The Machines in the Basement

Most AI-in-law coverage asks whether AI will replace lawyers. That’s the wrong question for this book, and probably for the next 10 years too. The question that matters is smaller. AI doesn’t have to replace experienced lawyers to hurt law schools. It just has to replace the work new lawyers do.

What New Lawyers Actually Do

Spend time inside a big law firm and you’ll notice something about first-year associates. They’re expensive to hire. They bill cheap. And they spend most of their time on work that looks nothing like what they pictured in law school. Document review. First-pass legal research. First drafts of memos. Summaries of deposition transcripts. It’s high-volume, repetitive work. A client pays a premium for it partly because it has a JD attached, and partly because, until recently, there wasn’t a faster way to get it done.

Steven Harper explained the logic in 2013.¹ Firms hire far more entry-level associates than they will ever make partner. They bill that junior work at about a 4-to-1 markup over what they pay the associate. Before the 2008 recession, more than 80 percent of associates were gone within 5 years.¹ That pyramid is the revenue model. Lots of juniors come in, most leave within 5 years, and they get billed hard the whole time.

That only works if the junior work justifies the headcount, and that’s what’s under pressure now.

What the Data Actually Says (No Overclaiming)

Look, I want to be careful here. Force 9 is the most speculative of the 9, and the book’s credibility depends on saying exactly what’s true and not 1 inch more.²

As of mid-2026, the hard data does not show AI gutting entry-level legal hiring. New law-graduate hiring numbers are about flat. Entry-level law-firm salaries dipped about 3 percent.² Firms are slowing the pace of associate hiring and trimming summer associate programs. But no one is pointing to a collapse in first-year class sizes and blaming a chatbot.²

The full hit has not landed. Write that down.

What is happening is the mechanism. AI tools now handle first-pass document review, initial research, and draft generation, at a speed and cost that didn’t exist 3 years ago. The work hasn’t disappeared from law firms yet, but the reason to pay a 4-to-1 markup for a human to do it is quietly getting weaker. A managing partner can now run the numbers on AI-assisted document review against a first-year billing several hundred dollars an hour. That calculation didn’t use to be this easy.²

The real thing under pressure is the training pipeline, and it matters more than this year’s hiring number. A junior associate learns to be a lawyer by doing the work. The doc review that feels like a grind is how a first-year learns to spot the clause that matters. The first draft that comes back covered in red marks is how they learn to write precisely. If that work gets automated before they do it, the path from new hire to competent attorney breaks. Firms are starting to feel this, and they don’t have a clean answer yet.

So here’s the honest description of Force 9 as of mid-2026. The mechanism is visibly in motion, hiring is slowing, the training pipeline is under structural pressure, and the full hiring impact is still ahead of us.

The Irony, and Why It Hits Specific Schools Hard

The 2025-2026 application surge is being driven partly by AI.³ LSAC has noted that applicants cite 2 main reasons for applying. One is the political climate. The other is a sense that AI is making other careers feel unstable, so a law degree looks like safer ground.³ A generation of smart people is running into law school to escape AI disruption.

They graduate in 2028 and 2029, into the market AI is most actively reshaping. They’re running straight at the disruption. We flagged this in Chapter 3, and it deserves more weight here. The schools that accept the most of these students, the ones discounting hardest to fill the biggest classes they can, are concentrated in the middle and bottom tiers.

That’s the real distribution problem. Top graduates from elite schools will be fine. The strong students, the ones who had options going in, have options coming out. Big Law. Federal clerkships. Strong regional markets. AI goes after routine volume work done by large classes of people hired because the pyramid needed bodies. A third-year associate with a federal clerkship and a strong research record at a top-20 firm isn’t doing that work.

The roughly $75,000 regional-firm associate job is the most exposed. That job absorbed the middle of every graduating class for decades. It was never glamorous, but it was steady, and it gave you a foothold. It exists because midsize firms had routine high-volume work and needed someone to do it at a manageable rate. AI handles a lot of that work now, or soon will. A midsize firm running lean doesn’t need to hire 4 associates when 2 and a subscription cover the same volume.²

The schools that filled seats by selling that middle job are the ones left holding the risk.

The Historical Proof of Concept

The 2009-to-2011 “lost classes” show what happens when the top of the legal market stops hiring and the shock ripples down.¹ When Big Law froze first-year classes in 2009, it didn’t just hurt Harvard and Columbia students. It cascaded. Elite-school students who didn’t land Big Law competed harder for second-tier jobs, which pushed the students aiming at those jobs down a tier, and the students at the bottom found the floor pushed out from under them. A disruption at the top pushes the whole salary distribution down.¹

Harper called those years the “lost classes” because those students lost more than a first job. They lost years of career compound interest.¹ The people who graduated in 2009 and 2010 and couldn’t find a footing never fully caught up to their counterparts from 2007 or 2012.

The recession did that on a cycle, and the market came back. AI is doing a structural version of the same thing. When the work is automated, the jobs don’t return with the economy, because the reason to hire has changed.

That’s the part the current data doesn’t capture yet. The 2026 hiring numbers are flat because firms are still figuring out how to restructure around the new tools. The 2028 and 2029 numbers are the ones to watch, and that’s right when the students flooding into law school today graduate. The timing is almost perfectly bad.

The schools absorbing this incoming class are discounting hard to fill seats. They’re building enrollment projections on the assumption that the jobs their middle students are counting on still exist in 2028. Those schools will look back on the 2026 application boom the way schools of 2010 looked back on theirs, as the thing that set up the fall.

Watch: what AI actually changes for lawyers

Will AI Replace Lawyers?

This chapter is from Anyone Can Beat the Law School Collapse by Steve Schwartz, LSAT Coach and Founder of LSAT Unplugged. I've been coaching the LSAT since 2005.

Published July 28, 2026.