🔍 Read the full analysis: The Harvard Study Asks What AI Can Do About App Switching 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 analyzed 103 million application events from 1,017 workers at eight companies in India. Workers switched apps less in the ten app uses after an AI use, but made about 200 more switches on days they used AI; the observational data cannot show that AI caused either pattern.
A Harvard Business School preprint analyzing 103 million application events from 1,017 office workers found that employees switched apps less in the ten app uses after using generative AI, yet made about 200 more switches on AI-use days overall. The contrasting results matter because the study does not establish that AI reduces workplace distraction: it tracks activity rather than testing whether AI caused changes in work patterns.
The researchers, Sumer S. Vaid and Ashley V. Whillans, counted 56.9 million app switches across 62,693 person-days, recorded from November 2025 through January 2026. Workers averaged 908.3 switches per day, while the median was 814. The middle half of employees recorded between 547 and 1,130 switches per workday. These are counts of changes from one application to another, not necessarily distinct tasks or interruptions.
In the ten application uses after an AI use, employees used fewer apps, switched less often, stayed roughly half a second longer in each app and moved between apps in a more predictable pattern, according to the paper. But the broader daily measure pointed the other way: AI-use days had about 200 additional app switches, and days with more minutes of AI use were more fragmented. These findings describe different time scales and are not necessarily contradictory.
The average time before a worker switched apps was 27.8 seconds; the median was 3.3 seconds. The paper compares these figures with earlier attention research, but says the studies use different methods, making them an imperfect basis for a direct historical comparison. The preprint describes its findings as observational and says it is under peer review; they may change.
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 Varies Across the Workday
The results caution against treating an AI tool as a proven way to improve focus. A short-term pattern after each AI interaction appears more orderly, while the whole day on which AI is used appears more fragmented. That distinction is relevant to employers evaluating productivity software: an immediate sequence of app activity may not represent the overall shape or quality of a worker’s day.
The paper also finds that day-to-day variation within the same person accounts for 44.6% of variation in fragmentation, compared with 35.8% attributed to differences between people and 19.6% to differences between organizations. The authors report that fragmentation peaked on Tuesdays and eased after weekends and holidays. These associations suggest that workload timing and communication patterns may deserve attention alongside individual habits, but the data do not show which workplace changes would improve concentration.
Nor does app switching measure whether work was good, useful or completed faster. The study records application activity, not output quality, employee well-being or the reasons someone changed apps. Its findings can inform questions about workplace technology, but they do not establish that high switching is necessarily harmful or that AI improves performance.
Products Worth Considering

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

Amazon Kindle Scribe (16GB) with Premium Pen, Tungsten

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
How the App Records Were Collected
The paper, Digital Fragmentation and Generative AI Use Across 103 Million Application Events, is by Vaid of Harvard Business School’s Artificial Intelligence Institute and Whillans of its Negotiation, Organizations and Markets Unit. It was first posted on arXiv on July 7, 2026, and revised on August 7. The manuscript identifies itself as a preprint under review and warns that findings could change during peer review.
The sample covers eight organizations in India across software, accounting, financial services, e-publishing, law and marketing. Data came from Flowace workforce-analytics software, which the paper says recorded the active application and associated timing when an employee changed apps. The employers determined which applications counted as productive. Researchers received anonymized records under a data-use agreement; Harvard’s human-subjects committee approved the protocol.
The authors state that the study had no external funding and that Flowace provided no commercial support and had no role in the design, analysis or interpretation. The paper also discloses that Vaid holds equity in an unrelated company that uses AI to create enterprise digital twins. The data included several AI tools, including Anthropic’s Claude, which was the smallest of the five AI tools in the sample.
“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 Coverage Remain Open
The study cannot establish whether AI use led to more fragmented days, whether workers turned to AI on days that were already more fragmented, or whether both patterns occurred. It was not an experiment, and the authors describe multiple possible explanations. The short sequence after AI use and the day-level association should not be read as proof that AI either improves focus or causes distraction.
The sample is limited to employees at eight organizations in India, observed through employer-used monitoring software from November 2025 to January 2026. It does not cover phone use, other countries or whether the work produced better results. Because employers set the definition of productive apps, the records also do not independently determine what counted as productive activity. The paper remains under review, and its estimates or interpretation may change.
Peer Review and Broader Evidence
The next formal step is peer review of the preprint; the authors have not specified in the supplied material when that process will conclude. Until then, the reported patterns should be treated as preliminary findings from this particular sample and measurement system.
Further research would need to test whether changing AI access or workplace routines causes app-switching patterns to change, and whether those changes affect output, concentration or employee experience. For organizations considering AI purchases, the study alone does not support marketing or policy claims that a tool will improve focus. It does point to a practical distinction to track: how activity changes immediately after AI use versus how fragmented the full workday becomes.
Key Questions
Did the study find that AI reduces app switching?
Only in a narrow sequence. In the ten app uses after an AI use, workers switched less. Across whole days when they used AI, they made about 200 more switches on average. The study does not establish that AI caused either pattern.
How many workers and companies were included?
The analysis covered 1,017 employees at eight organizations in India, across sectors including software, accounting, finance, law and marketing.
Was the paper peer-reviewed?
No. The paper was posted as a preprint under peer review. Its authors say findings may change through that process.
What does 908 app switches a day mean?
It is the reported average number of changes from one application to another per day. The study’s median was 814 switches per workday; a switch count does not by itself show how many tasks a worker completed or how often work was interrupted.
Does the study show that AI improves productivity?
No. The researchers measured application activity, not whether workers completed more work, produced higher-quality results or felt more focused. The study therefore does not establish a productivity benefit.
Source: ThorstenMeyerAI.com
