Published: September 4, 2026 | Reading Time: ~11 minutes | Channel: Business
Three of the biggest legal clients on planet Earth just looked at their outside counsel invoices, looked at the AI tools everyone's been bragging about for three years, and asked the only question that matters: so where's our discount?
Goldman Sachs, Morgan Stanley and Citigroup are pressing elite law firms to pass on the savings from AI-assisted legal work, according to a Financial Times report by Kaye Wiggins and Joshua Franklin published September 1, 2026.¹ Citi's global head of legal, Adam Meshel, put it in writing: if the number of hours on a matter has come down because of AI, "our expectation is for costs to come down significantly per transaction."² Morgan Stanley's general counsel Eric Grossman went further and called Big Law's traditional associate-heavy model "extraordinarily unstable," warning that AI could amount to "a fundamental altering of the revenue foundation for these mega firms."²
Here's the number that explains everything: average associate billing rates at top US law firms have climbed 33% since 2023, to $798 an hour this year. Partner rates rose 29% over the same period.² That happened during the exact window in which 94% of lawyers started using AI for legal work.³
Read those two facts together and you understand why the banks are done being polite. The most productive technology in the history of professional services arrived, adoption went vertical — and prices went up. Not sideways. Up by a third.
This isn't a legal industry story. It's the opening shot in the repricing of every profession that bills by the hour — including yours.
For three years, the AI story in professional services has been told from the seller's chair. Firms bought tools, published glossy "AI transformation" reports, and told anyone within PowerPoint range that artificial intelligence would revolutionize legal practice. The clients listened. Then the clients did math.
The math goes like this: contract review, document review, due diligence, research memos, discovery — the bread and butter of a first- and second-year associate's billable life — is exactly the text-heavy, pattern-based work AI tools were built to chew through. A junior lawyer once spent three weeks drowning in document review and billed every hour of it. Now a machine does the first pass in minutes, with a lawyer checking the result. The client looks at the invoice and asks the obvious question: why does the bill still look like nothing changed?
What's new this month is who's asking. General counsel have grumbled about this quietly for a couple of years. But Goldman, Morgan Stanley and Citi have the buying power to make the question expensive to ignore. And each bank is applying pressure differently:
Notice what's absent from that list: any suggestion that the banks want bargain-bin legal advice. Morgan Stanley's position, per the FT, is that it will still pay serious money for judgment, expertise, and the legal nerve required when a multibillion-dollar deal is wobbling at midnight.² What's being repriced isn't expertise. It's time — specifically, the fiction that an hour of junior-associate document review is worth the same in 2026 as it was in 2021.

Let's put the whole battlefield on one table. Every figure below is scraped and verified — sources numbered at the bottom.
The rate explosion (the client's grievance):
| Metric | 2023 | 2026 | Change |
|---|---|---|---|
| Avg. associate rate, top US firms | ~$600/hr (implied) | $798/hr | +33% ² |
| Partner rates, top US firms | — | — | +29% ² |
| First-year, Sullivan & Cromwell, FTX bankruptcy (2023 benchmark) | — | $1,310/hr billed | court filing ⁴ |
| Lawyers billing near $960/hr (2021 grads, 2023) | $960/hr | — | Bloomberg Law ⁴ |
The AI adoption curve (the firm's defense — and the client's ammunition):
| LexisNexis 2026 survey (n=543) | Share |
|---|---|
| Lawyers using AI for legal work | 94% ³ |
| Using it at least weekly | 74% ³ |
| Using it daily | 34% ³ |
| Dependent on AI to do their job | 10% ³ |
| Believe firms will change billing because of AI | 55% (vs 40% in Jan 2025) ³ |
| Using AI for legal research | 69% ³ |
| Document summarisation | 62% ³ |
| Drafting / document review | 53% each ³ |
The economics fight (both sides have numbers):
| Metric | Figure | Source |
|---|---|---|
| Law firm revenue growth, first 9M 2025 | +11.3% | Citi Law Firm Group / Hildebrandt ⁴ |
| Demand growth, same period | +1.9% | ⁴ |
| Expense growth, same period | +9.1% | ⁴ |
| Headcount growth | +2.9% | ⁴ |
| Demand growth, H1 2026 | +4.2% | Bloomberg Law ⁴ |
| Large firms saying AI already affected their pricing | ~half | Citi law-firm-group survey ² |
| Corporate legal depts saying AI will significantly reduce billable-hour reliance (2026 survey) | 62% | Wolters Kluwer ⁴ |
| Law firms saying the same | 57% | Wolters Kluwer ⁴ |
| Same question, 2024 survey (depts / firms, "affect prevalence") | 67% / 55% | Wolters Kluwer ⁴ |
And one more from the adoption data that should make any law firm managing partner sweat: the share of legal organizations with AI embedded in strategy and operations jumped from 17% in January 2026 to 29% today, and 46% of lawyers now say their career would be negatively affected if their organization failed to fully embrace AI — up from 28% a year ago.³ The professionals themselves have already surrendered to the future. Only the invoice is holding out.
The law firms' position, stated plainly: we bought the tools, we trained the lawyers, we carry the licensing bills and the hallucination risk — the efficiency gain is our margin, not your rebate. And to be fair, their recent financials give that argument some cover. Revenue up 11.3% while expenses rose 9.1% is a healthy spread, not a crisis.⁴
It's also an argument that collapses under its own arithmetic — for three reasons.
First, the productivity gain isn't the firm's to claim, because the client funded it. In the billable-hour model, the client pays for every hour of every learning curve. Decades of "know-how" inside elite firms — the templates, the playbooks, the precedent libraries — were effectively R&D paid for by clients at $798 an hour. AI didn't create the savings from nothing; it industrialized knowledge the clients already financed. Handing the client a portion of the savings isn't charity. It's a refund on duplicated tuition.
Second, the rate increases were the tell. A 33% associate rate hike during the largest productivity shock in professional services history isn't a pricing strategy — it's a liquidity event. Firms monetized scarcity (talent wars, senior-heavy staffing) while the scarcity itself was being automated away. That works until the biggest buyers in the market do what Goldman just did: demand the itemized AI savings and use them as negotiating leverage.⁴ The banks aren't inventing a new dynamic. They're noticing that the old magic trick — rooms full of junior lawyers billing time — is now visible from the balcony.
Third, the countervailing data is a head fake. Yes, demand grew 4.2% in the first half of 2026, per Bloomberg Law.⁴ But demand growth measures work volume, not willingness to pay yesterday's prices per unit of work. And the market's own expectations have moved: 55% of lawyers now believe AI will change how firms bill, up from 40% in January 2025.³ When half the people inside the cathedral expect the pricing model to crack, "nothing has broken yet" is not a rebuttal — it's a countdown.
The second-order effect nobody's talking about: this repricing won't stop at law. The FT reported the same day that companies are already pushing consulting firms to lower fees as AI reduces the need for outside work on IT projects and system integration.⁴ Procurement teams are pattern-matchers. Once they learn the "quantify your AI savings" question works on a $798-an-hour law firm, they'll run the same play on auditors, consultants, agencies, and every vendor whose invoice still treats time as the product. The billable hour isn't just a legal industry problem anymore. It's the template for the 2026-2030 procurement cycle.
Whether you run a company that buys professional services or one that sells them, here's the playbook — this week, not someday.
If you're the buyer:
If you're the seller — and if you run any services business, you are:

And the standard disclaimer, because this is analysis, not advice: none of this is investment or legal guidance. If you re-price your services business or renegotiate your vendor contracts based on a blog post, do it with your own advisors — ideally ones who've quantified their AI savings.
For three years, professional services firms told clients AI would change everything — and quietly kept the savings. This month, the three clients with the most buying power said "great, itemize it." The billable hour was never killed by technology; it will be killed by procurement. Ask the Goldman question before someone asks it of you.
All claims verified against Gold-tier (LexisNexis primary research release via GlobeNewswire) and Silver-tier (Financial Times via four independent syndications, Bloomberg Law data as cited) sources. Each cited URL was scraped and confirmed accessible on September 4, 2026. Discarded during verification: Seeking Alpha and Luxembourg Times (403 blocked). The original FT article is paywalled; every FT-derived fact is double-sourced through at least two independent syndications.
The clients heard you. They did the math. Send the invoice accordingly. 🎯