🔍 Read the full analysis: How To Understand The Harvard Study On AI And App Switching on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
A Harvard Business School preprint analysed 103 million application events from 1,017 office workers at eight organisations in India. It found fewer app switches in the ten app uses after an AI use, but about 200 more switches on days when workers used AI; the observational data cannot show that AI caused either pattern.
A Harvard Business School preprint tracking 1,017 office workers found that they switched applications less often in the ten app uses after using AI, but made about 200 more switches on days they used AI. The mixed results complicate a headline that spread on November 18 saying AI reduced app switching, and the study does not establish that AI caused either pattern.
Researchers Sumer S. Vaid and Ashley V. Whillans analysed 103 million application events recorded between November 2025 and January 2026. The sample covered employees at eight organisations in India, working in software, accounting, financial services, e-publishing, law and marketing. The paper, “Digital Fragmentation and Generative AI Use Across 103 Million Application Events,” was posted on arXiv and remains under peer review.
Across 62,693 person-days, workers switched applications an average of 908.3 times per day; the median was 814. The measure counts a switch when someone moves to a different application, not between two documents in the same app. An application remained active for a mean of 27.8 seconds before a switch, while the median duration was 3.3 seconds. These measures describe recorded app activity, not the quality or output of the work.
The AI findings 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 per app and followed more predictable sequences. Yet on days with AI use, they made roughly 200 additional switches, and days with more AI minutes were more fragmented. The authors say the data cannot determine whether AI changed how the day unfolded, workers used AI on days that were already different, or both.
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
Why AI Days Look More Fragmented
The findings matter because claims that workplace AI improves focus can influence software purchases and work policies. This study does not show that buying an AI tool will reduce interruptions: the short sequence after an AI use and the full day show different associations, and neither is a causal test.
It also suggests that app switching may depend heavily on the workday itself. The authors attribute 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 organisations. That pattern supports studying daily workload and communication demands, but it does not prove which changes would improve focus or results.
For workers and managers, the practical limit is clear: the paper measures application activity, not productivity, concentration or wellbeing. Its authors and the source account do not establish that fewer switches are always better, or that AI makes work more effective. The figures should be read as a description of monitored computer use in this sample, rather than a general measure of how all employees work.
Popular right now

KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Dried Rose (KSM150PS)

Breville Barista Express Espresso Machine, Black Truffle

Ninja Foodi Smart 6-in-1 Indoor Grill and Air Fryer, 4 qt, FG551
As an affiliate, we earn on qualifying purchases.
How Researchers Measured App Use
The event data came from Flowace, a commercial workforce-analytics service already used by the participating organisations. Its software recorded changes in the active application, including app names and timestamps; the paper says the researchers received anonymised data under a data-use agreement. Employers, rather than the researchers, decided which applications counted as “productive.” Harvard’s human-subjects committee approved the protocol, according to the paper.
The study includes workers at two software firms, two accounting firms and one organisation each in financial services, e-publishing, law and marketing. Each person contributed a median of 58 workdays. The paper states it received no external research funding and that Flowace did not fund or direct the analysis. One author disclosed equity in a private company unrelated to Flowace that uses AI to create enterprise digital twins.
The headline circulating on November 18 captured the high daily switching average but flattened the AI results into a single conclusion. The paper itself describes the manuscript as a preprint under review and warns that its findings may change through peer review. Its comparisons with earlier studies of app attention also involve different methods, so they are not a precise like-for-like measure of change over time.
“The manuscript is currently under review, and the findings may change based on the peer review process.”
— Sumer S. Vaid and Ashley V. Whillans, paper authors
What the Data Cannot Establish
The research is observational, not experimental. It cannot establish whether AI leads to more switching across a day, whether workers turn to AI when their workday is already fragmented or consolidated, or whether other factors explain the association. The authors explicitly leave open the possibility that AI reorganises the day, that workers use it in response to the day’s demands, or both.
The evidence also has limits in scope. It comes from eight organisations in India, covers computer applications rather than phone use, and does not measure whether work became better or worse. It remains unclear whether the patterns would hold in other countries, industries or monitoring settings, and how much app-switching frequency reflects meaningful task changes rather than interruptions.
The paper remains under peer review, and the reported findings could change. The source material also notes that the person behind the social-media account that spread the headline could not be confirmed. Neither that account’s identity nor the headline’s reach changes what the study itself measured.
Peer Review and Broader Testing
The immediate milestone is peer review of the preprint. The findings may be revised, and readers should distinguish any later version from the current draft. More research would be needed to test whether AI use causes changes in switching, ideally by comparing work patterns under conditions that can separate AI’s effects from differences in workload and the type of day.
For now, organisations can treat the results as a reason to examine daily work patterns, not as evidence that an AI purchase will improve focus. The authors’ analysis points to substantial variation from one day to another and reports that fragmentation peaked on Tuesdays and reset after weekends and holidays. Whether protecting concentrated mornings or adjusting midweek communication would improve outcomes is not tested by this paper.
The next questions are whether similar patterns appear in other countries and types of work, whether they persist in a peer-reviewed version, and whether reduced switching after AI use corresponds to better work. Until those questions are answered, the defensible conclusion is limited: this sample showed less switching immediately after AI use and more switching on AI-use days, without establishing why.
Key Questions
Did the Harvard study find that AI reduces app switching?
Only in a specific short window. In the ten application uses after an AI use, workers switched less. Across whole days, however, AI-use days had about 200 more switches. The study does not show that AI caused either result.
What does the figure of 908 app switches per day mean?
It is the mean number of switches between different applications per workday in the monitored sample. The median was 814. It is a count of recorded computer activity, not a measure of productivity or interruptions on its own.
Who took part in the study?
The analysis covered 1,017 employees at eight organisations in India, across software, accounting, financial services, e-publishing, law and marketing. Their computer application events were recorded from November 2025 through January 2026.
Has the study been peer reviewed?
Not yet, according to the source material. The paper was posted as a preprint on arXiv and says it is under review, with findings that may change during the peer-review process.
Does the study show that AI improves productivity?
No. It measures application events and switching patterns, not work quality, output or productivity. Its observational design also cannot establish whether AI caused the patterns it reports.
Source: ThorstenMeyerAI.com
