The flagship essay
The AI-native practice.
There is a quiet assumption running through every conversation about AI and law: that the firms best positioned to go AI-native are the big ones. It is backwards.
The practices actually rebuilding themselves around AI are not the large firms. They are solo practitioners and small shops most people have never heard of, doing fixed-fee work with automated intake and same-day turnaround, quietly profitable, quietly growing. Two years ago this category barely had a name. It will not stay small for long.
This essay makes three arguments. First, that large firms are structurally incapable of going AI-native, no matter how much they spend on tools. Second, that solo and small practices hold the exact opposite position: everything that blocks the large firm is absent, and everything that matters is already in their hands. Third, that AI-native is not a vibe or a marketing label but a specific, describable operating model, one any small practice can install, piece by piece, without raising money or hiring engineers.
Why the big firms cannot do this
The economics of a large firm run on leverage: partners at the top, a pyramid of associates underneath doing the billable hours that make the partnership profitable. The associate is not just labour, the associate is the product's margin. A firm bills a junior out at a multiple of what it pays them, and the spread between those two numbers, multiplied across hundreds of associates and millions of hours, is what funds the offices, the recruiting pipeline and the partner draws.
Now look at what AI actually automates first: document review, first-draft contracts, research memos, due diligence summaries. That is not a random list. It is, almost exactly, the list of things junior staff spend their billable hours doing. Every hour AI removes from that stack is not efficiency from the firm's perspective. It is revenue the billing model has no way to recapture, because the model prices time, and AI's whole contribution is deleting time.
This is not a willpower problem or a technology problem. It is structural. A managing partner who fully automates the junior layer is not modernising the firm, they are dismantling the machine that pays for the building, the recruiting pipeline and their own partnership stake. Institutions do not vote against their own equity, and it is not reasonable to expect them to.
There is a second structural block that gets less attention: the partnership itself. A firm with hundreds of partners cannot rebuild its pricing, intake and delivery model by decision. It can only do it by consensus, and consensus among hundreds of people whose personal income depends on the current model produces exactly the outcome you would predict: pilots, committees, innovation labs and press releases. Motion without movement.
So the large firm does what large incumbents facing a structural threat have always done: it adopts the tools and protects the model. Document review software here, a contract analysis assistant there, a research copilot for the juniors. Pricing stays hourly. Intake still starts with a partner call and a week of back and forth. The client signs the same engagement letter and receives the same experience, delivered marginally faster by people using better software. That is AI-augmented, not AI-native, and it is the overwhelming majority of what is happening at the top of the market.
None of this means large firms disappear. Incumbents with strong client relationships and bet-the-company work survive structural shifts for a long time. It means something narrower: the AI-native operating model will not come from the top of the market. It is being built at the bottom, and the bottom has advantages nobody priced in.
Why solo and small practices can
A solo practitioner or a five-person shop has no pyramid to protect. There is no leverage model extracting margin from junior hours, because there often are no junior hours. Nobody's equity depends on preserving that layer, because it does not exist. The constraint on a small practice has never been headcount politics, it has been time. One person, or five people, and more work than the hours can hold.
That flips the entire incentive structure. For the large firm, every automated hour is lost revenue. For the small practitioner, every automated hour is recovered life: another matter taken on without hiring, an evening returned, a practice that scales without the practitioner scaling their exhaustion alongside it. The same technology that threatens the large firm's margin directly expands the small practice's capacity. Identical tool, opposite economics.
Three advantages that rarely get named
Decision speed
A solo practitioner can decide to publish fixed pricing on a Tuesday and have it live Wednesday. No committee, no pilot, no change-management consultant. The entire operating model is one person's decision, so the practice can iterate at the speed of one person's conviction.
Nothing to unlearn
A small shop does not carry twenty years of process documentation, a document management system with a decade of custom configuration, and staff trained on the old workflow. The absence of infrastructure, long treated as a small practice's weakness, is now its edge. Nothing to migrate, nothing to sunset, no sunk cost voting against the rebuild.
Client proximity
Solo and small-practice clients are individuals and small businesses, exactly the clients most punished by hourly billing's uncertainty and most rewarded by a fixed price and a fast turnaround. The client base at the bottom of the market is the one that rewards the AI-native model first.
And here is the quiet, underreported fact about the current wave: most of the practices actually building this way are not venture-backed startups burning investor money. They are bootstrapped, self-funded operations that built an operating model and are doing real volume, profitably, without ever announcing a funding round. The funding-announcement stories get the press coverage, which distorts the picture badly.
That is the actual opportunity. Not whether a professional can use a chatbot. The question is what a fully rebuilt practice looks like, concretely, and whether an ordinary solo practitioner can install it. The answer to the second half is yes, and the first half is describable in seven pieces.
What AI-native actually means
Seven levers any small practice can pull.
01
Pricing is transparent, not hourly
The client knows the cost before the work starts, published or generated as an instant quote. This sounds like a billing detail. It is the keystone of the whole model, because fixed pricing is only survivable when delivery is systematised, so committing to it forces everything else into place.
02
Intake is automated
No back and forth to get a matter started. Online forms, automated checks, engagement letters generated without a phone call. The work around the work runs itself, so substantive work can begin the same day the client shows up.
03
The practice meets clients where they already are
Existing channels and workflows, not a proprietary portal that sends a password-reset email once a quarter. Low-friction engagement is a design choice, and the practices building this way make it deliberately.
04
AI runs first, on every matter, systematically
The difference is systematic versus incidental. A practice where every matter, every time, gets an automated first pass before a person touches it, with the professional reviewing, correcting and taking responsibility for the output, is running a different operating model entirely.
05
The practice gets measurably better over time
Because every matter runs through the same system, the system produces data: turnaround times, accuracy rates, how often outputs need revision, where the exceptions cluster. That data feeds back into the system, and the practice improves as an institution rather than as a set of individuals.
06
Turnaround is fast
Often same-day, sometimes near-instant. Not because anyone is working harder, but because the work starts the moment it is submitted rather than when it reaches the top of somebody's inbox.
07
Technical capability sits inside the practice, not next to it
Engineering and data literacy are not an IT department bolted onto the side. Technical capability is treated as core to what the practice is, the same way professional judgement is.
The wave is not staying small
The old assumption was backwards.
Right now this category numbers a small, tracked cohort of firms. It will not stay small. A founding cohort proves the model works, the model gets documented, and once the playbook exists outside the founders' heads, the constraint on category growth stops being invention and becomes adoption. The number of practitioners running the model goes from tens to hundreds to thousands, not gradually but in a rush.
Strip away everything above and the practical conclusion is short. You do not need to raise money, most of the practices already doing this did not. You do not need a committee or permission. What you need is the operating model: how pricing gets set so it can be published, how intake gets automated so matters start themselves, how AI runs a systematic first pass on everything, how the whole system gets measured so the practice improves as an institution rather than as a person.
The old assumption was that AI-native belonged to whoever had the biggest budget. The actual answer is the opposite. It belongs to whoever has the least legacy structure to defend. Right now, that could be you.
The operating model, named
The six primitives underneath every AI-native practice.
This essay describes what an AI-native practice looks like from the outside. The MATTER Method describes what it is built from underneath, and MatterOS and LexOS exist to install it.
Building the operating model, not just writing about it.