AI in law will reward proof, not enthusiasm
The firms that win will be the ones that can show where AI saves time, where it creates risk, and how that changes pricing.
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.
What does AI do to a law firm’s margin? That is the question a partner should be asking, and the easy answer is the one everybody is already selling: more speed, more output, more adoption. The better answer is that speed only matters if you can measure what it did to timekeeping, profitability, and return on investment. Thomson Reuters and Laurel said exactly that this week, tying AI use in tools like CoCounsel Legal to those numbers rather than to hand-waving about usage. That is the useful part of the news.
Attention is cheap. Proof is expensive.
The market is moving toward management, not demos
This is not happening in isolation. LexisNexis is rolling out new AI features across drafting, workflow automation, and agentic workflows in its legal and business products. Microsoft’s legal team is expanding Harvey across corporate, external, and legal affairs operations. Law.com reports that 132 people now hold AI ownership roles across law firms, with many firms naming a chief AI officer or assigning AI to innovation or knowledge leaders. Latham & Watkins says it has launched an AI Academy for its lawyers.
The real competition is operational
The firms that will matter are the ones that can tell a client, with a straight face, where AI saved time, where it introduced risk, and what happened to the bill. That sounds obvious until you notice how many firms are still treating AI as an internal pilot with a logo. Latham also says its digital infrastructure team surpassed $1 trillion in transactions during the first half of 2026 and continues to work on data centers, chip factories, and other AI infrastructure needs. JPMorgan Chase says its internal LLM Suite is used by more than 200,000 employees and that AI spending could reach up to $2 billion annually. Kirkland & Ellis says it is co-developing an AI platform with Palantir as part of a $500 million plan. Blackstone says it invested $50 million in Norm Ai. None of that is cosmetic.
Risk is already priced in
The cautionary side of the ledger is just as clear. Our database tracks 743 AI-related sanctions, with New York, California, and Oregon at the top of the list by state. That does not mean every AI mistake ends in discipline. It does mean the profession has already crossed the point where sloppy use of machine output is treated as a training issue only. A recent report from Howden says AI can magnify errors, omissions, inaccurate advice, and confidentiality problems if guardrails are not in place. Another report from Shumaker says insurers are already responding with AI-specific exclusions, narrower endorsements, sublimits, and tougher underwriting. The money is learning what the bar has known for a while: somebody eventually pays for bad output.
Measurement is the wedge
That is why the Thomson Reuters and Laurel move matters more than another announcement about features. Firms do not need more AI theater. They need to know whether the tool shaved time off a task, whether that time was billable or lost, and whether the pricing model captured the gain or gave it away. Our site tracks 239 legal AI tools, and the ones drawing the most attention right now are Claude for Legal, CoCounsel, and Harvey. Attention is cheap. Proof is expensive.
The firms that figure out measurement first will not just use AI better. They will know what to charge for it.