Standard rates at the largest US firms keep climbing, and top partners at elite shops are now billing four figures an hour. Over the same stretch, law firm technology spending has grown at its fastest pace in years. Read those two trends together and the tension jumps out: firms are charging more for lawyer time while spending heavily on tools built to shrink the time any given task takes.
That is the story reshaping the economics of legal work. Discovery, document review, and first-draft writing were the load-bearing walls of the billable hour. Generative AI is knocking on all three at once, and the pricing model built on top of them is starting to move.
Why Are Firms Charging More While Buying Tools That Do the Work Faster?
The short answer: demand is up and clients are still paying. Standard rates have climbed at the fastest pace since the financial crisis, even as AI absorbed work that used to fill timesheets. Firm leaders are, for now, defending existing rates rather than repricing around efficiency.
The longer answer is uncomfortable. Roughly 90% of corporate legal spend still flows through hourly arrangements more or less unchanged since the 1950s. When a research memo that used to take ten hours can be produced in one, the firm has three options: bill the hour anyway, bill the one and absorb the write-down, or reprice the deliverable. Each choice builds a different business.
What Actually Changes in Discovery and Document Review
Discovery is where the economics move first, because it is where volume lives. Predictive coding and technology-assisted review have been judicially blessed for more than a decade, going back to Judge Peck's Da Silva Moore ruling. Generative models are the next turn of the same wheel, and courts are beginning to handle them under the same acceptance framework, with EEOC v. Tesla among the first matters to acknowledge AI-assisted review inside an ESI protocol.
For the firm, the practical shift is that first-pass review stops being a staffing exercise. Contract attorneys and junior associates no longer sit in front of a review platform for weeks. A smaller team supervises a model that classifies, clusters, and summarizes at a speed no human queue can match. The billable inventory attached to that work shrinks.
Where Does Drafting Fit In?
Drafting is the second pressure point, and the one associates feel personally. Motion shells, discovery responses, deposition outlines, transaction checklists, first-cut NDAs, closing memoranda, the connective tissue of a matter, now come out in draft form from models trained on a firm's own precedent, then get edited by the lawyer whose name goes on the filing.
The output is not always right. It is often close enough to save real hours, which is the point and also the problem. Close-enough drafts push the value of the work up the stack: away from producing the document and toward judging it, tailoring it to the client, and standing behind it. That work looks nothing like the associate job description the pyramid was built around.
What About the Ethics of Billing for AI Work?
State bars and the ABA have started drawing lines, and those lines matter for pricing. ABA Formal Opinion 512 concluded that a lawyer generally cannot bill an hourly client for time the lawyer did not actually spend, even when AI produced the deliverable in a fraction of the previous time. Charging ten hours for one hour of work, on the theory that the output is worth ten, is not a defensible reading of most engagement letters.
That guidance nudges firms toward the alternative fee arrangements clients are already asking for. If you cannot bill phantom hours, and you cannot make the economics work by billing only the compressed hours at the old rate, the deliverable has to be priced. Flat and value-based fees are gaining ground for a mechanical reason, not a philosophical one.
How Should a Firm Actually Adapt?
The firms handling this well are not the ones with the flashiest chatbot. They are the ones running AI as infrastructure, governed, integrated, and measured, rather than as a plugin. That means private deployments over confidential matter data, human approval gates on anything that leaves the building, and honest internal metrics on where the tools save time and where they don't.
It also means rewriting the compensation and training pipelines that were built around a pyramid of junior hours. If juniors are not spending a year doing document review, they need another way to learn how a case gets built. If partners are pricing by matter rather than by hour, origination credit and profit-sharing have to follow. For a fuller walkthrough of how these pieces fit together across intake, review, drafting, and firm operations, Law.co's guide to how legal AI is changing the industry guide to how legal AI is changing the industry is a useful reference.
The billable hour is being unbundled rather than buried. Some work will keep the clock: bet-the-company litigation, novel regulatory questions, the kind of judgment calls no client wants a fixed fee to constrain. The rest is increasingly getting priced the way many clients have long preferred to buy it: by the thing, not by the minute.
