A letter from Max Junestrand, CEO and co-founder of Legora, which raised $550 million at an $8.8 billion valuation this week.
A large U.S. law firm wanted to switch to Legora but needed one specific task solved, and we couldn’t do it yet. One of our legal engineers, a former lawyer, flew from New York to Stockholm on a Saturday and spent a week and a half with our engineers until he built it. That is what we mean by partner.
When I started Legora with my co-founders, we made a rule: every new customer had to make the company stronger. It goes back to our partnership with Mannheimer Swartling, one of the largest firms in the Nordics, and our first big customer. We built the product alongside their lawyers, setting up Legora’s first office in one of their conference rooms, and we have built it with our customers ever since.
Three years on we have more than 2,400 customers, and the rule hasn’t changed. Each one made the product better and the economics healthier. That way of working is also why we raised a $550 million Series E, led by Atreides Management. I’ll make the full case at the end, but here is the short version of why I’m still optimistic: legal has always had more work than lawyers to do it, and AI hasn’t changed that.
The bet we made on models
We build on the best frontier models, and don’t train our own general legal model. That was a real debate when we started in 2023. The models were new, and plenty of people said the way to win in legal AI was to fine-tune your own. We didn’t have the money for it, and fine-tuning didn’t seem to work at our scale. Training a new generation of model costs billions, and most of what lawyers needed wasn’t in the model anyway. It was in everything around it: security, privacy, parsing, working with large sets of documents. So we bet on every frontier model and put our effort into making them work for lawyers. Since then a better, cheaper model has arrived every few months. We build boats, not houses: when the models rise, everything we’ve built floats higher. Legora routes each task to the model that does it best, and we only post-train for narrow jobs where frontier models fall short.
That bet also made us efficient. Every new model goes through our own evaluations before it touches customer work, so we know exactly where it’s strong. Customers get better performance at lower cost.
We build Legora together with our customers, and we wanted our pricing to reflect that. We get paid when customers get value. We were among the first legal AI companies to offer consumption-based pricing, so what a customer pays rises and falls with the work Legora does for them. We only grow with a customer when Legora is earning its place, which keeps our interests pointed the same way. It’s the rule we started with, written into how we charge.
To keep earning that place, we do more than ship software. Building technology is hard. Changing how a law firm works is harder. So when we launched, we didn’t hand lawyers a product and wish them luck. We had former lawyers ready to sit with them and show how it fit their practice area. We call them legal engineers. We now employ more than 150, and they work inside customer teams on their hardest matters.
We now have more lawyers than software engineers. And daily active users now spend more than two hours a day on Legora.
None of this happens in a vacuum. Legal AI is now a crowded, well-funded market, which tells you how much law firms and investors expect it to change. When every serious company builds on the same frontier models, the models stop being the difference. The difference is what sits around them: how well a system knows a firm’s work, and whether a partner will trust it with a deal.
The hard choices
Early on, we didn’t sell Legora until we knew lawyers could rely on it, even though we could have started sooner. We chose to build on other people’s models rather than our own, and decided to charge in a way that only pays off if customers keep getting value. We win on the product and the service, so we don’t chase lower prices to close a deal.
They have paid off. Our annual recurring revenue now exceeds $200 million. We’d make the same trades again.
What the money builds
Legal work is roughly a trillion-dollar industry, and still today only about 4% of it goes on software. Agents can change that, and getting them into lawyers’ hands is half the job. We meet lawyers where they already work, in Microsoft Word or in person at their offices.
That system is the Legora aOS, our agentic operating system for legal work. A law firm already runs on one: its playbooks, how juniors are trained, who signs off on what. The trouble is that it lives in people’s heads and across too many disconnected tools, so every piece of work depends on someone coordinating it by hand. The aOS brings it into one system that sits between the models and the firm. Agents can take a whole task from start to finish, such as diligence on a deal or discovery in a dispute. Lawyers step in where judgment is needed.
What a firm teaches the system stays the firm’s. The models underneath will keep changing, and the aOS is built so that each new one makes it better.
Where we go from here
Legal has always had more work than lawyers to do it. AI hasn’t changed that. In fact, I’m convinced that AI will grow the pie. If you could review every contract, you would. Many of the firms we work with are hiring more lawyers, because there’s more work to go around. And our users tell us they love being a lawyer again. Some say they would never have left the profession if they’d had this technology when they practiced.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
