A Harvard Study Found AI Is Making Junior Lawyers Faster — and Partners Irrelevant

When you enter a large law firm in Manhattan or London on a weekday evening, you’ll still notice that the lights are on, that associates are bent over screens, and that the glass meeting rooms smell like cold coffee. The actual content on those screens has changed. It was a Word document and a Westlaw tab six years ago. A question that used to take a junior associate four hours can now be answered in about forty seconds thanks to the chat window.

A finding from Harvard-affiliated researchers and a parallel survey from LexisNexis released in February both center on that change. The headline figure is almost too clear: 58% of attorneys who use AI tools claim to be producing legal work more quickly, and that number rises to 65% among those who pay for premium legal AI platforms. The more intriguing figure, however, is buried deeper: according to 72% of respondents, the largest skill gap among juniors is in deep legal reasoning. Yes, faster. Not necessarily sharper.

Speaking with professionals in the field gives me the impression that the productivity narrative is the simple part of this. What happens to partners is the difficult part. The law firm pyramid was successful for decades because partners assigned associates to handle the laborious, painful, formative work, such as first drafts, document reviews, and case research. The work was intentionally inefficient. That’s how a 26-year-old with a $400 million deal came to be trusted. If you remove that, the pyramid loses its meaning.

Given that they still maintain client relationships and that clients continue to pay for a person with gray hair on the call, it’s possible that the partners aren’t entirely irrelevant. However, their economic reasoning—the part where you pay a junior $600 per hour to do something that now costs almost nothing—is quickly disappearing. In private, some managing partners have been open about this, but less so in public. Last winter, one informed an industry researcher that his juniors were responding to questions more quickly than he could process them. He sounded unproud of it.

A layer that most legal-tech conferences omit is added by the UC-Berkeley research. According to the study, generative AI increases work rather than decreases it. Tasks grow. Boundaries become hazy. Because typing into a chat box doesn’t feel like work, lawyers put drafting into the evenings. In a field that already consumes its youth, burnout worsens rather than improves. The promise of freedom turns into a slightly faster treadmill.

For more than a year, David Wilkins, the director of Harvard’s Center on the Legal Profession, has been predicting this kind of instability. In general, he believes that the demand for attorneys is increasing rather than decreasing, but the type of attorney who prevails is evolving. integrated thinkers. Individuals who can collaborate with legal tech companies, the Big Four, and whatever comes next. It’s possible that the outdated apprenticeship model, which taught judgment through thousands of repetitions of low-value labor, won’t last for another ten years.

As this develops, it’s difficult not to question the true origins of the upcoming generation of senior attorneys. What exactly are juniors becoming if they never struggle with a poor first draft, never understand why a clause is risky on their own, and never experience the slight embarrassment of a partner red-penciling their memo at midnight? Yes, faster. Perhaps hollowed out. In 2035, the companies that figure this out first and approach AI as a collaborator rather than a quick fix are likely to survive. The others will realize a bit later that efficiency without discernment is just costly noise.