AI-native law firms are regulated practices designed around AI from the outset rather than ordinary practices that bought a few tools. The Law Society Gazette reported on 7 August 2026 that a directory kept by Matt Pollins now carries 58 listings. The United States holds 59 per cent and the United Kingdom 16 per cent, which puts around nine of them here. That is a small population set against the profession, and no longer one you can treat as an experiment happening somewhere else.
What the model looks like in practice
Pollins draws the line between a firm designed around AI and a firm augmented by it, and the distinction shows in how a firm is built rather than in how it describes itself. An AI-native firm builds its pricing, intake, delivery and team structure on the assumption that software does the first pass. Fees are fixed or charged by subscription, intake runs online and returns a quote without anyone picking up the phone, and AI produces the work before a lawyer reviews it and signs it off. That last step is not optional courtesy. Regulation here requires a named person to take responsibility for the advice, whatever the automation does upstream.
The pricing reaches you before the competition does
The part of this that touches your practice first is not the technology. It is the quote. A client with a debt to recover or a contract to review now has the option of an online form that returns a fixed price in minutes. Your own process runs to a call, a scoping exercise and a client care letter that lands days later. You do not need to lose the work to feel the effect, because that quote sets what the client thinks the job is worth before they speak to you. We looked at the same pressure from the buyer's side in what clients now expect AI to do to your fees, and the AI-native firm is where it gets a price attached.
Speed of adaptation is the other advantage worth naming. A firm whose intake, pricing and workflow were all built in the last two years changes any of them in weeks. A firm running a practice management system chosen in 2015 does not.
The model reaches family work and conveyancing
A British instance arrived this month. The divorce and separation service amicable, founded in 2015 and operating outside regulation, has applied to the SRA for authorisation to run a separate law firm in England and Wales, an application Legal IT Insider reported on 13 August 2026. The new firm will take family work its couples-based model cannot reach, including where one partner will engage and the other will not. It will also handle transfers of equity and the remortgages that follow for separating couples. AI will draft, triage and research, with solicitors reviewing the advice and the court documents.
Read that as a direction of travel rather than as one competitor, because two staples of high street work are the target and the business aiming at them already holds the client relationship.
Where the model runs into professional obligation
The Gazette also reported an argument running alongside the headcount, and it matters more. Crosby, an AI-native firm working on commercial contract negotiation, wants to buy professional liability insurance for its AI agents so they can do autonomous legal work. Its chief executive Ryan Daniels wrote that lawyers review every output today and that agents have improved enough for that to stop being necessary. Jake Sendar, who founded the autonomous transactions system Walton, answered that insured but unsupervised output reduces an AI-native firm to selling software with insurance attached, and that the appeal of the model was machine efficiency combined with professional accountability.
For a firm regulated in England and Wales the argument settles itself. Insurance pays for loss after the event and transfers nothing about who owes the client a duty, so the solicitor who put their name to the work still answers for it. What the exchange tells you is where the commercial pressure points, and the firms building this model are already asking how much review they can strip out. Your own answer belongs in writing, which is the ground covered in AI and your firm's professional indemnity cover.
Two pieces of work follow from this, and neither takes long. Take one standard matter you handle every week, price it as a fixed fee on your real costs, and see whether the number survives what an automated quote would offer for the same thing. Then write down, for each type of work where you use AI, the point at which a named person reads the output and takes responsibility for it. The first tells you whether your pricing has a problem coming. The second is what your regulator will ask for, and what an AI-native competitor has already written down.
Joanna Goodman set out the count, the directory and the insurance argument for the Law Society Gazette on 7 August 2026 in AI natives and rogue agents, and the Gazette reported the amicable application on 14 August. Both are free to read.
If you want to know how your own pricing and sign-off would stand against a firm built this way, that comparison is the sort of work we do before a client changes anything: talk it through with us.