AI helps new patent lawyers write better, but they learn less from it

Junior patent lawyers wrote noticeably better drafts with an AI assistant, but when it was taken away, they showed no lasting gain. Senior lawyers did keep an edge. The finding comes from a working paper from the National Bureau of Economic Research, a nonprofit U.S. economics research group. It raises a question for any workplace where new hires lean on AI.

How the study worked

Researchers followed 133 patent attorneys at eleven U.S. intellectual property law firms for three months. The groups were assigned at random. Two-thirds got an unreleased Google Labs assistant called InFlow. The rest worked without AI. Google funded the study.

There were three tasks. In two, the attorneys drafted patent claims, and only the AI group could use InFlow. In the third, called redlining, everyone reviewed an existing patent draft for flaws without AI. Drafting measured work done with AI help. Redlining measured the lawyer's own judgment, which attorneys use daily without tools. Blinded attorneys at an independent firm scored the results.

What the results showed

In drafting, the AI group beat the unaided group at both the 10-day and 90-day marks. Higher scores came mostly from fewer bad ratings and more mid-range ones. Top scores barely moved. Lawyers with fewer than seven years of experience gained much more than seniors.

Redlining flipped the picture. The AI group scored better overall, but the advantage came entirely from senior attorneys. Juniors saw no benefit, even though they had gained the most with AI. Their scores spread out, with more low marks and more high ones. The authors suggest AI sped up learning for some juniors and replaced it for others. The study does not explain why.

A possible explanation

A text analysis offers a clue. Senior attorneys more often restructured the flawed draft and tied their changes to legal principles. Juniors often spotted serious flaws but did not fix them, leaving notes that read like instructions to an AI assistant. That also happened in the group that never used AI. So the authors see it as a gap in junior skills that AI did not close, not a habit AI created.

Limits to keep in mind

• The sample was small, partly because top attorneys bill at high rates.
• The participating firms do not represent professional services as a whole.
• Skills were not measured before the study began.
• Three months is short next to the years it takes to build patent expertise.

Why it matters beyond law

The authors say the tension matters as long as work requires judgment without AI. They note that AI likely will not be available in real time during client meetings and court proceedings.

Other research points the same way. An Anthropic study found younger developers who learned a programming library with AI scored 17 percent worse on a follow-up test. The MIT Media Lab found signs of cognitive debt from outsourced thinking in essay writing. A 2023 BCG study found weaker consultants improved more with GPT-4 than strong ones did.

For managers, the practical point is simple. AI can make new staff look more capable on the page. It is worth checking whether they can still do the core work on their own.