AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: AI And The Workplace: Making Sense Of 900 Daily App Switches on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

A Harvard Business School preprint analyzed 103 million application events from 1,017 workers at eight Indian organizations. Workers averaged 908 app switches a workday; they switched less in the ten app uses after an AI use, but made about 200 more switches on days they used AI. The observational study cannot show that AI caused either pattern.

A Harvard Business School preprint analyzing 103 million application events found that 1,017 office workers switched apps an average of 908 times per workday, with a mixed pattern around generative AI: switching fell in the ten app uses after an AI use, but workers made about 200 more switches on AI-use days. The study is observational, so it cannot establish that AI caused either change.

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. It draws on records collected from November 2025 to January 2026 by Flowace, a commercial workforce-analytics service used by the participating employers. The authors say the data comprised 56.9 million switches across 62,693 person-days.

Across those days, the mean was 908.3 switches per day; the median worker recorded 814. The middle half of employees ranged from 547 to 1,130 daily switches. The paper defines a switch as a move to a different application, not a change between documents within one app. It reports an average app-use period of 27.8 seconds and a median of 3.3 seconds before a worker moved to another application.

The AI results depend on the time scale being measured. 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. At the day level, however, workers made about 200 more switches on days when they used AI, and days with more AI-use minutes were more fragmented. These are associations in logged activity, not evidence that AI itself produced the increase.

At a glance
reportWhen: Preprint posted July 7, 2026, and revis…
The developmentA Harvard Business School preprint reports that workers switched apps less immediately after AI use but more on days when they used AI.

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.

10 app uses before AI median 7 seconds 10 app uses after Whole AI day: about 200 more switches
  1. 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. 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. 3 · The next ten app usesFewer different apps, fewer switches, about half a second longer per app, more predictable. Real, but small.
  4. 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 a simple claim that AI tools give workers more focus time. The short sequence after an AI interaction looks less fragmented, while the broader daily measure points the other way. Both patterns can be true: a person may move through fewer apps immediately after asking an AI tool for help, yet have a more interrupted day overall.

For employers weighing AI purchases or workplace policies, the distinction matters. This paper does not show that AI improves concentration, productivity or work quality, and it does not establish that AI causes fragmentation. The authors suggest that AI may reorganize how a day is structured, that people may use AI on days that are already becoming more consolidated, or that both processes may be involved. The data cannot distinguish among those possibilities.

The study also points to the importance of daily working conditions. 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 workers and 19.6% to differences between organizations. The researchers report that fragmentation peaked on Tuesdays and eased after weekends and holidays. Those observations support looking at communication and scheduling patterns, but they do not prove that changing a schedule will improve outcomes.

Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

As an affiliate, we earn on qualifying purchases.

How the Workplace Data Was Collected

The sample covered 1,017 employees at eight organizations in India, across software, accounting, financial services, e-publishing, law and marketing. Each person contributed a median of 58 workdays of records. The monitoring software recorded the active application and related details when a worker changed apps. Employers, rather than the researchers, determined which applications counted as productive.

The authors describe the records as anonymized before they were provided to the research team under a data-use agreement. They say Harvard’s human-subjects committee approved the protocol. The paper states that it received no external funding and that Flowace had no role in the study’s design, analysis or interpretation. It also discloses that Vaid holds equity in a separate private company that uses AI to create enterprise digital twins.

The manuscript was posted on arXiv on July 7, 2026, and revised on August 7. It identifies itself as a preprint under peer review and warns that findings may change through that process. The authors are Harvard researchers, but the paper’s status is not the same as a peer-reviewed publication.

“The manuscript is currently under review, and the findings may change based on the peer review process.”

— Vaid and Whillans, in the paper’s preprint

What the App Logs Cannot Show

The study does not establish whether AI use leads to more switching, whether workers turn to AI on days that are already more fragmented, or whether other factors account for both patterns. The authors explicitly leave open the possibility that AI reorganizes the day, that workers select AI in response to the day’s existing structure, or that both are happening. No experiment was conducted.

The records measure application activity, not the quality or value of the work. They do not show whether employees completed tasks faster, made fewer errors, felt more focused or experienced more stress. The paper’s app-switching measures also cannot, by themselves, establish whether a move between applications interrupted meaningful concentration.

The sample is limited to eight organizations in India and to activity recorded on work computers using Flowace. The source material does not establish how well the results apply to other countries, workplaces without this monitoring software or work on phones. The paper also notes that earlier studies of time spent in applications used different methods, so its comparisons with past attention research should be treated as broad signals, not precise measurements of change over time.

Peer Review and Better Workplace Measures

The immediate next step is peer review, which may lead to changes in the manuscript or its conclusions. Until then, employers and workers should treat the results as evidence about app activity in this sample, not as a settled measure of AI’s effect on productivity or attention.

Further research would need to separate the effect of AI from the reasons workers choose to use it, ideally by measuring work outcomes alongside app activity and testing whether changes in AI access or work practices alter those outcomes. For teams considering practical steps now, the data can prompt questions about meeting load, communication and midweek interruptions. It does not, on its own, support buying an AI tool as a focus aid or claiming that AI reduces workplace fragmentation.

Key Questions

Did the study find that AI reduces app switching?

Not across the whole day. Workers switched less in the ten app uses after an AI use, but made about 200 more switches on days they used AI. The study cannot establish cause and effect.

What does the 908 figure mean?

It is the mean number of switches to a different application per worker per day in the study. The median was 814, and switching between two documents in the same app did not count as a switch.

Was this a peer-reviewed study?

No. The paper was posted as a preprint under peer review. Its authors state that findings could change during that process.

Can the results show that AI makes people less productive?

No. The study measured application activity, not productivity or work quality, and its observational data cannot show that AI caused the switching patterns.

Who was included in the research?

The records covered 1,017 employees at eight organizations in India in software, accounting, financial services, e-publishing, law and marketing. The data came from workplace computers monitored using Flowace.

Source: ThorstenMeyerAI.com

You May Also Like

Is AI causing a repeat of Front end’s Lost Decade?

Analysis of how AI’s impact on programming echoes past shifts in frontend development, raising questions about skill erosion and industry changes.

King Luther Capital Management Corp Has $129.43 Million Stake in Meta Platforms, Inc. $META

King Luther Capital Management boosted its Meta holdings by 5.5% in Q4, now owning $129.43 million worth of shares, according to SEC filings.

Chinese scientists identify degradation pathways in low-silver heterojunction solar cells

Chinese researchers identify how interdiffusion causes long-term degradation in low-silver heterojunction solar cell electrodes, impacting reliability.

Google DeepMind Unionization Talks Are Off to a Rocky Start

Initial unionization negotiations at Google DeepMind in London have encountered significant setbacks, raising questions about employee representation efforts.