Weil Gotshal has reportedly lost close to 30 partners this year, including the chair of its corporate department, who left for Cravath with five M&A partners and a pay package reported north of $20 million. Merger talks came and went. I want to be direct about why I think this matters to people who will never set foot in a Big Law office: the same force that is pulling Weil apart is about to make a four-lawyer defense shop more capable than it has ever been. Here is the argument, in the order I'd make it to a skeptic.
What just happened
Big Law has become a free-agent market. Client relationships belong to individual partners, not to firms, and the firms are now bidding for those partners the way sports teams bid for players. Reporting on the Weil exits put guaranteed packages for the top names at $20 million to $40 million.
The part of that story I'd underline is the one that got less attention. Leadership could not sell its senior partners on a world where high-end legal advice is worth less, because AI is doing more of the work and $10 million-plus packages are eating what is left. The AI half of that is my market.
Every managing partner is now squeezed from two directions. Partner compensation pushes on margin from the top. AI pricing pushes on it from the bottom. When that happens, the question in every firm becomes cost per matter, and litigation is where the associate pyramid is most exposed, because litigation is where the hours pile up.
Why it happened
Look at what the rainmaker model actually needs. It needs the relationship and it needs the rate premium. It does not need 20 associates. It has 20 associates because the hourly model made leverage, the ratio of associates billing under each partner, the only way to monetize the relationship. The partner's income was never really their own hours. It was the spread on everyone else's. The pyramid was never the product. It was the billing mechanism.
So run the thought experiment. A $100 million private equity relationship gets serviced by 8 lawyers instead of 40. The client keeps paying for the partner's judgment, now under a fixed fee. The rainmaker and the firm get richer, not poorer. That is the Kirkland future, and it is why the top of the market will buy AI aggressively, not reluctantly. Some already are. Our partners at Reed Smith saw the writing on the wall early. The enterprise features we built on the platform with them exist because they wanted to be paid for what they deliver to clients, not for the hours underneath it.
The firms with the hardest adjustment are in the middle, and not because they lack judgment. A partner at a 200-lawyer firm may read a judge as well as anyone at Kirkland. But she bills $700 an hour, not $2,000, because the client picked her firm partly on price. That means her firm's margin leans on the volume of associate hours under her, and when AI takes most of those hours out, she cannot make it up by tripling her rate. Price was part of why she won the work. Kirkland loses hours it never needed. The middle loses the hours that were carrying the margin, and has to reprice with less room to do it. Some will manage it well: smaller teams, fixed fees, a higher partner rate justified by faster outcomes. The ones in trouble are the firms whose only real pitch was cheaper hours and enough people to bill them.
There is a quieter consequence, and I think it is the one that compounds. If associates are no longer the profit engine, nobody has a business reason to develop them. The pyramid was also the training program: you learned to try a case by carrying the bags on forty of them. Meanwhile the partners whose judgment took twenty years to build, the ones doing the work clients pay the premium for, leave to start smaller firms where lawyers lawyer: tighter teams, better outcomes, lower cost. Big Law gets more expensive at the top and weaker underneath at the same time.
And then there is the long tail, which is where I come from and where TrialKit lives. Here there is no rainmaker economics at all. The client is assigned by a court or lands in a public defender's caseload. The fee is capped. The discovery volume shows up whether anyone wants it or not. Same tool. Opposite economics. And here is the part I did not expect: the product we built for a buyer with no pyramid to protect and have been shipping to public defenders and defense firms for two years, is the thing the top of the market is now scrambling to figure out.
The asymmetry
AI is worth more to the firm that lacks the infrastructure than to the firm that already has it.
Kirkland with Harvey gains a cost reduction. A real one, but a reduction on a line item it already had covered with associates, contract reviewers and vendors. It may not even want to pass that saving through to clients.
A four-lawyer defense shop with TrialKit gains a capability it never had. Before, the body cam footage got watched by the lawyer at 11pm or it did not get watched. The 40,000-page production got searched by keyword or it got skimmed. There was no associate pyramid to replace, so the tool does not replace anything. It adds a capacity that previously did not exist at that price point.
And that capability does not stop at the cost line. A defense lawyer's revenue is gated by hours, not by demand. There are always more clients, more court appointments, more cases a solo would take if they could physically get through the discovery. If the body cam review that took 400 hours now takes 40, they take the next case. Or they take the bigger case, the one with the cellphone dump and 60 hours of footage that they would have turned down last year. Fewer hours per case and more cases per year is a revenue story, not just a margin story.
Same software category. One buyer saves money on work it was already doing. The other buyer does work it could not do before, and gets paid for it.
The objection, and the answer
I'll make the objection for you, because it is a fair one. The reason investors fixate on enterprise legal is not stupidity. Small-firm legal software has historically been a terrible business: low contract values, high churn, price-sensitive buyers, no procurement budgets, and every sale a founder-led conversion. Clio is the exception, and it took a decade.
If my pitch were "the VCs are wrong about enterprise," the next question would be "then why are small-firm economics different this time?" Here is the answer.
In litigation, discovery volume is imposed from outside. The prosecution or the opposing party dumps the terabytes, and the defense has to review them. That is true whether you are a public defender on a salary, a CJA panel attorney on a capped voucher, or a solo on a flat fee. Nobody on the defense side chose that volume, and nobody gets paid more for it.
Which means AI in my market cuts a cost, not a revenue line. Nobody on the defense side wants to bill 400 hours of body cam review. Most of them cannot bill it at all. The hours are a loss they absorb to do the job right, or a corner they cut because they have no choice.
That is the exact inversion of the Big Law problem. At Weil, AI threatens the thing the firm sells. In a public defender's office, AI removes the thing that was eating the office alive. One of those buyers has every reason to drag their feet. The other has none.
The whole thing, compresed
AI collapses the associate pyramid, and the hourly model cannot absorb that without repricing. Value accrues to whoever is paid for outcomes rather than hours: boutique litigation firms, public defenders, contingency and flat-fee litigators, government. Discovery is the sharpest case, because the volume is externally imposed and the cost is unrecoverable, so for the defense buyer AI is margin recovery on every existing case and capacity for the cases it could not take before. Nothing about it cannibalizes revenue.
Weil is what happens to a leverage-only business when the leverage stops being valuable. I spent years as a public defender on the other side of that pyramid, drowning in the discovery it produced. I did not expect the same technology to threaten one end of the profession and rescue the other. But that is what the economics say, and I'd rather build on the end where the buyer wants the tool to work.