A study by MIT economist David Autor, published as Working Paper 35720 by the National Bureau of Economic Research, examined the impact of AI on 133 patent lawyers. The three-month trial involved participants from 11 US intellectual property firms using a custom Google tool called InFlow, which has since been integrated into Gemini Notebook.
During the trial, drafting quality improved by 0.34 standard deviations after 10 days and by 0.38 standard deviations after 90 days. Junior lawyers completed tasks an average of 18 minutes faster. However, the researchers noted that while AI improved output, it did not necessarily enhance the underlying skills of the users.
When tested without AI, participants who had used the tool outperformed the control group by 0.32 standard deviations. Senior lawyers showed a 0.45 standard deviation advantage over their peers. Among junior lawyers, the study observed a polarization effect, where mediocre results decreased while both strong and weak performances became more frequent.
The findings suggest that foundational expertise is necessary to derive long-term value from AI. The researchers, including co-authors Tanya Rodchenko and Josh Martin, noted that the study was limited to patent law and remains a working paper. It remains unclear if these patterns apply to other professions such as coding or writing.
