Nearly six in ten in-house counsel say they have seen no meaningful savings yet from their outside firms' use of generative AI, even as those same firms tout the technology in every pitch meeting. That figure, from recent ACC/Everlaw research, is the single most important number a CEO can carry into the next conversation with their general counsel. It reframes the whole promise the industry has been making.
The technology works and the efficiency is showing up inside firms, but the invoice, so far, has not caught up. Understanding why is now part of the job of anyone who signs the legal budget.
The Missing Savings Are the Story
Hold that six-in-ten figure in view for the rest of this piece. It is not proof the technology fails. It is proof the pricing model has not caught up to what the technology does.
A first-year associate used to spend fourteen hours on a document review that a well-tuned model now finishes in twenty minutes with a lawyer checking the output. Under hourly billing, the firm bills the twenty minutes plus the review time, and the client sees a smaller line item on that task.
What the client rarely sees is the total. Saved hours get redirected to other work on the same matter, or the rate on the remaining hours ticks up to protect firm revenue. The savings exist somewhere in the system, they just are not landing on the invoice.
The Invoice Hides the Efficiency
The reason has a name inside the profession: the efficiency paradox. Faster work under an hourly model produces fewer billable hours per matter, which pressures firm revenue, which pressures rates. And rates keep climbing. Worked rates at U.S. firms grew 7.3% in 2025, more than double inflation, according to Thomson Reuters and Georgetown Law.
A CEO reading the year-end legal spend sees the top-line number go up while their firm reports strong internal productivity. Both things can be true at once.
There is also an ethics wrinkle worth knowing. Recent ABA guidance reminds lawyers who bill hourly that they can only bill for time actually spent, not for the time the task would have taken without AI. In practice, that puts a firm using AI heavily under real pressure to either lower the hours it bills or move away from the hourly model entirely.
The Number Should Push You to Ask for More
Savings that do not show up on their own have to be negotiated in, and that is the concrete thing a CEO can do this quarter. A short list of asks, aimed at the firms already on the roster:
- A fixed fee on repeatable work. Diligence, standard contract review, first-pass document review, routine filings — anything the firm has done more than a dozen times is a candidate. If the firm has automated it, the price should reflect the automation, not the associate hours it displaced.
- An AI provision in the engagement letter. More companies are writing outside counsel guidelines that require disclosure of AI use, human review standards, and data handling. Ask for it in writing before the next matter opens.
- A blended rate that assumes the tech. If a matter still runs hourly, price the blended rate as if AI-assisted work is the baseline. It already is inside the firm.
- Line-item transparency on AI-assisted tasks. You do not need a technical audit. You need to know which tasks were AI-accelerated and what the firm believes that saved you.
The Quality Question Sits Underneath All of This
Price is one half of the conversation. Accuracy is the other, and any CEO pushing on price has to push on quality in the same breath. Leading legal research tools still hallucinate — invent citations, misstate holdings — at meaningful rates, and the general-purpose chatbots consumers use are worse.
That is why the firms taking AI seriously are investing in review layers, private deployments, and audit trails rather than turning associates loose with a public chatbot. Mashable's coverage of the challenges facing legal coverage of the challenges facing legal AI adoption adoption is worth passing to your general counsel before the next vendor conversation.
The practical read for a CEO: cheaper legal work is only cheaper if it holds up. Ask the firm what happens between the model's first draft and the version that reaches your desk. If they cannot describe that step in a sentence, the price cut is not worth taking. The CEOs who see a different invoice next year are the ones asking for a different invoice this year — fixed fees where the work is repeatable, written commitments on how AI is used and reviewed, and a frank conversation about what the firm's rate card assumes about its own productivity.