AI At Work: A Harvard Study On App Switching And Productivity
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: AI At Work: A Harvard Study On App Switching And Productivity on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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 analysed 103 million application events from 1,017 workers at eight Indian companies. Workers switched apps less in the ten uses after an AI interaction, but made about 200 more switches on AI-use days; the observational study cannot show that AI caused either pattern.

A Harvard Business School preprint analysing 103 million application events found that workers switched between apps less in the ten app uses after an AI interaction, yet made about 200 more switches on days they used AI. The results, drawn from 1,017 employees at eight companies in India, describe associations rather than cause and effect and do not establish whether AI improved productivity.

The researchers, Sumer S. Vaid and Ashley V. Whillans, examined computer activity recorded from November 2025 through January 2026. Across 62,693 person-days, employees made a mean of 908.3 switches per day; the median was 814. The study counted a switch when a worker changed the active application, not when they moved between documents within the same app. The middle half of employees recorded between 547 and 1,130 daily switches.

The AI findings point in two directions. In the ten application uses after an AI use, workers used fewer apps and switched less, while staying about half a second longer in each app and following a more predictable sequence. Across whole days, however, AI-use days had about 200 more switches, and days with more minutes of AI use were more fragmented. The paper does not resolve why those patterns differ.

The records came from Flowace, a commercial workforce-analytics service whose monitoring software logs application changes on employees’ computers. Researchers received anonymised data under an agreement. The employers determined which applications counted as productive. The study covered software, accounting, finance, e-publishing, law and marketing organisations; it did not measure work quality or establish whether switching helped or hindered employees’ output.

At a glance
reportWhen: Preprint first posted July 7, 2026, rev…
The developmentA Harvard Business School preprint reports that AI use was associated with fewer app switches immediately afterward but more switching across the full workday.
Crypto market snapshot
Fear & Greed Index
12/100 — Extreme Fear
Bitcoin BTC$61,568▼ 3.5%
Ethereum ETH$1,745▼ 3.3%
Tether USDT$0.9992▼ 0.0%
BNB BNB$570.15▼ 3.4%
USDC USDC$0.9996▼ 0.0%
XRP XRP$0.9572▼ 2.9%
Solana SOL$72.11▼ 3.2%
TRON TRX$0.3085▼ 1.2%
Live data · CoinGecko · alternative.me (24h 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.

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 Fewer Switches

The findings complicate claims that adding AI tools automatically gives workers more focus. The immediate sequence after an AI interaction looked less fragmented, but the day-level association ran the other way. A short-term pattern around a tool cannot by itself establish what happens to a person’s workload, attention or output across a full day.

For employers, the distinction matters because the study measured application activity, not productivity. A high switch count is not proof that work was ineffective, and a lower count is not proof that work improved. The paper also suggests that fragmentation varies substantially within the same worker: 44.6% of its variation came from day-to-day differences, compared with 35.8% from differences between people and 19.6% between organisations. That points to the workday and its demands as relevant factors, not just individual habits or company policies.

The evidence therefore does not support purchasing an AI product on the promise that it will reduce interruptions. It can inform questions about when workers face heavier communication demands or need protected concentration time, but answers about the benefits of any particular tool would require further evidence, including measures of work quality and research capable of testing causation.

Products Worth Considering

Trezor Safe 7 Crypto Hardware Wallet with Bluetooth and Color Touchscreen

Trezor Safe 7 Crypto Hardware Wallet with Bluetooth and Color Touchscreen

Apple MacBook Pro 14.2-inch with M5 Pro, 15-Core CPU, 16-Core GPU, 24GB Memory, 2TB SSD, Wi-Fi 7, Space Black

Apple MacBook Pro 14.2-inch with M5 Pro, 15-Core CPU, 16-Core GPU, 24GB Memory, 2TB SSD, Wi-Fi 7, Space Black

As an affiliate, we earn on qualifying purchases.

How the Harvard Study Was Built

The paper, titled 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 first appeared on arXiv on July 7, 2026, and was revised on August 7. It remains a preprint under peer review; its authors say findings may change during that process.

Flowace software recorded the active application and the start and end of its use, alongside other activity details. The sample comprised 1,017 employees at eight Indian organisations, with a median of 58 tracked workdays per person. It did not include phone activity or workers in other countries. The study received no external funding, according to the paper; Flowace provided no funding or commercial support and had no role in study design or analysis, the authors state.

The reported average focus period in one application was 27.8 seconds, while the median was 3.3 seconds. The authors compare this with earlier attention research reporting longer periods, but differences in methods mean those figures should be read as a broad comparison rather than a directly measured historical decline.

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

— Sumer S. Vaid and Ashley V. Whillans, the paper’s authors

Causation and Productivity Remain Unknown

The study was observational, not an experiment. It cannot show whether AI caused more switching, whether workers reached for AI on already fragmented days, or whether another factor contributed to both. The authors explicitly leave open the possibility that the day’s work pattern influences AI use, that AI changes the pattern, or that both are true.

The analysis also does not establish whether workers completed tasks faster, produced better work or experienced more distraction. The sample is limited to employees at eight organisations in India and activity on work computers tracked through employer-run monitoring software. The authors’ results remain subject to peer review, and the paper warns they could change.

Peer Review and Better Outcome Measures

The next formal milestone is peer review, which may lead to revisions or changes in how the results are presented. The paper does not specify a review timetable. Further research would need to test whether AI use changes switching, and assess outcomes such as task completion, work quality and employee experience alongside application activity.

Until then, the study offers a description of recorded computer activity in a defined sample, not a general verdict on AI’s effect on workplace productivity. Readers and employers should treat its app-switching figures as measured patterns, not proof that an AI tool either improves focus or causes fragmentation.

Key Questions

What did the Harvard study find about app switching?

Across the tracked workdays, employees averaged 908.3 switches between applications per day, with a median of 814. The study covered 1,017 employees at eight companies in India.

Did workers switch apps less when they used AI?

The answer depends on the time frame. In the ten application uses after AI use, workers switched less. Across the full day, AI-use days had about 200 more switches. The study does not explain the difference or establish cause and effect.

Does the study show that AI makes workers more productive?

No. The researchers measured application activity, not task quality, output or overall productivity. The findings cannot establish whether AI improved or harmed work outcomes.

Is the research peer-reviewed?

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

Who supplied the activity data?

The data came from Flowace, a workforce-analytics service used by the participating organisations to record application activity on employees’ computers. The authors say the data were anonymised before they received them.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

AI Automation Software For Small Businesses: A Buyer’s Guide

A buyer’s guide compares Zapier and Make for small businesses, from setup and integrations to AI workflows, costs and human review.

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

Anthropic reveals that effective AI Skills are structured as folders containing instructions, scripts, and assets, transforming prompt engineering into durable organizational assets.

How Mini PCs Fit Into Low-Clutter Trading Desks

Mini PCs seamlessly integrate into low-clutter trading desks, offering a sleek, organized setup that enhances efficiency—discover how they transform your workspace.

OpenSSH SSHole Bug: Where Overlong Passwords Meet Server Security

CVE-2026-123158 can bypass SSH passwords on specific musl-based Linux setups. OpenSSH 10.6p2 fixes it; administrators should update and check logs.