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

🔍 Read the full analysis: App Switching, AI, And Work: Five Questions The Harvard Study Raises on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

A Harvard Business School preprint analyzed 103 million application events from 1,017 employees at eight Indian organizations. Workers averaged 908.3 app switches per day; switching was lower in the ten app uses after an AI use, but AI-use days had about 200 more switches overall. The observational data cannot show whether AI caused either pattern or whether work quality improved.

A Harvard Business School preprint analyzing 103 million application events from 1,017 employees found that workers switched apps an average of 908.3 times per day, while the relationship between generative AI and switching varied by time scale. App use was less fragmented in the ten app uses after an AI interaction, but employees had about 200 more switches on AI-use days than on days without recorded AI use. The study is observational and does not establish that AI caused either pattern.

The paper, titled Digital Fragmentation and Generative AI Use Across 103 Million Application Events, was written by Sumer S. Vaid of the Harvard Business School Artificial Intelligence Institute and Ashley V. Whillans of its Negotiation, Organizations and Markets Unit. It examined records collected from November 2025 to January 2026 at eight organizations in India. The authors describe the manuscript as a preprint under peer review and warn its findings may change.

The workers made 56.9 million app switches across 62,693 person-days. The reported mean was 908.3 switches per day; the median employee made 814, and the middle half of employees ranged from 547 to 1,130. The average period spent in an application before switching was 27.8 seconds, while the median was 3.3 seconds. These measures count changes between applications, not changes between documents within the same app.

The AI findings differ depending on the measure. In the ten application uses after an AI use, workers used fewer apps, switched less and stayed roughly half a second longer in each app, with activity following a more predictable sequence. Across whole days, however, days with AI use were associated with about 200 additional switches, and days with more AI minutes were more fragmented. The authors say these patterns do not establish which came first or whether AI changed the workday.

At a glance
reportWhen: Preprint posted July 7, 2026, revised A…
The developmentA Harvard Business School preprint based on workplace monitoring data reports high app switching and mixed associations between generative AI use and 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.

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 Shows Two Switching Patterns

The findings matter because claims that AI reduces workplace distraction can depend on which window is measured. The sequence immediately after AI use appeared less fragmented, while the full day of AI use was associated with more switching. Treating only one of those observations as proof that AI improves focus would overstate what the study shows.

The paper also points to variation within the same worker: 44.6% of fragmentation variation was attributed to day-to-day changes for an individual, compared with 35.8% to differences between people and 19.6% to differences between organizations. That suggests workload and daily conditions may be relevant alongside individual habits. The study did not measure whether switching harmed productivity, well-being or the quality of work, so those outcomes remain unanswered.

For employers, the evidence is not a basis for buying an AI product as a focus intervention. The authors’ results describe associations in a particular monitored workforce, not a controlled test of an AI tool. The data may help frame questions about how workdays are organized, but it does not show that reducing app switching by itself improves performance.

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

De’Longhi All-in-One Espresso and Coffee Maker with 10-Cup Pot and Milk Frother

De’Longhi All-in-One Espresso and Coffee Maker with 10-Cup Pot and Milk Frother

Samsung 65-Inch 4K OLED Smart TV with Dolby Atmos, 120Hz Refresh Rate, and Amber Protection Bundle

Samsung 65-Inch 4K OLED Smart TV with Dolby Atmos, 120Hz Refresh Rate, and Amber Protection Bundle

As an affiliate, we earn on qualifying purchases.

What the Workplace Data Covers

The sample included employees at two software companies, two accounting firms, and organizations in financial services, e-publishing, law and marketing. The source was Flowace, a commercial workforce-analytics service used by the participating employers. According to the paper, its software recorded when an employee changed the active application, along with app names, times and other window information. Employers, rather than the researchers, decided which apps counted as productive.

The authors say the data were anonymized before reaching them under a data-use agreement, and that the research protocol received human-subjects committee approval. The paper states it 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 a private company unrelated to Flowace that uses AI for enterprise digital twins.

The researchers also compare their app-use durations with earlier attention research, including estimates of about 2.5 minutes two decades ago and 47 seconds in the last decade. Because the studies used different methods, those figures should be read as a broad comparison rather than a directly measured change over time. The study covers computer applications at Indian workplaces, not phone use or employees in other countries.

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

— The paper’s authors

What the Study Cannot Establish

The study did not randomly assign workers to use AI, so it cannot show whether AI led to more switching, whether workers turned to AI on already fragmented days, or whether both patterns occurred. The authors explicitly leave that direction unresolved. A lower switching rate immediately after AI use also does not prove that people concentrated better or completed work more effectively.

Other limits include the sample’s scope and the source of the records: eight organizations in India using one workforce-monitoring service. The paper does not assess phone activity, represent every kind of workplace, or report whether output quality improved. Its measures also depend partly on employer choices about which applications were classified as productive. Peer review is still underway, and findings could change.

Peer Review and Broader Testing

The next formal milestone is peer review of the preprint; the paper does not state when that process will conclude. Readers should look for any revised analysis, methods or qualifications when a reviewed version becomes available. Further studies across different countries and types of work could test whether these patterns hold beyond the eight participating organizations.

To establish whether AI causes changes in switching or improves work, future research would need designs that can distinguish AI’s effect from workload, task type and other daily conditions, and measures of work quality as well as app activity. For now, the paper supports a limited conclusion: app switching was frequent in this sample, and its relationship with AI varied between short sequences and whole workdays.

Key Questions

What did the Harvard study find about app switching?

In the monitored sample, workers switched applications an average of 908.3 times per day. The median employee made 814 switches per workday.

Did AI reduce app switching?

Not consistently. Switching was lower in the ten app uses after an AI use, but days with AI use were associated with about 200 more switches overall. The observational study cannot establish that AI caused either pattern.

Who and what did the study cover?

It analyzed computer application records from 1,017 employees at eight organizations in India, collected from November 2025 through January 2026 using Flowace workforce-analytics software.

Has the study been peer-reviewed?

No. The paper is a preprint under peer review. Its authors say the findings may change during that process.

Does the study show that AI improves productivity?

No. It measured application activity and switching, not whether work quality or productivity improved.

Source: ThorstenMeyerAI.com

You May Also Like

Your 2026 AI Automation Roadmap Starts Here

Learn how to develop your AI automation strategy for 2026 with key tools, platforms, and best practices outlined in this comprehensive guide.

A Skill Is A Folder, Not A Prompt: What Anthropic Learned Running Hundreds Of Them

Anthropic reveals that skills are folders containing instructions, scripts, and assets, transforming how organizations build durable AI capabilities.

Cutrova: Edit the Words, Not the Timeline

Cutrova introduces a local-first, transcript-based video editing tool that simplifies editing by focusing on words, not timelines, enhancing privacy and accessibility.

Maximizing Speech Recognition AI Accuracy With Benchmark Performance Metrics

Hugging Face researchers introduce tests revealing that leading open-source speech recognition models may overfit to benchmarks, impacting real-world accuracy.