Generative AI promises to streamline routine legal work. Nevertheless, BigLaw clients are buying more of their lawyers’ time. At Bloomberg Law, Roy Strom reports that demand for BigLaw attorneys’ time rose 4.2% in the first half of 2026, compared with an historic norm of about 1.5%. These numbers follow significant demand growth for law firms’ services in both 2024 and 2025, and they support an even greater level of first-half revenue growth (11.7%).
The widespread integration of AI into law firm workflows raises questions about AI’s promised time savings. But, obviously, greater efficiency and more billable hours can coexist. The more interesting questions are how lawyers and firms deploy AI-related time savings across professional activities and clients, as well as who captures the gains when the costs of production fall. Context, some basic economics, and the missing data that affect everything, below the fold.
The context, of course, is that some clients are pressuring BigLaw firms to “cut fees”—or are taking matters into their own legal departments—to reflect the ability of AI tools to assist in routine legal work. AI-native law firms promise to automate commoditized tasks such as compliance, while still contributing a judgment layer from experienced attorneys. And one of the core law firm functions that’s not easily commoditized—M&A advising—has proven a huge driver for large firms in the first six months of 2026.
So perhaps the story is something like the following: A firm handles ten matters, each requiring 100 hours pre-AI. Total hours: 1,000. If an AI tool reduces the time per matter to 80 hours—a 20% time savings—each matter costs a client less and yields less revenue for the firm. But there’s more demand for these lower-priced matters, and the firm’s workload grows by 50%. Now, fifteen matters yield 1,200 billable hours. Assuming constant billing rates, the firm’s hours and revenue grow by 20%, even with the AI tool’s efficiencies. There’s no necessary inconsistency between AI-related cost efficiency and more billable hours—or greater revenue.
There are other possibilities. Strom notes that law firms may be “using the extra time to provide extra services or more thorough work for clients.” That is, the hypothetical firm handles the same ten matters for 120 hours per matter, and the legal work itself is more comprehensive or higher quality. This type of billing shift, of course, implicates the uncomfortable question of why the firm wasn’t doing its most comprehensive and highest quality work for each client pre-AI. Alternatively, matters may take 80 hours with AI but require 20 additional hours for a human lawyer to verify AI output. Essentially, clients and lawyers are fighting over a surplus that doesn’t yet exist but may emerge as technology improves. Finally, the market for legal services may not have reached equilibrium. As firms and clients adjust to the new technology, prices and quantities should move toward a new equilibrium, with possible openings for opportunism during the transition.
But if AI causes a rightward shift in the supply curve for legal services, and law firms’ response is to take on more matters, then whether the story has a happy ending depends significantly on the nature of those new matters. To the extent that the incremental matters reflect a previously unmet need for legal services, the profession is doing good work by billing more with AI. But if the additional matters are manufactured churning or pettifoggery, then the profession should have reason to police socially wasteful increases in legal work. The problem, of course, is that the matter-level data to address these questions aren’t readily available or easy to analyze systematically.
The profession really needs a reportable measure of legal output—work performed, its complexity, and its quality—to distinguish genuine efficiency gains from more pernicious alternatives. This measure also would bear on the overall allocation of AI-related gains among lawyers and clients. In the absence of such a measure (and none appears forthcoming), the profession will have to rely on anecdotes and informal impressions to trace AI’s effect on legal services.
Related TaxProf Blog coverage:
- Tax Advice and AI (Sept. 13, 2026)
- AI Slop Floods Legislative Counsel After Decades of Declining Congressional Capacity (Aug. 22, 2026)
- Vibe Coding for Lawyers and Law Students (Apr. 25, 2026)
- Bloomberg: Why Polsinelli Lawyers Won’t Get Billing Credit for AI Training (Apr. 11, 2026)
- Muller: AI Is Not Yet Disrupting the Robust Job Market for Law School Graduates (Apr. 5, 2026)
- Legalweek New York: AI and the Billable Hour (Mar. 21, 2026)
- The New Places of Legal Expertise in an AI World (Feb. 17, 2026)
- Will AI End the Billable Hour, Finally? (Dec. 27, 2025)
