Discover more from Off The Record
AI will invert the biglaw pyramid
an existential threat to the traditional model of biglaw firms
I’ve previously written about why law firms are adopting generative AI quicker than expected. Cece Xie, writer of , extends this line of thinking in today’s guest post. “Software is not a replacement for humans; it’s a replacement for some tasks previously performed by humans,” she writes, “Anyone who proclaims the former is just trying to bait you—or doesn’t have a solid understanding of the value of lawyers.” Her essay, below, discusses the traditional model of biglaw firms and how AI will challenge that age-old structure.
For decades, the economics of biglaw firms has been:
Equity partners get together to share overhead costs and (ideally) cross-sell each other’s services to their clients.
Hire an army of associates to bill at least 2000 hours annually.
Bill clients; get paid; make big money.
This simple model has weathered the vicissitudes of life surprisingly well. Firms delayed or deferred associate start dates after the 2008 recession and the current market slowdown. Partnerships abandoned lockstep compensation structures to appease rainmaker partners defecting to competitors. Even technological advances like the advent of e-discovery didn’t significantly reduce the number of hours billed by associates.
The pyramid structure of biglaw has stood steadfast through it all—shuffling at the top, hiring at the bottom, and depending on the natural rate of attrition in between.
But thanks to AI, this cozy arrangement will be coming to an end—and faster than we think.
I know, I know—it’s so alarmist to hypothesize that risk is coming to a risk-averse profession. And A.I. has existed for years. Hell, a New York Times article from 2011 headlined, “Armies of Expensive Lawyers, Replaced by Cheaper Software.” But as someone who worked on litigation with massive e-discovery as a baby lawyer, I can assure you: there were still armies of expensive lawyers. Maybe fewer than before, but still overseeing armies of cheaper lawyers or predictive-coding results.
Which is all to say: software is not a replacement for humans; it’s a replacement for some tasks previously performed by humans. Anyone who proclaims the former is just trying to bait you—or doesn’t have a solid understanding of the value of lawyers.
In order to grasp the implications of AI, we have to boil down “lawyering” into its essential tasks:
Read. Lawyers read a lot of statutes, opinions, contracts, statements, emails, briefs, memos, articles, etc. etc.
Identify and categorize. As lawyers read, they need to issue-spot and determine whether what they’re reading is relevant to their clients’ situations. In due diligence for a transaction, is the contract term a yellow flag? A red flag? In motion practice, is the precedent cited in the brief applicable or distinguishable?
Analyze. This is the so what of it all. What does whatever you’ve read and identified mean for a body of law? What new obligations or responsibilities attain? The GDPR, for example, placed new notice obligations on certain companies and gave individuals new rights.
Synthesize. This is the second so what of it all. Clients don’t really care what something means for companies or people generally—they just don’t. They care about what it means for them. This is the observable line between a good lawyer and a great lawyer—great lawyers synthesize the law for clients instead of just analyzing law for them. It’s the difference between an off-the-rack piece of clothing and a tailored version.
Draft. Whatever analysis or synthesis the lawyer did, now it has to make its way onto paper (or email). Bring out the Bluebook! Or don’t, if you’re a transactional lawyer—then just bring out the charts.
Schmooze and sell. Law is, at the end of the day, a people business. Lawyers need clients in order to make money, and clients are people (at least for a little while longer1). Schmoozing and selling distinguishes the competent lawyers from the rainmaker lawyers (who may or may not be competent).
Can you already see where I’m going with this? (If so, there might still be time before the AI overtakes you yet!)
E-discovery platforms and the ability to OCR and hit Ctrl+F obviated the need for Task 1, reading. Lawyers, however, still needed to do some reading in order to complete Task 2, identify and categorize. The AI of 2011 then came along and partially eliminated the need for lawyers to spend so much of their energies on Task 2. Instead, lawyers could oversee the AI as it performed Task 2.
ChatGPT was a revelation because it’s capable of not only Tasks 1 and 2 but Tasks 3 (analyze), 4 (synthesize), and 5 (draft). Not perfectly, and not all the time—but it still could give those assignments a good ol’ college try. GPT technology is particularly adept at analyzing and drafting. An attorney used CoCounsel, a software integrated with ChatGPT, to review “more than 400 pages of documents, and the software quickly reviewed them and wrote a summary that pointed him to an important gap in the defense’s case.” In minutes.
These tasks that GPT can now handle are, coincidentally, common tasks for junior associates. From company and transaction summaries to legal research and drafting memos, analyzing and drafting have long been the purview of bright-eyed, bushy-tailed new law grads.
If we follow the capitalistic impulse of biglaw firms to its logical conclusion, this means that junior associates may soon face obsolescence. Why spend an hour figuring out how to explain an assignment to a first-year associate when you can just ask CoCounsel in five minutes? And the initial output will likely be better than a first-year’s initial work product, too.
Given the immense cost-savings that legal GPT products can confer, I suspect the rise of AI in legal tech will coincide with smaller junior associate classes. Gone are the days of 50+ junior lawyers all working on the same document review or due diligence. Instead, a fraction of those junior lawyers will be hired to oversee and QC the AI’s outputs. Junior associates will edit more than they do currently and manage more than they do right now. Juniors will effectively be more like midlevels from the get-go.
This might not sound that bad—after all, it’s great to give associates more responsibility earlier on—if not for the glaring succession problem that such a system sets up. The existing biglaw structure only works because there are more associates jockeying for partnership than there are invites to buy into the partnership. What happens when there are more partners than associates,2 such that even promoting all associates won’t be enough to serve the firm’s existing clients? What then?
I don’t claim to know the answer but am extremely interested to see how law firms tackle this quandary. Will the decrease in supply of associates lead to higher wages and better working conditions? Will biglaw partnerships recruit more heavily from midsize and boutique firms? Will new law grads displaced from biglaw by AI be more inclined to hang up their own shingle? There’s a lot of uncertainty in the future—as there always is—but one thing’s for sure: lawyers will need to embrace their inner entrepreneurs more because the conveyor belt path of biglaw is about to face an existential threat. ◆
For a limited time, you can subscribe to debrief for 28% off of annual subscriptions:
If you are a student without disposable income, un- or under-employed, or a minimum-wage worker, just fill out this form and Cece will comp you a free subscription, no questions asked. If you’d like to donate one of these subscriptions, you can do so here.
I’m looking forward to the day when AI lawyers must pitch to AI clients, while the human lawyers and human clients will either be on endless vacations eating canapés or enslaved by the AI to be 24/7 maintenance technicians.
Apart from the glaring staffing nightmare. If you’ve ever been the only associate in a practice group, you can already predict the potential pitfalls. Lean associate staffing is often code for LOL good luck.