AI in law is becoming a hiring and training bill, not a toy
The money is moving into training, infrastructure, and internal systems, and that means the real fight is who pays to make lawyers faster without making them sloppier.
An AI-assisted editorial, reviewed by a human before publishing. It reasons over our own tracker data (and cited context) — a point of view, not legal advice.
The strongest argument against my view is simple: the profession has not broken by adopting too much AI, it has broken when lawyers used it badly. A New York City lawyer filed a brief riddled with fake cases after using Microsoft Copilot and got fined and referred for discipline. Defense counsel in Coomer v. Lindell filed nearly 30 defective citations and cases that do not exist, then was sanctioned after admitting unverified AI use. A California custody fight over a dog ended with a $5,000 fine after counsel filed briefs with fabricated cases. That is the part of the story everyone should keep in view.
I still think the bigger shift for a working lawyer is not courtroom embarrassment. It is a budget fight. Firms and legal departments are putting real money into AI training, internal platforms, and control layers because they now expect AI to sit inside ordinary work, not outside it. Kirkland & Ellis says it has earmarked $500 million of its revenues to build its own platform. Latham & Watkins says it has launched an AI Academy. Microsoft’s Corporate, External, and Legal Affairs organization, about 2,000 lawyers and compliance professionals, will use Harvey for legal and compliance work. That is not gadget buying. That is payroll strategy.
The profession does not need a miracle. It needs systems that make review cheaper than correction.
Who spends first
The first money is going to the institutions that can spread the cost. Kirkland’s $500 million commitment is the bluntest signal in the material. Cleary Gottlieb bought Springbok AI and brought AI engineers in-house. Tarter Krinsky & Drogin created a firmwide Office of AI and Innovation. Capital Group is building a global AI enterprise program inside Legal & Compliance. Even Latham’s AI Academy points in the same direction: the expense is no longer just software, it is training people to use it without creating sanctions risk or a mess for the next reviewer.
The market side matches that. Gartner says corporate legal departments need to reinvent talent, data, and workflow strategies because of AI, flat budgets, and rising workloads, and it predicts legal technology budgets will double by 2028. Reuters reported, through the Bloomberg Law piece on associates, that firms are making competitions out of AI use and giving young lawyers billable credit to learn it. That is a sign the cost is being pushed into the training of junior lawyers, which has always been the cheapest place to hide a strategic bet.
Who charges, who gets charged
The firms that charge for judgment are trying to preserve the part clients still cannot do themselves. Wilson Sonsini is advising on AI infrastructure deals. Latham is pointing to AI-related capital solution and financing work. Those are matters where the lawyer is paid to understand risk allocation, not to grind through paper. But the ordinary pressure lands lower down the chain, where the hours used to live. If AI does the first pass, the associate who used to spend all night in the document room is no longer being paid for raw throughput. The client will still pay for the lawyer who can verify, shape, and sign off. They will resist paying the same rate for the busywork.
That is why the training story matters more than the tool catalog. We already have 99 legal AI tools tracked, and the highest-scored ones in the catalog include Harvey, CoCounsel Legal, and RelativityOne at 90 out of 100. Those numbers tell me something plain: the market has plenty of capable software. The bottleneck is not whether a tool exists. It is whether a firm can build the habits, supervision, and internal review so that the tool does not turn into another sanction report.
The real bill lands on the client
The client pays twice if lawyers get this wrong. First through the cost of retraining firms that did not prepare their people. Then through the cost of cleaning up mistakes that should have been caught before filing. That is why the sanctions matter so much even though they are scattered. Our database tracks 28 AI-related sanctions, with New York leading at 5, California at 4, and Oregon at 3. The pattern is not random. It says courts are already charging lawyers for treating machine output like authority.
There is a better way to read the spending now underway. Watts Law says its AI-powered case-qualification platform still keeps every case subject to lawyer review. Turbo Law, which publishes this site, says its platform proposes and the team verifies. That phrasing is boring on purpose, and it should be. The profession does not need a miracle. It needs systems that make review cheaper than correction. If lawyers do not invest in that now, they will spend the next few years billing less for real work, and more for apologies.