The AI-native law firm won't look like a law firm

When firms imagine their AI-native competitor, they picture a law firm that’s good at AI. A familiar shape - partners, associates, practice groups, a brand - just with better tools bolted on. They benchmark themselves against that imagined rival and conclude they’re not far behind.

I think they’re benchmarking against the wrong thing entirely. The business that does real damage to traditional firms won’t be a law firm that’s good at AI. It won’t look like a law firm at all - and the reason is a pattern that has played out in every industry software has reshaped.

The disruptor never looks like the incumbent

The thing that eats a category almost never looks like the category. The business that hollowed out travel agencies wasn’t a better travel agency. The one that took the taxi industry’s lunch wasn’t a taxi company. In each case the incumbents kept scanning the horizon for a better version of themselves, and the threat arrived from an angle that didn’t register because it didn’t share their shape.

There’s a structural reason for this. An incumbent is organised around the old way of delivering value, and that organisation is load-bearing - it’s how they make money, how they’re regulated, how their people are trained and paid. They can bolt new technology onto the old structure, but they can’t easily let the technology reorganise the structure itself, because the structure is the business. A challenger born after the technology has no such constraint. It organises around the new way from day one. And “organised around the new way from day one” produces something that looks alien to the incumbent - which is exactly why they don’t see it coming.

What the legal version actually looks like

So what does a business organised from scratch around productised, AI-delivered legal work look like? Not like a firm. A few features I’d expect:

Its core asset is software, not people. A traditional firm’s balance sheet is its lawyers - their judgement, their relationships, their hours. The AI-native legal business’s core asset is a system that delivers a category of legal outcome, with people in supervising and edge-case roles around it. The headcount-to-revenue ratio is unrecognisable. It can serve ten times the volume with a tenth of the lawyers, because the lawyers aren’t doing the work - they’re standing behind the system that does.

It sells outcomes and products, not time. There’s no billable hour because there’s no leverage pyramid to bill. There’s a product, with a price - per output, per seat, per outcome. The economics are software economics: high fixed cost to build the system, near-zero marginal cost to serve the next customer. That’s a completely different financial animal from a firm, and it can price in ways a firm structurally can’t match on the work it targets.

It’s narrow on purpose. It doesn’t do “law.” It does one category - one type of work, deeply productised - and owns it. Breadth is the incumbent’s strength and the challenger’s trap. The AI-native business wins by going a mile deep on a category an inch wide, taking it completely, and only then expanding. By the time it’s visible across the whole market, it already owns several categories outright.

It treats regulation as a design constraint, not a moat. Incumbents quietly rely on regulation to keep challengers out. The smart AI-native business doesn’t fight the regulation or hide from it - it designs around it from the start, builds the accountability and supervision in as a feature, and turns “we’re built to be compliant by design” into part of the product. Regulation slows it down. It doesn’t stop it, because the challenger treats it as a spec rather than a wall.

Why incumbents will misjudge the timing

The cruel part is that for a long time, the AI-native challenger will look unthreatening to the incumbent - small, narrow, operating in some unglamorous corner the firm doesn’t care about. The firm will note it, file it under “not real competition,” and carry on benchmarking against the imagined better-version-of-itself.

And the challenger will be fine with that, because the corner is the beachhead. It takes one narrow category completely, builds the muscle for productising legal work, and then moves to the next category with everything it learned. Each category falls faster than the last. By the time the cumulative effect is visible to the incumbents - by the time it shows up as lost work across several practice areas at once - the challenger has compounded for years and the gap is structural, not catch-up-able.

What I’d do if I ran a firm

I would stop benchmarking against a better version of my own firm, because that comparison is comfortable and useless. I’d benchmark against the thing that doesn’t look like me: the narrow, software-cored, outcome-pricing, regulation-by-design business that could take one of my high-volume categories completely.

And then I’d ask the genuinely hard question - the one most firms avoid because the answer is threatening: in the categories where that challenger could win, should I be building the AI-native business myself, on my own terms, before someone who doesn’t look like me does it to me? The firms that survive the next decade won’t be the ones that got good at AI. They’ll be the ones that were willing to grow a thing inside themselves that didn’t look like a law firm - and let it eat the parts of the old firm that were always going to be eaten anyway.