🔍 Read the full analysis: Workers Switch Apps 900 Times A Day. Does AI Cut That? What The Harvard Study Really Shows on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A Harvard Business School preprint analyzing 103 million application events from 1,017 workers found an average of 908 app switches per workday. Switching was lower in the ten app uses after an AI use, but workers switched more on days they used AI; the observational study cannot show that AI caused either pattern.
A Harvard Business School preprint tracking 1,017 office workers found they switched applications an average of 908 times per day, while its AI findings point in different directions: switching fell in the apps used immediately after AI, but rose on days when workers used AI. The study is observational, so it cannot establish that AI caused either pattern or that it improved work.
Researchers Sumer S. Vaid and Ashley V. Whillans analyzed 103 million application events recorded by workforce-analytics software at eight organizations in India from November 2025 through January 2026. The dataset included 62,693 person-days and 56.9 million switches between applications. A switch counted a change to a different app, not a move between two documents within the same app.
The average was 908.3 switches per day; the median employee recorded 814. The middle half of employees ranged from 547 to 1,130 switches per workday. Researchers reported a mean app-use duration of 27.8 seconds and a median of 3.3 seconds. These figures describe computer activity, not necessarily sustained attention or the quality of the work.
The AI results depend on the time window. In the ten application uses after an AI use, workers used fewer apps, switched less, stayed about half a second longer in each app and moved through apps more predictably. Yet workers made about 200 more switches on days with AI use, and days with more AI minutes were more fragmented, according to the paper. Those are associations, not evidence that AI itself increased switching.
App switching and AI · one moment
What happens around one AI query?
The study zoomed in on every AI use: the ten app uses before it, and the ten after. Up close, work calms down a little. Zoom out to the whole day, and the picture changes.
- 1 · A workday of switchingWorkers changed apps 908 times a day on average. One app held their focus for 27.8 seconds on average, 3.3 seconds at the median.
- 2 · AI comes in at a busy momentPeople used AI a median of 13 times a day, for a median of 7 seconds. Fragmentation peaked in the hour of AI use.
- 3 · The next ten app usesFewer different apps, fewer switches, about half a second longer per app, more predictable. Real, but small.
- 4 · Zoom out: the whole dayOn days with AI use, the same people made about 200 more switches. Whether AI calms work or busy moments bring AI: unknown.
Source: Sumer S. Vaid, Ashley V. Whillans (Harvard Business School), “Digital Fragmentation and Generative AI Use Across 103 Million Application Events”, arXiv:2607.06681v2, preprint under review. Blocks are schematic, not to scale.
Graphic: Powered by Thorsten Meyer AI
AI Use Does Not Equal Less Switching
The findings complicate the idea that adding an AI tool will automatically create a more focused workday. A short stretch after using AI looked less fragmented, while the day as a whole was more fragmented on days when AI was used. For employers weighing AI purchases, the paper offers no evidence that AI is a focus tool or that it improves productivity.
It also suggests that the conditions of a particular day may matter more than a worker’s usual habits or employer. The authors attributed 44.6% of variation in fragmentation to day-to-day changes within the same person, compared with 35.8% to differences between people and 19.6% to differences between organizations. The paper reports that fragmentation peaked on Tuesdays and reset after weekends and holidays. These patterns may help frame questions about scheduling and communication, but they do not establish why those patterns occur.
The limits matter for workers, too. Monitoring software records which application is active, not whether a person is thinking, collaborating, reading away from a screen or producing good work. The study measured app switching, and did not measure whether work improved.
Products Worth Considering

Apple MacBook Air 13-inch with M5 Chip, 16GB Memory, 512GB SSD, Midnight

DJI Mini 4K Camera Drone Combo with 2 Batteries

Sonos Era 300 Wireless Smart Speaker with Dolby Atmos, Black
As an affiliate, we earn on qualifying purchases.
What the Monitoring Data Captured
The paper, Digital Fragmentation and Generative AI Use Across 103 Million Application Events, was posted to arXiv on July 7, 2026, and revised on August 7. It labels itself a preprint under peer review and says findings may change. The authors are affiliated with Harvard Business School; the study is not a completed peer-reviewed finding.
The data came from Flowace, a commercial workforce-analytics service used by the participating organizations. Its software recorded changes in the active application, along with related computer-use details. Employers determined which apps counted as productive. The sample covered software, accounting, financial services, e-publishing, law and marketing organizations, with a median of 58 workdays recorded per employee.
The authors say the human-subjects protocol was approved by Harvard’s committee and that researchers received anonymized data under a data-use agreement. The paper says the research had no external funding, and that Flowace provided no funding or commercial support and had no role in study design, analysis or interpretation. One author disclosed equity in an unrelated company using AI for enterprise digital twins.
“The manuscript is currently under review, and the findings may change based on the peer review process.”
— Sumer S. Vaid and Ashley V. Whillans, in the preprint
Causation and Work Quality Remain Unknown
The study did not assign workers to use or avoid AI, so it cannot tell whether AI led to more daily switching, whether workers reached for AI on already fragmented days, or whether both patterns reflected other factors. The authors explicitly leave that direction of influence unresolved. The shorter-term pattern after AI use does not settle the question.
The sample is limited to employees at eight organizations in India using employer-installed computer monitoring. It does not cover other countries, phone activity or workers outside these participating workplaces. The app-event data also cannot show whether a switch was necessary, distracting or productive. The paper does not establish how switching relates to work quality, output or employee well-being.
Peer review is still pending, and the authors warn that results may change. The source material does not specify when review will be completed or whether a final paper will report revised estimates.
Peer Review and Better Causal Evidence
The immediate next milestone is peer review of the preprint. Readers should check for a later version or journal publication before treating the estimates as settled, particularly because the authors say the findings may change.
For now, the results support caution rather than a product verdict. Employers can distinguish app activity from productivity when evaluating workplace tools, and should not infer from this study that buying AI will reduce interruptions. Establishing whether AI changes switching would require research that can separate tool use from the reasons people choose to use it; this paper does not provide that causal test.
Key Questions
Did the study find that workers switch apps 900 times a day?
It reported a mean of 908.3 switches per day across the recorded work activity. The median employee made 814 switches per workday, so 908 was the average, not the typical worker’s exact count.
Did AI reduce app switching?
The results were mixed. Switching was lower in the ten app uses after an AI use, but workers made about 200 more switches on days they used AI. The study cannot show that AI caused either result.
Who took part in the research?
The data covered 1,017 employees at eight organizations in India, across sectors including software, accounting, financial services, publishing, law and marketing. Activity was recorded from November 2025 through January 2026.
Is the Harvard study peer-reviewed?
No. The paper was posted as a preprint under peer review. Its authors state that findings may change during the review process.
Does the paper show that AI improves productivity?
No. It measured application events and switching, not productivity, work quality or employee well-being. The study does not establish that AI improves focus or output.
Source: ThorstenMeyerAI.com
