Loading page…
An AI workspace for lawyers built by founders with no legal background — a document-by-query review grid plus a Word add-in that runs a firm's negotiation playbook against a contract and marks it up
tens of thousands of lawyers using it daily · $80M Series B led by Iconiq and General Catalyst, with Y Combinator, Benchmark and Redpoint participating · headcount 10 → 100 in roughly a year · $675M valuation appears only in the video title, not in the interview
"We don't know exactly where the future's going, but neither do you. So let's work together to make sure that we're both winners in whatever happens." That sentence is how three non-lawyers sold AI to law firms that had already been burned.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Legora went through Y Combinator's Winter 2024 batch. Thirteen months later the founder says he feels five years older, and the company has gone from ten people to a hundred. The founding observation was mundane and specific: a co-founder's friend, a lawyer, spent four months of one summer doing nothing but summarising court cases for a big law firm. The team had been playing with AI and legal work before ChatGPT, using early BERT models from Google, which were decent in English and "horrendously bad in Swedish." When GPT-3.5 was released to developers they started building. The first thing they shipped was a stock option reader that explained how a stock option contract worked — "as startup founders with no legal background, that seemed reasonable." The focus changed fast, from a single tool to an end-to-end system a legal professional would open every day. Building for Europe meant clearing compliance hoops first: all data hosted inside Europe, nothing used for training, no retention, and exemption from human review in the way Azure and AWS are structured. They jumped through all of it and shipped a system a law firm could legally use. The moment it landed is a good scene. They deployed into the largest law firm in the Nordics, whose managing partner had said in a newspaper that AI was "more artificial than intelligent." In the meeting, the founder opened a laptop and asked him to type a query. Legora was wired to Swedish legislation through a RAG system. It answered perfectly. "You kind of see it in his eyes — it's the aha moment." None of the founders were lawyers. He now calls himself a hobby lawyer.
Three non-lawyers started by buying access. The founder messaged lawyers on LinkedIn offering to pay their hourly rate for lunch, interviewed 100 of them, and says none of them ever made him pay.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
Data credibility: Founder interview on Y Combinator's own channel, so treat it as a friendly-venue account rather than a neutral one. The most important thing to flag: no revenue, ARR, customer count or churn figure is stated anywhere in the conversation. The $675 million valuation and the founder's age appear only in the video title, never in the interview — do not present them as sourced from the transcript. The strongest verifiable claim is the $80 million Series B led by Iconiq and General Catalyst with Y Combinator, Benchmark and Redpoint participating, which the founder states directly and which was announced publicly, so it is checkable. 'Tens of thousands of lawyers daily' is a founder's own round number with no dashboard shown. The headcount arc (10 people out of the batch, ~25 in early October, ~100 six months later) is self-reported. Two names are garbled in the auto-generated transcript and should be verified before publication: the founder is rendered 'Max Junior Strand' (we render it Max Junger Strand), the product is inconsistently transcribed as Legora, Lagora, Lora and Agora, and the Spanish law firm in the courtroom anecdote appears as 'Perorka', which we have not attempted to correct. The compliance posture (EU data residency, no training on customer data, no retention, human-review exemption), the Azure-first then multi-model stack, the classification-based model routing, the four-to-five-month sales pause, and the top-down-only sales motion are all as described by the founder and are unaudited.
No revenue, ARR or customer count is disclosed anywhere, so what is visible is the shape rather than the number. The first thing the team shipped was a stock option reader that explained how a stock option contract worked, which is the kind of artifact that earns a meeting rather than an invoice.
Two loops, both structural. Inside an account the product spreads because the playbook does: a playbook is a set of approve and disapprove rules with example language and ordered fallbacks, and once the legal team writes one it becomes the firm's standard.
The instructive stretch here was chosen rather than endured. After raising, at the first board meeting, he told the board Legora would not sell anything for the next four to five months, and remembers the look on some of the directors' faces.
Two surfaces, and the second one is the interesting bet. The web application began as chat over your own documents and files and has become an agent that can call the app's other endpoints plus external tools to run multi-step work. His example: write a memo, where step one is research, step two is conforming that research into the firm's standard language, and step three is writing the report. It does all of it. Alongside the chat is what they call tabular review — a grid where every document is a row and every query is a column. Load a hundred employment agreements, ask whether each includes an IP clause protecting the company's intellectual property, and it "starts to rattle and it goes yes yes yes no no no yes," always linking back to the citation. He is emphatic that the innovation here is not prompting: it is running the thing at scale. "How do you run 100,000 queries in parallel at the same time and make sure nothing breaks, all the citations are correct" — with chunking, RAG inside very long individual documents, and legal-specific intricacies like always including definitions and resolving cross-references between clauses. The second surface is a Microsoft Word add-in, which he describes as "cursor for lawyers." Lawyers draft and review in Word or PDF, and the constraint is brutal: "you can't fork Word and you can't take up all the real estate you want." You get the right-hand column. "It's basically like designing a mobile app, because that's all the real estate you get." Inside it, an assistant reads the document and creates edits — renegotiate this MSA for the buyer, using my internal checklist or playbook. Our read: the Word add-in is the sharper insight. Meeting the user inside the tool they already live in beats asking them to move.
No revenue figure is disclosed anywhere in the interview, and we are not going to model one. What is described is the shape of the commercial motion, which is more transferable anyway. Enterprise seats sold to law firms and corporate legal teams on one- or two-year contracts. He is explicit that longer terms are gone: "we're not seeing anybody want to lock in a five-year contract, because the world is moving so fast." Even law firms, who historically loved vendor lock-in, now sign one or two years. And the evaluation criterion has shifted in a way worth internalising: "yes, people look at the technology, but even more so they're zooming out and they're looking at your rate of change." Buyers are choosing a partner who will get them from A to B, not a feature list. The expansion mechanic is the most valuable part. Playbooks — collections of approve/disapprove rules with example language and ordered fallbacks — turn out to travel outside the legal department. At Legora itself, every sales rep uses Legora to negotiate NDAs before sending them to the legal team. At a very large Nordic bank the product moved from legal to compliance to risk and then to sales. The reason it spreads is not just speed: "the legal team creates the playbook and that becomes the standard that everybody uses," so quality and consistency increase alongside throughput. On the funding side: $80 million Series B led by Iconiq and General Catalyst, with Y Combinator's continued participation plus Benchmark and Redpoint. The $675 million valuation in the video's title is never stated in the conversation.
Top-down only, and he is unusually direct about why. "It's impossible to do a bottom-up motion in our industry because you don't procure software individually." Everything goes through procurement and IT, with security checks and data privacy checks you must clear before client data can touch your system. No amount of product-led growth routes around that. Who buys depends on firm size. The biggest firms have a partner group plus an innovation department whose influence varies; a strong innovation department procures software itself and owns the whole innovation agenda. He says he gets the most energy from those people, and notes the structural oddity that they drive efficiency across the firm without being the users themselves. In mid-size firms there is no innovation department, so partners decide — and getting an entire partnership to agree is hard. His answer to the founder question of how you sell into a firm like that: "you have to convince everybody, or you start smaller. You say, let's work with this partner and their team and make them rock stars. And then everybody else looks at them saying, what's that guy doing? That looks awesome. We also want in." Why law firms became sellable at all around two years ago: legal work is low-differentiation, so when one firm takes a new approach the equilibrium breaks and clients — already under price pressure with high legal fees — switch quickly. Firms are therefore incentivised to move first rather than be disrupted. His analogy is lawyers adopting computers: walk to the library and find the right book, or press Ctrl+F. The one deliberate anti-growth decision: after raising, they stopped selling for four to five months. "The first experience of a legal professional logging in is the one chance you have. If you mess that up, they're not coming back." They fixed reliability and scalability until they could comfortably onboard a thousand lawyers a day, then let it rip.
Azure from the beginning, chosen because it was what their customers were already on — and because early on, GPT was effectively the only model you could serve through Azure. They now use AWS, Claude, Gemini, GPT and Mistral interchangeably. The architectural principle behind that is the reusable part: "the biggest thing there has been how do we build everything in such a way where we can hot swap the models whenever we want, and also build it in such a way that as the models become better, everything improves." That second clause is the strategy, not a nicety. It is the same instinct he describes as building things as boats "so that when the tide rises, just everything gets better." On cost control they route by difficulty. Classification models decide which model a query deserves: "if you do a simple query we'll serve you a simple model, if you do a complex query we'll serve you a complex model." His stated reason is margin — "sometimes you don't need a bazooka when you just need a water gun." No margin figures are given, but the fact that model routing is framed as a margin lever tells you inference is a material cost line in a product that runs a hundred thousand parallel queries against long documents. On the compliance side of the stack: data hosted in Europe, no training on customer data, no retention, and exemption from human review in the Azure and AWS configurations. For European legal buyers this is not a feature, it is the price of admission. Headcount economics are worth noting too. "Our ability to out-ship teams of thousands of engineers with just 30 is insane." On marketing he wants generalists using AI to do ten times more: "where you might have needed a 30-person marketing team, you now need five."
You are not copying this. It is venture-scale, enterprise-sold and compliance-heavy, and it has $80 million behind it. Four things do transfer. Build inside the tool they already use. The Word add-in is the highest-leverage decision in the product, and the constraint that came with it — one narrow column, no forking the host app — forced better design, not worse. Whatever workflow you are attacking, the equivalent question is which application your user already has open all day, and whether you can live inside it. Sell to seniors first, then let envy do the work. Pick one partner or one team, make them visibly excellent, and let the rest of the organisation ask what they are doing. This works in any org where individual purchase is impossible. Build for the models getting better, not for today's models. "It's interesting to build these products knowing that the models will get better." Where a capability is not good enough yet, ship the scaffolding around it — playbooks with explicit rules and fallbacks rather than 'let the model negotiate' — so the product improves for free when the model does. And take his own advice on positioning: "don't get locked in with a provider and don't compete with the AI labs. If you're just starting out, you've got to realise that you do not have the capacity to outperform any of those companies." Find a narrow category the models will not reach, or use them in a way nobody else has. His example is AI scribing, which needs custom prompting to produce genuine medical language — the same as law, where "you needed to write clauses in a way that a lawyer would write a clause, not just what the model spits out as the most probable answer." What to change: skip the enterprise motion. The playbook mechanic applied to a small, self-serve niche — contractor agreements for agencies, vendor terms for small businesses — is the indie-scale version.
He names the biggest risk himself, and it is the platform above him. "The expectations on companies like us are pretty clear: whatever comes out of a model lab is kind of expected, and then everything else we're adding on top is kind of like icing on the cake." Asked whether the labs will attempt law, his answer is that they are becoming platform companies rather than model providers — Google building Workspace with Gemini, Anthropic running hard on MCP as a universal entry point into applications. A vertical AI company's entire value has to live in the part of the stack the labs will not build. Model dependency is structural. The hot-swap architecture is a real mitigation, but the product's quality, cost and margin are all set by third parties whose pricing can move without notice — which is precisely why he treats model routing as a margin lever. Contract length now cuts against the incumbent. Buyers on one- and two-year terms and judging vendors on "rate of change" is exactly what let Legora displace the legacy legal-tech M&A machines. It will also be what lets the next entrant displace Legora. There is no lock-in left to hide behind. The pricing floor is moving underneath the customer. He observes that "clients are also not really that excited to pay for very simple contract review when they know that AI can do 99% of it," and that due diligence is becoming a commodity. If the billable hours Legora compresses are the same hours funding its customers, the buying budget is being squeezed by the product's own success. Hiring at two people a week is its own failure mode: he says teammates must scale exponentially, because "if people scale linearly, at some point it's a really large delta and then things aren't working out anymore." Finally, the evidence gap. No ARR, no revenue, no customer count, no churn. "Tens of thousands of lawyers daily" is a founder's own round number, and the valuation comes from a headline, not the interview.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,435 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click