oct 15, 2026

There Will Be No AI-Native Law Firms

A new neofirm is launching every week. Sure, law is a trillion-dollar industry with plenty of room. But there is something strange about an 'AI-native' business: it assumes the non-AI alternatives will continue to exist.

They won't.

There will be no AI-native law firms in the future. Because there will just be law firms.

The first wave of legal AI was software. Harvey and Legora sold to traditional law firms.

Now, those firms are building the technology themselves. Big Law is partnering with frontier labs to build in-house. Kirkland committed $500 million to its own AI platform. Cooley has a new business named Cooley AI.

The second wave is the neofirm. Rather than selling technology to lawyers, startups are building law firms around it.

The irony is that the same technology threatening the first wave's advantage could eventually threaten the second's too.

And Big Law already has the clients and the name. I've heard of VCs telling their portfolio companies to use established law firms for financing rounds instead of neofirms (even while investing in those same neofirms!) because when millions are on the line, 'the name matters'.

But not every legal matter is a financing round.

I am extremely bullish on neofirms. They get to build around AI from day one, while traditional firms have to adapt. While that head start won't last forever, it gives them an opportunity to pioneer something much bigger.

I think the third wave will be fully autonomous legal services: entire categories of legal work delivered without a lawyer approving every matter.

And I think that's a much bigger opportunity than simply making lawyers more productive.

The Review Ratio

AI can already review contracts, propose revisions and negotiate within a client's instructions. But a lawyer must still approve the work before it goes out.

The third wave removes that final checkpoint for specific workflows. Think an entire service: receiving a contract, negotiating its terms and returning it ready for signature.

I call it the review ratio: how much human review is required before AI-generated legal work can go out to a client. As the ratio approaches zero, lawyers might still handle exceptions or conduct audits, but they no longer need to approve every matter.

But law is a regulated profession; the firm remains responsible for its work regardless of whether a lawyer reviews it. I think we'll eventually see legal services delivered almost entirely by AI, with professional accountability, licensing and liability insurance still in place. I'm not going to predict which regulators will allow what or when but there are a few questions I find quite interesting:

Does human review actually make the work better?

We've always assumed so but as models improve, a lawyer could end up “correcting” something the AI got right in the first place. The benchmark here is whether the same workflow performs better with or without a lawyer reviewing it.

What happens when the AI gets something wrong?

A model could get 99% of contracts right and still leave a firm facing enormous losses. 1 faulty workflow could make the same mistake across 1000s of client matters before anyone catches it.

This is why loss modelling matters. You'd need to understand not just how often things go wrong, but how much they could cost, especially if the same error keeps happening.

What would it take to actually send that work out?

The firm is still responsible. So how much of the exposure can be insured? What evidence would an insurer need to underwrite it? How much autonomy would professional rules allow?

The questions above are a simple outline of the complex and exciting landscape we will have to navigate. And obviously, the answers will be different for different workflows. But for some, I think we will see the third wave coming.

That changes what kind of business a law firm can be.

The Opportunity for Neofirms

To be honest, I used to be a little skeptical of the VC case for neofirms. Yes, AI makes lawyers dramatically more productive and revenue per lawyer could be enormous ... but you're still selling a service. Even with much better margins, can you realistically build a billion-dollar business in a few years rather than the decades it took Big Law?

And I know the whole neofirm pitch is that AI breaks the relationship between revenue and lawyer headcount. I'm just not convinced that making lawyers 10x or even 100x more productive is enough.

What changed my mind is the possibility of fully autonomous legal workflows: repeatable legal services (think: negotiating routine NDAs or recovering undisputed debts) delivered end-to-end.

Instead of building a more efficient law firm, you're effectively selling software that delivers a legal service, with a professional license and liability attached, for some workflows.

I think this is where neofirms have a real advantage. Big Law can buy the same models but rebuilding an entire institution around autonomous delivery is a different problem.

That's the opportunity a technological head start gives neofirms: the chance to build a completely different kind of law firm.

For the first time, I can see a plausible path to a billion-dollar neofirm on VC timelines.

So there are really two reasons I don't think AI-native law firms will exist as a distinct category for very long:

1. Big Law will eventually become AI-native
2. The neofirms that succeed might stop operating like traditional law firms altogether

Which is kind of funny, because we've come full circle.

The first wave of legal AI sold software to law firms. The current wave is building law firms around software.

The third wave might be law firms that scale like software.

first published on x