Contents
ActivTrak’s analysis of 9,140 employees at 467 companies found that AI users worked up to 90 minutes longer per day than non-users. The findings show why companies should measure both completed work and working hours before claiming that AI saves time.
Key Takeaways#
- AI users worked 8.4–9.1 hours per day, compared with 7.6 hours for non-users.
- Frequent AI users had shorter uninterrupted work periods than infrequent users: 37.8 minutes versus 48.2 minutes.
- Users of embedded AI switched apps less often than users of browser-based AI tools.
- The study shows an association; it does not prove that AI caused longer working hours or reduced concentration.
What the Study Measured#
ActivTrak’s Productivity Lab analyzed workplace activity data collected from April 1 to September 25, 2026. Employees were classified as AI non-users or grouped by usage level.
Infrequent users spent less than 10 minutes per day using AI. Regular users spent 10–45 minutes. Frequent users spent at least 45 minutes and used AI in at least five sessions per day.
These categories describe recorded usage, not work quality. Employees who spend longer using AI may be handling harder tasks, reviewing poor-quality output, or completing more work. Usage time alone cannot distinguish among these possibilities.
Longer Hours, Shorter Focus#
The AI user group with the longest average working day worked 9.1 hours, compared with 7.6 hours for non-users—a difference of 1.5 hours. This is the basis for the “up to 90 minutes” finding. It does not mean that every AI user worked an extra 90 minutes.
The share of work performed outside scheduled hours was similar: 32.3% for AI users and 31.9% for non-users. Because AI users worked longer overall, they reportedly also worked 18 minutes longer outside scheduled hours.
Frequent AI users also averaged 37.8 minutes of uninterrupted work, about 22% less than the 48.2 minutes recorded for infrequent users. Meanwhile, meetings accounted for 3% of AI users’ working hours, compared with 5% for non-users. In this data, therefore, less meeting time did not translate into shorter working hours.
Where AI Is Used Matters#
Employees using AI embedded in their primary work apps switched apps an average of 0.6 times per hour. Browser-based AI users reportedly switched 1.7–4.2 times per hour.
This difference suggests that workflow design is worth examining. Moving between a work app and a separate AI tool may require copying information, restating context, and transferring responses. These steps may consume some of the time saved on the task itself.
Start by choosing one recurring task and counting the required information transfers. If drafting a reply requires moving information between multiple windows, test whether an approved integration can reduce those transfers. Check accuracy and completion time. Fewer app switches alone do not prove better results.
What These Findings Cannot Prove#
This is an analysis of observational data, not a randomized experiment. It cannot determine whether AI lengthened working hours, employees with heavier workloads were more likely to use AI, or another factor influenced both.
Differences between roles also complicate comparisons. The report states that software engineers tended to use embedded AI, while business operations managers tended to use browser-based tools. Because these roles require different types and amounts of coordination, the entire difference in app switching cannot be attributed to where AI is placed.
Activity logs can address some weaknesses of self-reported estimates, but they do not fully measure productivity. The findings presented do not show whether longer working hours led to more approved deliverables, higher quality, or less rework. These outcomes need to be checked before drawing conclusions about AI’s value.
Measure Outcomes#
A useful AI evaluation needs to answer a practical question: Can this workflow produce results that meet acceptance criteria while reducing total effort?
Choose one recurring task and run a narrowly focused comparison. For comparable work, record these four measures.
- Time from starting work to approval of the result, including AI interactions and review.
- Error corrections and revisions required before approval.
- Total daily working hours, including work outside scheduled hours.
- Number of comparable tasks completed and approved.
Compare results within the same role wherever possible. If tasks are completed faster but working hours increase, investigate whether employees are being assigned more work or spending extra time reviewing output. Saving time on individual tasks and shortening the working day are separate outcomes.
Next Step#
Choose one recurring task that uses AI and compare it with the existing process for two weeks. Record completion time, corrections, working hours, and approved outcomes before deciding whether to expand its use.
Source: https://gigazine.net/news/20261113-ai-users-have-long-worktime-than-non-ai-users/
Contents
