some outlets describe the whole range as applying to AI users, while others tie it to heavy users. It is plausible that the two ends reflect usage tiers, but without the full report that remains an inference.

The after-hours finding

ActivTrak reports that heavy users spend almost the same share of their workday outside normal hours as non-users do: 32.3% versus 31.9%. Yet because their days are longer, that share translates into about 18 more after-hours minutes daily. In other words, the extra time is not mainly a shift of work into evenings; the whole day appears to stretch.

Whether that stretching reflects an earlier start, a later finish or breaks spread through the day is not clear from the material available. The identical after-hours share is consistent with proportional stretching, but that is a reading of the numbers, not something ActivTrak has demonstrated.

Which employees are most affected

Heavy users

Heavy users are the group that stands out in ActivTrak's release. They switch between applications 1.1 times per hour, which ActivTrak puts at 15.2% more often than non-users, who switch 0.95 times. (The two rounded rates imply a slightly larger gap, so the company's figure presumably rests on unrounded values.)

Focus appears shorter as well. Uninterrupted focus sessions averaged 37.8 minutes for heavy users and 48.2 minutes for light users, a 22% difference. The comparison is between heavy and light users, not between heavy users and people who avoid AI altogether, so it says less about AI versus no AI than the headline might imply.

Two further figures come from secondary coverage. A summary on the AI Commission site, attributed to ActivTrak's research, says light users spend more active time in core job applications than heavy users do, 81.4% versus 78%. A news report of the study, relayed by Mezha, says heavy users spent 3% of their day in meetings against 5% for non-users; that number rests on a single derivative source.

Embedded versus browser-based tools

How the AI is delivered seems to matter as much as how often it is used. Employees whose AI is built into their main work applications averaged 0.6 application switches per hour, fewer than non-users. Those relying on browser-based tools switched between 1.7 and 4.2 times per hour, and at 4.2 that is more than four times the non-user rate.

ActivTrak calls this a "swivel-chair tax," with hedged wording suggesting that moving work between a separate AI tab and the primary application may carry a cost. The data shows the switching pattern; it does not establish what that switching costs workers in time or effort.

Job roles

Role-level evidence is thin. In TechRadar's account of the release, software engineers and operations managers are among the heaviest AI users. Engineers tend to use embedded tools, averaging 0.6 switches per hour and spending 78.5% of active time in core coding applications.

Operations managers lean toward browser-based tools and spend 75.5% of active time in core applications. They average 1.1 switches per hour, rising to 4.1 during periods of heavy coordination. Both role findings concern switching and focus. No breakdown of the workday-length gap by role, seniority, age, industry or company size turned up in the sources found, so the question of which employees work the longest hours is, for now, unanswered.

Why the study cannot say AI causes longer days

The design is observational and compares groups at a single point in time. People who already work long hours might be the ones who adopt AI most readily. AI might also encourage more work, or reviewing and correcting AI output might eat time. Differences in job type, seniority, company culture and remote status could contribute too, and the data cannot separate these explanations.

Selection is a second concern. The employees come from organizations that buy an employee-monitoring product, and such organizations may differ systematically from others. Classifying someone as a non-user is also imperfect, since a person could use AI on a personal device and still be counted in that group.

The third is incentive. ActivTrak sells monitoring and "workforce intelligence" products, and its release argues that leaders should measure how deeply AI is integrated into work, not just how much it is used. That is a reasonable point, and it is also one that favors the company's business. The study has not been peer reviewed, and no methods appendix, statistical tests or confidence intervals were found in the material available.

The window is short as well, about five months, so it cannot show whether the gap persists. Productivity also complicates the picture: ActivTrak reports both longer days and higher productive hours for AI users, and the activity data cannot show how much valuable work was completed.

How the finding fits with other evidence

ActivTrak's other reports

The closest thing to a within-person comparison from ActivTrak is a before-and-after cohort of 10,584 employees at 376 companies, tracked for 180 days before and 180 days after adopting AI. According to ActivTrak, time in every measured work category rose, by between 27% and 346%, and none fell. Email time rose 104% and chat time 145%. That shows added activity after adoption, though not directly longer days.

The same company's other data points the opposite way in places. A September 2026 ActivTrak report says AI users averaged 6h34m of productive hours a day in the second quarter of 2026, against 6h17m for non-users. It also says users at "Stage 3," meaning deep integration, logged a shorter average workday span (6h46m) than Stage 2 users (7h05m). Only 2% of the workforce, the report says, has reached the stage where AI transforms workflows.

A separate analysis of ActivTrak data points to a possible "sweet spot." Employees spending 7% to 10% of work hours in AI tools show the highest productivity, at 95%, but only 3% of users fall in that range. The largest segment, 57%, spends under 1% of work hours in AI tools.

Some outlets describe the October results as contradicting an earlier ActivTrak finding that AI cut workdays by about 9 minutes. That framing deserves caution. The 9-minute figure matches the broad 2023 to 2025 drop in the average workday, from 8h53m to 8h44m across 163,638 employees at 1,111 organizations, which ActivTrak reported in March 2026. That decline was not specific to AI users in the pages reviewed.

One puzzle remains unresolved. The 7.6-hour figure for non-users in the new study is lower than ActivTrak's 2025 overall average of 8h44m. The definitions, the cohorts or the way span is calculated may differ, which makes comparisons across the reports risky.

A field study of work intensification

Independent support for the workload-creep idea comes from a Harvard Business Review article by Aruna Ranganathan and Xingqi Maggie Ye of Berkeley Haas, published in February 2026. Their in-progress study followed a technology company with 200 employees for eight months, with fieldwork running April to December 2025. They reported that employees worked faster, took on a broader range of tasks and stretched work into more hours of the day, often without being asked.

The researchers identified three mechanisms: task expansion, blurred boundaries between work and home, and more multitasking. An Inc. report on the research describes a feedback loop in which AI sped up certain tasks, which raised expectations for speed. The caveats are substantial: one firm, qualitative methods, unfinished work, and publication in an editorial outlet rather than a peer-reviewed journal. One aggregator gives a different headcount of 40 workers, which may refer to interviews, but that is unconfirmed.

Studies that find time savings

Research that asks about time saved paints a more favorable picture, with limits. In a national survey analyzed by Alexander Bick, Adam Blandin and David Deming for the St. Louis Fed and NBER, 33.5% of daily generative AI users said it saved them four or more hours, versus 11.5% of people using it one day a week. Across all workers, including non-users, generative AI saved 1.4% of total hours. These are self-reported figures from a working paper, not workplace logs.

A Microsoft Copilot field study (NBER working paper 33795) used logged behavior. Workers who used Copilot in more than half of the sample weeks spent 3.6 fewer hours, or 31% less time, on email each week; the intent-to-treat estimate was 1.3 hours. A Korean worker survey, whose peer-review status is unconfirmed, found that 51.8% of workers use generative AI for work and that it reduces working time by 3.8%.

What happens to saved time is its own question. A Lausanne study reported in HBR in March 2025 found that generative AI saved managers about 2 hours and 50 minutes a week on average, and that 83% of the AI users who saved time admitted to wasting at least a quarter of it. Saved time, in other words, does not automatically become leisure or extra output.

The two bodies of evidence are compatible. None of the time-saving studies tracks total working hours, so AI can shorten specific tasks while overall days stay the same or lengthen if freed time is absorbed by more work. That is a plausible reading supported by the Berkeley ethnography, ActivTrak's before-and-after cohort and the Lausanne findings, but it has not been tested causally at scale.

What the findings mean, and what to watch

The claim that AI adds hours rather than saving them is too simple for the evidence in hand. ActivTrak's data shows that AI users, especially heavy users and those working across browser tabs, have longer and more fragmented days. Independent studies show real, if modest, time savings on specific tasks. The open question is where the saved minutes go.

The most useful evidence to come would track the same workers through an AI rollout, with matched controls and logs of total hours. Until then, ActivTrak's own advice, that integration matters more than raw usage, is worth weighing alongside the fact that the company profits from measuring it. For workers and managers, the practical signal is to watch what rises when AI arrives: email, chat, scope and expectations, not just speed.

Frequently asked questions

Does using AI at work make the workday longer?

In ActivTrak's study of more than 9,000 employees, AI users had workday spans of 8.4 to 9.1 hours versus 7.6 hours for non-users. The study compares groups and does not show that AI causes the difference. Other research finds AI saves time on specific tasks, so the net effect on total hours is unresolved.

How much longer do AI users work, according to ActivTrak?

The release describes a gap of up to 90 minutes a day, which is the top of a range; the low end works out to roughly 50 minutes. Heavy users also logged about 18 more after-hours minutes a day than non-users. The sources do not say which usage tiers sit at each end of the range.

Who is most affected by AI-related workload changes?

ActivTrak's data points to heavy AI users, defined as more than 45 minutes and 5 or more sessions a day, and to users of browser-based tools, who switch applications most often. Software engineers and operations managers are named as heavy users, but no breakdown of workday length by role was found.

What does workday span mean?

It is the term ActivTrak's release uses for the length of the workday. The full definition could not be retrieved, but it is probably elapsed time rather than active time, which is unverified. A longer span does not necessarily mean more hours of focused work.

Can the ActivTrak study be trusted?

The figures are consistent across the release and coverage, but the study is a company report that has not been peer reviewed, and no methods appendix was found. ActivTrak sells monitoring software and has a commercial interest in how the results are framed. The numbers are best read as a signal, not a verdict.

Sources

1. ActivTrak Productivity Lab: AI Users Have Workdays Up to 90 Minutes Longer Than Non-Users , PR Newswire (via HealthTechnologyNet) (2026-10-29)
2. ActivTrak Productivity Lab: AI Users Have Workdays Up to 90 Minutes Longer Than Non-Users , PR Newswire (via Morningstar) (2026-10-29)
3. ActivTrak Productivity Lab: AI Users Have Workdays Up to 90 Minutes Longer Than Non-Users , PR Newswire (via ADVFN) (2026-10-29)
4. 'The goal can't simply be AI usage': New report finds people using AI are working up to 90 minutes extra every day , TechRadar
5. AI Users Are Working Up to 90 Minutes Longer Every Day, Study Finds , Gizmodo
6. ActivTrak Study: AI Users Work Longer Hours , HR Tech Edge
7. 'The goal can't simply be AI usage': Report finds AI users work up to 90 minutes longer each day , Mezha.net
8. Does AI actually save employees time? Data is contradictory , AI Commission
9. 2026 State of the Workplace , ActivTrak (2026-03-11)
10. ActivTrak Productivity Lab Report: Only 2% of Workforce Has Reached the Stage Where AI Transforms Workflows , ActivTrak
11. Does AI Save Time at Work? How to Measure It , ActivTrak
12. AI Doesn't Reduce Work—It Intensifies It , Harvard Business Review
13. AI Promised to Save Time. Researchers Find It's Doing the Opposite , Inc.
14. The Rapid Adoption of Generative AI (NBER Working Paper 32966) , National Bureau of Economic Research
15. The Impact of Generative AI on Work Productivity , Federal Reserve Bank of St. Louis
16. Shifting Work Patterns with Generative AI (NBER Working Paper 33795) , National Bureau of Economic Research
17. Generative AI and the Reallocation of Time: Productivity, Leisure, and Fulfilling Work , ResearchGate
18. How Is Your Team Spending the Time Saved by Gen AI? , Harvard Business Review