The thesis
For three years the question in legal technology was whether AI can do legal work. That question is settled; it can, and in volume. The question now is who gets to run a practice built on that fact. The market's current answer is: only the firms born that way. That answer is wrong, and the wrongness of it is the largest unclaimed opportunity in legal services.
An AI-native firm is not a firm that uses AI. Every firm uses AI now, the way every firm uses email. The defining feature is architectural: in an AI-native firm, the machine does the first pass of the legal work and humans supervise it, rather than humans doing the work with machines assisting. Everything else flows from that single inversion. Pricing moves from hourly to fixed, because the cost of production no longer scales with time. The leverage pyramid of trainees and juniors collapses into what commentators call the obelisk: a small number of senior lawyers supervising machine output. The unit economics start to resemble software, which is why venture capital, which never had any use for law firms, suddenly cannot get enough of them.
The test is simple. Remove the AI and ask whether the firm could still deliver its service. If yes, it is a law firm with good tools. If no, it is AI-native.
Legal technology raised a record $2.4 billion in 2025, and the first quarter of 2026 alone produced $2.34 billion across 103 deals. Harvey closed a $200 million round at an $11 billion valuation; Legora raised $600 million at $5.6 billion. Manifest OS raised the largest Series A in legal tech history explicitly to scale "the world's first AI-native law firm model". Lawhive raised $60 million on revenue that grew sevenfold in a year. Eve passed a $1 billion valuation converting American plaintiff firms, one practice area in one country. And Moritz raised $9 million in four days on a model where AI does most of the work and named lawyers review, approve and carry the liability.
Read those numbers together and a pattern emerges. The tools layer is a capital war between two giants. The newbuild layer is proving the operating model, one firm at a time. And the conversion layer, the business of turning existing practices AI-native rather than replacing them, is almost empty.
The most under-reported fact in this story is that the regulators of both major common law markets have already said yes. The SRA authorised Garfield.Law in May 2025 as the first firm permitted to deliver legal services through AI, authorised LawFairy in February 2026, and joined the government's AI Growth Lab in June 2026. In the United States, Arizona abolished the rule against non-lawyer ownership and its Supreme Court approved Eudia Counsel, a venture-backed AI-augmented firm.
So the permission exists, on both sides of the Atlantic. And while the regulators open the front door, the courts have started closing the back one. The Upper Tribunal has held that using public AI tools on client matters can waive legal professional privilege, and survey data suggests a majority of UK lawyers are using AI without their firm's authorisation. Read those two facts together and supervision stops being best practice. Unsupervised AI use is now capable of destroying the single most valuable protection a lawyer offers a client, and much of the profession is doing it anyway.
Here is the assumption buried in every new AI firm's pitch deck: that AI-native firms have to be born, not made. But look closely at how the most celebrated newcomers actually deliver their work. Moritz's lawyers are co-counsel recruited from elite firms, reviewing machine output and carrying full professional liability. Manifest does not build firms at all; it partners with practising attorneys, more than a hundred of them. Strip away the branding and what remains, in both cases, is existing lawyers from the existing profession, doing AI-native work, because someone gave them the software and the governance to do it safely.
The "born, not made" story is marketing. The operating reality of the newbuilds is conversion; they have simply kept the conversion mechanism proprietary and charged a newbuild's valuation for it. And their model has a ceiling: a firm scaling at venture speed cannot afford to employ its lawyers, so growth is rationed by recruitment, one elite lawyer at a time. The benches are dozens; the profession is hundreds of thousands; and the clients, the brands and the insurance arrangements sit with firms, not individuals. The newbuild play is constrained by recruitment. The conversion play is constrained only by adoption, and adoption is the constraint that scales.
Any firm in England can licence world-class legal AI tomorrow with a credit card. What no firm can buy, anywhere, is a defensible answer to the regulator's question: who supervised this machine's work, how, and can you prove it? The SRA expects compliance officers to own the risk of new technology. The courts will strip privilege from unsupervised AI use. The American sanctions cases, where lawyers filed machine-invented citations, all reduce to the same diagnosis: AI errors are supervision failures, not technology failures.
The market's current answer to those requirements is a PDF checklist and a quarterly sign-off done by hand. That is not infrastructure; it is homework. NativeLaw is the infrastructure: a five-layer supervision process that takes AI-generated work through structured, recorded, professionally-owned review, ending in an immutable compliance record and a named lawyer's signature. The machine drafts; a named lawyer professes; the record proves it; the firm transforms without ceasing to be itself.
A necessary honesty here: the software alone does not finish the job. The supervision layer makes conversion survivable; the firm must still do the operational surgery on pricing, staffing and workflow. That is why NativeLaw has teamed with Lawyery, an SRA regulated firm that has spent four years living the method, so the firms that buy the infrastructure also get the playbook, and the clients who want the destination today can instruct a firm already operating there.
The profession's entire claim to exist is that someone answers for the work. Nothing about artificial intelligence changes that claim; it raises the stakes on it. The new firms believe the future of law belongs to practices born with machines inside them. We believe the future of law is the profession we already have, the nine thousand firms of England and Wales and the hundreds of thousands beyond, given the infrastructure to do what lawyers have always done: let the clerks draft, and sign only what they can stand behind. The clerks are silicon now. The signature is still human. NativeLaw is the bridge between those two facts.
AI errors are supervision failures, not technology failures.
From the thesis
This page condenses a longer working thesis; the principle it rests on, the manifesto and the full argument for professed, AI-native law, lives at professed.ai.
Get in touch See the platform