🔍 Read the full analysis: The Harvard Study On Whether AI Can Help With App Switching on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A Harvard Business School preprint analysing 103 million application events found that workers switched apps less in the ten uses after an AI tool, but made about 200 more switches on days they used AI. The observational data cannot show whether AI caused either pattern, and the study did not measure work quality.
A Harvard Business School preprint analysing 103 million application events from 1,017 office workers found that app switching was lower in the ten app uses immediately after an AI tool was used, but higher on days when workers used AI. The study matters because it complicates claims that AI reliably reduces workplace distraction: its data show associations, not that AI caused either pattern.
The paper, Digital Fragmentation and Generative AI Use Across 103 Million Application Events, was written by Sumer S. Vaid and Ashley V. Whillans of Harvard Business School. The researchers analysed records collected from November 2025 through January 2026 at eight organisations in India, covering software, accounting, financial services, e-publishing, law and marketing. The paper is an arXiv preprint and says it is still under peer review, so its findings may change.
Across 62,693 person-days, workers made a mean of 908.3 changes between applications per day; the median was 814. The researchers also reported an average app-use period of 27.8 seconds, with a median of 3.3 seconds. These measures describe recorded app activity, not necessarily uninterrupted attention: a switch in the active application does not establish what a worker was thinking or whether the work was productive.
The AI findings point in different directions depending on the comparison. In the ten application uses after an AI use, workers used fewer apps, switched less, stayed slightly longer in each app and followed a more predictable sequence. Yet on days when AI appeared in the logs, workers made about 200 more switches per day; days with more AI-use minutes were also more fragmented. The source reports no causal estimate, and the paper does not establish whether AI use led to more switching or whether workers used AI on already busy days.
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 and Fragmented Workdays
The results are relevant to employers weighing AI tools as a way to improve concentration. They do not support a simple conclusion that AI reduces interruptions across a workday. The short sequence following an AI interaction looked less fragmented, while whole days involving AI use showed more switching. Both observations can be true, and the study cannot determine why.
The researchers also found that 44.6% of variation in fragmentation came from day-to-day changes within the same person, compared with 35.8% from differences between workers and 19.6% from differences between organisations. That points to the workday itself as an important part of the pattern, rather than treating app switching only as a stable personal trait. The data also showed fragmentation peaking on Tuesdays and resetting after weekends and holidays.
For workers and managers, the practical limit is clear: this paper does not show that buying or deploying an AI product will improve focus, output or wellbeing. It measures application activity, not the quality of completed work, and the researchers did not run an intervention assigning some employees AI tools and others none. Any workplace decision based on these results should distinguish a recorded association from evidence of benefit.
Products Worth Considering

Samsung 32 Odyssey G5 QHD Gaming Monitor

LG UltraGear 34G600A-B 34-Inch Curved WQHD Gaming Monitor

Jackery HomePower 3600 Plus Portable Power Station
As an affiliate, we earn on qualifying purchases.
How the Activity Was Recorded
The study used data 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 timing; the paper says records could also include window titles and, where applicable, web addresses. The participating employers decided which applications counted as productive. Researchers received anonymised data under a data-use agreement, and the paper says the research protocol received human-subjects committee approval.
The sample included two software firms, two accounting firms and one organisation each in financial services, e-publishing, law and marketing. Each participant contributed a median of 58 workdays. The authors say the study received no external funding and that Flowace did not fund or direct the research. One author disclosed equity in a private company unrelated to Flowace. The paper names several AI tools in its records, including Claude, but these logs do not by themselves establish how workers used each tool or what effect it had on their tasks.
The headline that circulated on November 18, 2026, highlighted the average of roughly 900 daily switches and suggested switching dropped after AI. That captures the immediate post-use comparison but leaves out the separate finding that AI-use days had more switches overall. The paper itself was posted to arXiv on July 7, 2026, with a revision dated August 7; it identifies itself as a preprint under review, not a peer-reviewed finding.
“The manuscript is currently under review, and the findings may change based on the peer review process.”
— Vaid and Whillans, in the preprint
What the App Logs Cannot Show
The study is observational, so it cannot establish whether AI caused more switching across whole days or less switching immediately after an AI interaction. The authors say AI may reorganise the workday, workers may turn to AI when a day is already consolidating, or both. The data do not resolve those explanations.
The sample covers a limited set of organisations in India and does not include phone activity. It also does not show whether frequent switching harmed performance, whether workers completed tasks faster or better with AI, or how they experienced the monitoring. The reported averages describe this sample and measurement method; they should not be assumed to represent all workers or workplaces.
Peer review remains pending, and the preprint warns that findings may change. It is also not clear from the reported results whether the observed patterns differ substantially by AI tool, job role or particular task. The study records software activity but does not provide a controlled test of AI use or a direct measure of work quality.
Peer Review and Better Tests
The immediate next milestone is peer review of the preprint. Review may lead to changes in the analysis or interpretation; no publication decision or timeline is established in the supplied material.
To test whether AI changes app switching, future research would need designs that can separate cause from selection—for example, comparing similar workdays with and without assigned AI access while also measuring task completion and quality. Until such evidence is available, this study supports a cautious conclusion: app activity was less fragmented in a short period after AI use but more fragmented on AI-use days overall, with the reason still unknown.
Key Questions
What did the Harvard study find about app switching?
In records from 1,017 workers, the mean was 908.3 app changes per day. Switching was lower in the ten app uses after an AI use, but workers made about 200 more switches on AI-use days. These are different comparisons, and neither establishes that AI caused the pattern.
Does the study prove AI reduces workplace distraction?
No. The research is observational and measures application activity, not distraction or work quality directly. It cannot show that AI caused the lower switching immediately after use, or explain the higher switching seen across AI-use days.
Who took part in the study?
The sample comprised 1,017 office workers at eight organisations in India, across software, accounting, financial services, e-publishing, law and marketing. The data covered November 2025 to January 2026.
Is the study peer-reviewed?
No. The paper is an arXiv preprint under peer review. Its authors state that the findings may change during that process.
Did the researchers measure whether AI improved work?
The reported study examined application events and switching patterns. It did not establish whether AI improved productivity, accuracy, task completion or worker wellbeing.
Source: ThorstenMeyerAI.com
