Five predictions for legal in five years

I spoke to a partner at the firm I trained at this week. His sentiment was that they’re excited to see how my generation transforms the industry.

He’s right, it will look dramatically different than what it looks like today. Also, telling.

Some opportunities have already emerged. But there’s also a lot of anxiety among lawyers at every level that doesn’t get the publicity it deserves. We’re too proud to say it in public, but private conversations I’ve had around the barrier to entry, job security and the new gatekeepers tell a starkly different story.

One thing I’d hold onto: the pace of change in AI, and in legal tech specifically, will plateau. That has been true of every technological advance and this one won’t be different. The question is what the industry looks like once it settles.

Here are five predictions. In no particular order.

1. The verification tax gets solved

LLMs are inherently probabilistic. Lawyers are trained to heat seek errors. Those two facts sit badly together.

AI has made legal work cheap to produce. It hasn’t solved the verification tax. While the output is capable of hallucination, a lawyer has to check all of it, which means the saving is smaller than most demos suggests.

For this to stop being a problem through model quality alone, foundation models would need to score 100% on every eval. That isn’t happening, and law is a bad candidate for it. Legal work is contextual and complex. It isn’t linear. It follows the commercial objectives of a business, and those objectives are often unstated and sometimes contradictory.

So we won’t solve this by building perfect agentic lawyers. We’ll solve it with process. Something that closes the gap between output and approval, and my guess is it looks a lot like old fashioned deterministic logic rather than another model.

It isn’t purely a tech problem either. It requires a shift in how lawyers think about putting their name to output they didn’t personally produce. That’s a professional question, not an engineering one, and it will move slower than the technology.

2. Tech becomes table stakes

This one is already happening.

The cost and expertise required to build a product with high adoption, high utility and reasonable pricing are already low, and falling. The legal tech that feels frontier today will be the entry point within a couple of years.

What will keep driving the wow moments is not the tooling. It’s how lawyers use it to solve real problems inside businesses, and what that finally reveals about their own value.

Outside of protecting their data, companies do not care about tech stacks. They care about legal outcomes. And they’ll keep expecting services to come with a cost saving attached, because they know the tech is doing some of the work.

For scale, the Future of Professionals Report 2026 from Thomson Reuters found 44% of professionals now use AI multiple times a day. I’d expect that to be close to 100% by the early 2030s. At that point it feels like Microsoft Office feels today.

3. The individual lawyer model emerges

Generalists and specialists with genuine experience advising companies will use that experience and go direct. We’re already seeing it at Correm.

Companies benefit because they no longer have to make a permanent hire until a much later stage. Having a very experienced counsel who manages your legal function stops being the luxury it is today. Most companies will have one, and most will be able to afford one.

Fractional lawyers with portfolio careers make that possible. It already seems obvious to the people and companies working this way. Within five years it’s becomes the norm.

Ultimately,  if legal had always been reasonably priced, companies would never have needed to vibe draft anything. AI doesn’t mean commercial people can finally do all the law they’ve been wanting to do. Nobody wanted that. It means there’s now a cheap alternative to the cost and speed inefficiencies of traditional law.

An AI native fractional lawyer closes that gap without asking a founder to become their own legal department.

4. Law firms remodel

Companies pay law firm rates for a lot of reasons. Prestige, insurance, complexity. That isn’t going to die and it shouldn’t.

What changes is that value gets priced on experience rather than time. If it takes a partner 20 minutes to do something with AI, the argument is that it took 30 years of experience to be able to do it in 20 minutes. In my view that’s fair.

So the cost doesn’t necessarily collapse. It gets reconfigured.

The real problem was never really the headline rate. It was where the model was perceptibly unfair, which is the traditional pyramid and the cost consequences that flow from it. Nobody wants to pay associate rates for work that AI now does very well.

The analogy I’d use here is healthcare. Most people handle minor things themselves, over the counter. For anything real they see a GP. If it gets serious they end up in a hospital with a surgeon.

AI is the over-the-counter self-medication. The fractional general counsel is the GP. The law firm is the hospital and the surgeon.

Most legal tech being built at the frontier is built for the hospital. Harvey and Legora are both, essentially, hospital software. That’s a real market and they’ll do well in it. Correm is building the GP practices, which is where most people actually spend most of their time.

5. Foundation models enter the game

It might look like tokenised legal. You subscribe to access legal in the way you subscribe to anything else.

Personally I don’t think you can make a codex or a Lovable for legal. It’s too complex and it depends on variables that are hard to know in advance. Lawyers understand that context patterns are not always linear, which is exactly the thing that resists being codified.

I think the success of this depends entirely on who it’s aimed at.

Aimed at consumers, I can see it really disrupting things like small claims. Volume, repeatable, and the alternative today is expensive and slow.

Aimed at non-legal corporates, it probably takes up more time than it saves. Again, corporates are not desperate to do more legal work. They wanted it to be quick and reliable, and handing them a subscription to do it themselves misreads the problem.

A lot has to happen first on risk underwriting, consumer trust and regulation, so this may take the longest to materialise. But it happens.

What actually changes

The biggest step change won’t be any single one of these. It’s the service delivery model.

The clients we speak to want quality legal services at a reasonable price for 90% of the work. They remain perfectly happy to pay for the 10% that genuinely requires a specialist firm.

That split has been obvious for years. What’s new is that it’s finally possible to build for it.

This list could have been double the length. These are the five I think stand out the most.

Prediction 4 names Harvey and Legora. Cut that sentence if you’d rather keep the healthcare analogy without the comparison.