AI Users May Work 90 Minutes More. Boundaries Beat Speed
Key Takeaways
- Track what AI removes from your day, then decide what will not be refilled.
- Prefer AI workflows that reduce app switching and review debt, not just first draft time.
- Ask managers to define outcomes and workload boundaries before AI speed becomes extra work.
Why it matters
- ProductProduct leaders should design AI features that reduce switching and review debt, not merely increase tool usage.
- InvestorsInvestors should treat AI productivity claims cautiously when the measurement excludes longer hours or hidden effort.
New workplace data suggests AI fluency should include workload limits, delegation rules, and proof of outcomes.
The promise was clean: give workers AI, remove drudgery, shorten the day. The messier version now showing up in workplace data is that faster task completion can quietly become more tasks, more switching, and a longer workday. For learners trying to show AI fluency on a resume, that changes the assignment. The skill is not merely using the tool, it is designing a workflow that does not turn speed into unpaid sprawl.
The 90 minute finding is a warning label, not a verdict
Infobae reported that employees using AI worked between 8.42 and 9.08 hours per day, compared with 7.58 hours for employees who did not use AI. That is the headline number, a gap of up to 90 minutes per day, but the source matters as much as the gap. Infobae said the ActivTrak Productivity Lab reviewed data from 9,140 office workers at 467 companies between April 1 and September 25, 2026, using actual application records rather than a survey.
That makes the finding harder to wave away as vibes, but it still is not proof that AI caused longer hours. Infobae noted that ActivTrak used cautious language and did not demonstrate causation. In career terms, treat this as a signal about how AI gets absorbed into office systems: if a team saves time but has no agreement about what happens next, the calendar often finds a use for the surplus.
The ActivTrak Productivity Lab release sharpened that point by saying the difference is between adopting AI tools and integrating them well. Gabriela Mauch of ActivTrak put it plainly: “the goal can't simply be AI usage.” That is useful advice for job seekers, too. Listing tools is weaker than showing a before and after workflow, with the boundary condition included.
App switching is the tax nobody puts on the job post
ActivTrak reported that heavy AI users switched between applications more frequently and had shorter periods of uninterrupted focus. Infobae added a clearer cut of that pattern: heavy users averaged 37.8 minutes of uninterrupted focus per session, versus 48.2 minutes for light users. That 10.4 minute difference amounted to 22% less uninterrupted time, according to Infobae.
This is where many AI productivity claims get sloppy. A browser chatbot can make a single task feel faster while making the total workflow more fragmented. Infobae reported that employees using AI embedded in their usual programs switched applications 0.6 times per hour, while those using browser based assistants switched between 1.7 and 4.2 times per hour.
For professionals building AI skills, that distinction is practical. A good portfolio project should not only say, “I used AI to draft, summarize, or analyze.” It should show where the AI sits in the workflow, what handoffs it removes, which checks remain human, and how much review debt it creates. Hiring managers may ask for tool familiarity, but the better screen is whether you can reduce friction without hiding it somewhere else.
Managers may refill the saved time
CIO reported that ActivTrak’s latest findings conflict with earlier ActivTrak research that found AI users worked less, which is exactly why tidy productivity slogans deserve suspicion. CIO said the discrepancy may reflect different collection periods and sample sizes, but the bigger lesson is that measuring AI gains remains difficult. If your employer says AI saved time, the follow up question is simple: saved for whom, and what happened to it?
CIO also reported ActivTrak’s warning that faster task completion can lead managers to assume employees have capacity for more responsibilities. Without clear workload boundaries, time savings can be redirected into more work. That is not a reason to avoid AI. It is a reason to negotiate the operating rules before the pilot becomes the norm.
Those rules do not need to be theatrical. Define which tasks AI may draft, which tasks require human review, when outputs are batched, and what work will be removed if AI adds a new review queue. If you are early in your career, that may mean documenting your process so your speed is visible without volunteering for endless overflow. If you are midcareer, it may mean making the boundary explicit with your manager before your expertise becomes the cleanup layer for everyone else’s prompts.
Build AI fluency around outcomes, not hours
ActivTrak defined heavy AI users as employees who used AI for more than 45 minutes and at least five sessions per day, while light users used AI for less than 10 minutes per day. That definition is a useful reminder that intensity is not the same as maturity. More sessions can mean more leverage, or it can mean more hovering, checking, copying, pasting, and repairing.
The strongest AI upskilling plan now looks less like collecting certificates and more like proving workflow judgment. Pick a recurring task, write down the old process, add AI only where it reduces a real bottleneck, then measure the result in output quality, review time, handoffs, and after hours spillover. If the tool saves ten minutes but creates a new habit of evening checks, the workflow is not done.
Watch this space for better evidence on causation, especially studies that separate job type, tool design, and management expectations. For now, the practical lesson is already clear enough. AI literacy that belongs on a resume is not just speed with a shiny interface, it is the ability to set delegation rules, protect focus, and prove that productivity gains do not simply expand the workday.
Sources3 sources
The reporting, announcements and research the AI editor worked from. Links open the original publisher.
