The development
What happened
Google Research reported on an existing patent-lawyer trial that separates AI-assisted output from unassisted judgment. Drafting improved with tool access, but later error-correction gains were concentrated among experienced lawyers.
Key facts
The three-month trial involved 133 patent lawyers at eleven firms doing regular, non-exclusive business with Google; two-thirds per firm received AI access.
AI access increased assisted drafting-quality scores by 0.34 standard deviations after 10 days and 0.38 after 90 days.
On a day-90 patent-correction task without AI, treated seniors with at least seven years' experience scored 0.45 standard deviations above controls.
Juniors showed no average improvement on the unassisted task, with more high and low scores. The October post discusses an earlier NBER working paper.
The context
Why it matters
The distinction matters for workplace training: producing stronger drafts with AI does not by itself demonstrate better independent judgment. Teams assessing adoption should measure both assisted output and performance without the tool. This specialized, three-month study supports that evaluation approach; it does not establish lasting skill loss among junior workers or effects across every profession.