The Fixed Fee Has Arrived in Legal. Watch Where the Savings Go.

In March, Harvey raised $200 million at an $11 billion valuation. Legora raised $550 million at $5.55 billion.¹ Together, two companies that did not exist five years ago are now worth more than most publicly listed legal technology incumbents combined. Neither is profitable. Both sell into a licensed profession that has, for most of its modern history, been structurally resistant to productivity-driven pricing.

I want to take the valuations seriously, because I think they're telling us something specific, and I don't think that something is "investors are excited about legal software." At $11 billion, Harvey is not priced against the roughly $40 billion legal software market. It is priced against the roughly $1 trillion global legal services market, a market twenty-five times larger.¹ The bet is not that Harvey will sell more tools to law firms. The bet is that Harvey, and products like it, will compress what law firms can charge for the work itself.

The bet is that Harvey, and products like it, will compress what law firms can charge for the work itself.

And the bet is starting to look right. At the Stanford CodeX Future of Law conference on 16 April, Harvey's CEO Winston Weinberg said client expectations have "already shifted."² The specific shift he was pointing at is this: general counsels at large corporates are, for the first time at scale, signing fixed-fee alternative fee arrangements into their budgets, flat fees for defined matters, rather than the hourly billing that has structured outside legal work for half a century.¹ This is the pricing signal I've been watching for. AFAs have been discussed in the legal trade press for more than a decade. Law firms have resisted them for exactly as long, because a fixed fee transfers productivity risk from the client to the provider. AI productivity flips that calculus. A firm that can produce a first-draft contract in minutes instead of days can absorb fixed-fee risk comfortably. A firm that can't, can't.

That's the surface story. The story underneath it, the one I want to spend this piece on, is about who captures the efficiency gain, who absorbs the cost, and what general counsels will start asking for next. Because the people who gain from this transition and the people who pay for it are not, in any sense I can find, the same people.

What GCs are actually buying now

Talk to anyone selling into corporate legal departments right now, or read the pitch decks, and a specific picture of the new GC emerges. They are no longer buying "a relationship with a firm." They are buying measurable throughput at a defined price, with audit trails.

The consequence is visible in what GCs are starting to require from outside counsel, in tenders and in scoping conversations. I'd flag five shifts, all of which follow from the same underlying logic:

First, disclosure of AI tooling in the matter. GCs are asking which AI systems a firm uses, on which tasks, with what human supervision. This is partly ethical hygiene, several US state bars and the UK Solicitors Regulation Authority require human supervision of AI outputs³, and partly economic. A GC who knows Harvey did the first-pass drafting has a defensible basis for negotiating the associate time out of the bill.

Second, fixed fees on defined scopes, with carve-outs. The AFA itself is not new; what's new is that GCs now have a credible benchmark for what an AI-assisted firm should be able to produce a contract or a diligence memo in. The benchmark isn't precise, but it doesn't need to be. It just needs to be credible enough to make the hourly bill look indefensible.

Third, panel consolidation toward firms that have visibly invested in AI infrastructure. Firms that can show Harvey or Legora deployments, internal tooling, and AI governance policies are moving up panels. Firms that can't are moving down. Thomson Reuters found in 2024 that only 23% of law firms had formal AI governance policies.³ That number is going to matter a great deal, very quickly, as a selection criterion.

Fourth, blended "AI plus talent" offerings as a live alternative to the traditional firm. Axiom Law's 8 April partnership with Harvey, bundling Harvey's tooling with Axiom's flexible legal talent, sold directly to in-house teams, is the first major product of this kind.⁴ It reframes the question a GC asks. It's no longer "which firm should I send this to"; it's "should I send this to a firm at all, or to an AI-plus-contract-lawyer bundle that costs a third as much?" The alternative legal services provider market was already around $28.5 billion in 2023.³ Bundled AI makes the ALSP pitch materially sharper.

Fifth, and this is the one I think is most underpriced in the current coverage, GCs are starting to ask for outcome data. Error rates on AI-drafted clauses. Time-to-completion benchmarks. Comparative performance between firms on matched matters. The move from "trust my firm" to "show me the numbers" is the biggest cultural shift here, and it is downstream of the fact that, for the first time, the numbers exist.

If you are a managing partner reading this, the implication is not subtle. The firms that win the next procurement cycle will be the ones that can answer these five questions with receipts. The firms that can't will lose panel spots to firms that can, or to Axiom-style bundles, and they will lose them faster than the historical pace of legal industry change would suggest is possible.

Who's paying

This is where I want to be careful, because the easy version of this essay, "AI is coming for the lawyers", is both wrong and right in ways that matter.

It's wrong at the top. Partners at major firms bill at $1,000 to $2,000 an hour or more, and their work, judgment calls, client relationships, bet-the-company litigation strategy, is not the work Harvey does well.³ Partners will capture a meaningful share of the efficiency gain, at least in the near term, because fixed-fee work completed faster at the same price is simply more profitable per partner-hour. This is the point that partner-track lawyers at the top firms make to themselves in the lift, and they are not wrong to make it.

It's right at the bottom. Associates at major firms bill at $400 to $700 an hour, and a very large share of what they do, document review, first-draft contracts, discovery summarisation, regulatory research, is work Harvey does in seconds.³ Junior associate hiring at large US firms fell 6% in 2024.⁵ That is not yet a cliff. But set it against the fact that US law school enrolment in 2023–24 hit 114,520, the highest since 2010,⁵ and the structural shape of the next five years becomes uncomfortable to look at directly. More entrants, fewer entry-level roles, and the specific work that entry-level roles exist to do is the work most exposed to compression.

The people being priced out of the profession are not being consulted about the pricing. The AFA deals are being signed by GCs and managing partners. The law students borrowing $200,000 to train for a profession whose entry rungs are being sawn off are not at the table. Neither are the paralegals, whose work overlaps even more heavily with what Harvey does than associate work does, and who have less professional mobility when the work contracts.

I've seen the counter-argument, and I want to take it seriously. The argument is that legal work is demand-elastic: make it cheaper and there will be more of it, because there are vast quantities of legal need that currently go unmet because hiring a lawyer is too expensive. Small businesses that don't paper their contracts properly, individuals who don't contest unfair dismissals, tenants who don't challenge bad leases. This is a real argument, and it has history on its side, legal services have expanded faster than most forecasts predicted for thirty years. Thomson Reuters' 2025 survey of law firm leaders found that "client demand and value," not AI efficiency, was cited as the primary driver of AFA adoption.³ That suggests AI is accelerating a trend, not creating one.

But I don't think demand elasticity, even if it's real, solves the distributional problem. An expanded market for legal services means more work, not necessarily more traditional-track jobs. If the expansion comes through Axiom-style bundles, flexible contract lawyers plus AI, sold at a fraction of firm prices, then the work exists but the stable, career-progression-bearing job that used to do that work does not. The efficiency gain has been captured somewhere. It just hasn't been captured by the person doing the work.

Which brings me to Axiom.

The bundle problem

The Axiom-Harvey partnership is, on its merits, an interesting product. Corporate legal departments genuinely struggle with the peaks and troughs of legal demand; a flexible pool of contract lawyers augmented with AI tooling is a credible way to smooth that. For the GC buying it, it is probably a better product than what existed before. For some contract lawyers, it is probably better-paid and more interesting work than they would otherwise have. I'm not trying to argue the product is bad.

I am trying to argue that the structure of the product is worth looking at. Axiom's contract lawyers are not firm associates. They are not on partner track. They do not accumulate the institutional equity, the client relationships, the expertise in a firm's way of doing things, the eventual partnership, that has historically been the compensating promise for the brutal hours of early-career law. They are, in the parlance, fungible.

When you bundle AI with fungible human labour and sell it to a sophisticated buyer, the buyer captures most of the efficiency gain, because the human labour has limited bargaining power and the AI has none. A Big Law associate, grim as their hours can be, is inside an institution that has every incentive to retain them, train them, and eventually promote them. An Axiom contract lawyer is inside a market. The market's incentive is to pay them as little as the market clears at.

This is the distributional question the valuation surge is a bet on, even if the investors haven't framed it that way. If legal work shifts at scale from firm-employed associates to AI-plus-flex-talent bundles, then a very large share of the $1 trillion market's gross margin moves from being captured by human legal labour (as wages, partner draws, associate bonuses) to being captured by the owners of the tooling and the platforms. That is what a $16 billion combined valuation is pricing.

The thing GCs should see, and probably won't

There's one more consequence worth naming, because I don't see it in the trade coverage and I think it's coming.

General counsels are, right now, the clear winners of this shift. They gain leverage over outside firms. They gain data on legal production. They gain flexible alternatives. They look, from where I'm sitting, like the most structurally empowered they have been in decades relative to the firms that serve them.

But the same tooling that lets a GC credibly tell outside counsel "this matter should cost a third of what you're quoting" will, within a cycle or two, let a CFO credibly tell a GC "your in-house team should be half its current size." The logic that's being used to compress outside legal spend doesn't stop at the law firm door. The same productivity evidence, the same benchmarking, the same AI-plus-flex-talent alternatives. The efficiency dividend rolls uphill until someone powerful enough to stop it stops it, and in most corporates that person is not the general counsel.

I'd like GCs to notice this, because they are currently the most effective advocates inside large corporates for a humane transition. They're also going to be in the position to shape the terms of their own compression, if they start doing it now, and to get run over by it if they don't.

What to watch

Three things, in order of how quickly I expect them to matter.

First, the first major malpractice or disciplinary case involving AI-drafted legal work. Bar ethics rules require competent supervision of AI outputs; what "competent supervision" means in practice is not yet settled. The first hard case will shape procurement requirements for years, and will either accelerate the shift toward firms with real AI governance or slow it sharply.

Second, the panel decisions at FTSE 100 and Fortune 500 legal departments in the next two review cycles. These are the decisions that will reveal whether the AI-disclosure and outcome-data demands I described above are a real selection criterion or a performative one.

Third, the associate hiring numbers for the 2026 and 2027 classes. Six percent down in 2024 is a signal. Another two years of declines against record enrolment is a structural problem, and at some point it becomes the story, not the valuations, not the deals, but the generation that trained for a profession that repriced before they arrived in it.

I don't think this is going to go well for them. I'd like to be wrong.

---

Footnotes

Footnotes

1. "Legal Tech Valuations Surge In 2026 Because of AI," Broadband Breakfast, April 2026. https://broadbandbreakfast.com/legal-tech-valuations-surge-in-2026-because-of-ai/ ↩ ↩² ↩³

2. Winston Weinberg, remarks at Stanford CodeX Future of Law Conference, 16 April 2026, reported in "Legalweek 2026: Five Takeaways on AI and the Future of Legal," Harvey.ai Blog, 2026. https://www.harvey.ai/blog/legalweek-2026-takeaways ↩

3. Thomson Reuters Institute, 2024 Report on the State of the Legal Market and Future of Professionals Report 2024, 2024. https://www.thomsonreuters.com/en/reports/future-of-professionals.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

4. "Axiom Expands AI Tech+Talent Portfolio with Harvey for In-House Legal Teams," Axiom Law press release, 8 April 2026. https://www.axiomlaw.com/resources/press-releases/axiom-expands-ai-techtalent-portfolio-with-harvey-for-in-house-legal-teams ↩

5. NALP, "Associate Hiring Trends 2024"; American Bar Association, "ABA Profile of the Legal Profession 2024." https://www.americanbar.org/groups/legal_education/resources/statistics/ ↩ ↩²

¶