- Who it helps
- Managers and educators evaluating how AI affects professional learning.
- Availability
- The October 7 research account is public. It reports a three-month randomized study with 133 patent lawyers at eleven firms.
What the study found
The authors report better assisted work overall. In later unassisted redlining tasks, senior participants improved; juniors showed no average gain and their outcomes diverged.
What a team can measure
Evaluate a finished deliverable and a later unaided task separately. Keep the difficulty comparable and look at experience levels. This helps identify whether an apparent productivity gain also improves understanding.
What you can do
- Read the study’s task and participant limits before designing a learning evaluation for your own team. Open reference ↗
Where the evidence stops
This is one profession and a three-month study. It does not establish that AI improves or harms learning in every setting.
Sources read. Public source material was read October 17, 2026 UTC. No model inference or third-party project was executed.
Sources & further reading
Web · paper
Does better work always mean better workers? ↗
October 7 account reports a three-month randomized workplace study with 133 patent lawyers.
Google Research · page checked 2026-10-17 · Published 2026-10-07
Coverage note: The cited bundle covers the available primary explanation. Matching original X posts, a complete video comparison and independent execution evidence were unavailable.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
A Google Research study separates the quality of work produced with AI from what people learn to do without it. Those two outcomes can move differently.