Legora launched with 3 founders and 0 lawyers.
Tom Blomfield laughed at them in their YC interview. Not politely, an actual laugh, because they couldn’t answer “what type of lawyers are you serving?” They didn’t know lawyers came in types.
3 years later, over 3% of the world’s lawyers use the product they built anyway. The Swedish VC who passed on their pre-seed, specifically because the team had zero legal background, still brings it up whenever he runs into Max Junestrand at the airport.
“And so you can just do things.”
That’s Junestrand, looking back on the moment he skipped the safe path everyone expected of him and built instead.
I watched the full interview so you can skip it. Here are the 10 takeaways that matter.
Together with Vanta:
Legora didn’t stop at one market. 3% of the world’s lawyers now use it, across borders most legal tech never crosses.
Going global changes what “secure” means. A US customer wants SOC 2. A European one wants GDPR answered before the second call. An enterprise buyer anywhere wants ISO 27001, the standard that says your security program holds up outside your home market too.
Vanta built the checklist for exactly that moment: the ISO 27001 checklist for teams ready to go global.
▫️ What the standard actually requires, without the consultant markup
▫️ The gaps that stall international deals, flagged before a prospect finds them
▫️ Free, and built for teams scaling past their first market
1. Why Legora's Founders Weren't Lawyers
Domain expertise felt like table stakes for a legal AI company. Legora didn’t have any, and built the category leader anyway.
“I think that you need a willingness to learn about the markets, which we did very, very early by exposing ourselves and spending lots of time with lawyers, as much as we possibly could. But I don't think you need domain expertise.”
A Swedish VC passed on Legora’s pre-seed for exactly this reason: 0 lawyers on the team. He’s still bitter about it, years later, whenever their paths cross.
That’s the tell. Domain expertise looks like a moat right up until someone without it learns faster than the people who have it.
Junestrand’s caveat matters, though. He’d want deep technical grounding before walking into quantum computing or fusion, where the barrier to even understanding the problem is genuinely high.
Legal wasn’t that kind of barrier. It was learnable, so the founders learned it, on lunches with lawyers instead of resumes with the right credentials on them.
“Domain expertise is a lagging indicator. Learning speed is the leading one.”
2. The YC Question Legora Couldn't Answer
They applied to YC with a pitch about querying every legal document with LLMs. The first question undid them.
“One of the first questions that we got asked was, what type of lawyers are you serving? And not very knowledgeable about the legal space yet. Our response was, what do you mean? Are there different type of lawyers?”
Tom Blomfield laughed. Not a polite laugh, an actual one, in the middle of the interview, and Junestrand knew right there they were done.
They were. The day after brought nothing, but the day after that, they were back to building.
2 months later they reapplied under a different name, with a different platform, and got in.
The actual lesson isn’t resilience as a slogan; it’s narrower than that. They used the rejection as a forcing function to close the exact gap they’d been laughed at over, instead of polishing the pitch deck and reapplying with the same hole in their knowledge.
3. Legora Raised $35M, Then Went 6 Months With 0 New Sales
Benchmark had just wired $9.51 million. Redpoint pre-led the Series A 3 weeks later. Then the founders stopped selling.
“When you work with lawyers, you only really get one chance to get it right. If you show up and the product doesn't work, or if the sort of lag in the system is too high, or if a system goes down when there's too much traffic on your Assure instance, you are toast. And after that sixmonth sales freeze was when we really started ripping.”
$35 million in the bank, 10 people on the team, and a month where interest on that cash outpaced revenue from customers. That’s not a great sign for a startup; it’s a sign you’ve accidentally become a bank.
In their first board meeting, they made a call that looks reckless on paper: 6 months, 0 new sales, while every competitor raced to close logos off fresh funding.
It wasn’t caution for its own sake. It was a bet that in a trust-based vertical like law, one bad first impression closes a door permanently, and a closed door in a market this small doesn’t reopen.
The curve from $1 million to $100 million in ARR starts right after the freeze ends. Correlation isn’t proof, but the founders read it as cause.
If you’re weighing your own freeze-or-scale call right after a raise:
▫️ Why Your Series A Might Kill Your Startup
▫️ What Top-Tier VCs Actually Look For in 2026
4. The Leia Product Manifesto Turned $1.3M ARR Into $100M
Before the freeze, Legora decided what to build by team vote. It went about as well as that sounds.
“For anyone who has built software before, you know that if you have too many chefs in the kitchen, that typically does not make one very good dish. And that was exactly what was going on.”
At $1.3 million ARR, with competitors doing 10 times that on products with a fraction of the feature set, the founders wrote down what they called the Leia Product Manifesto in October 2024. 1 document, shared with all 25 people at the company. No more votes.
It wasn’t a mission statement. It was a filter, built to survive a platform that had to keep adapting as the underlying models, and the agentic frameworks built on top of them, kept shifting.
That refocus gave Legora enough product momentum to justify the move from Sweden to the US, and to start competing for the biggest logos in the market.
5. Bloomberg Spent Millions Fine-Tuning. Legora Didn't.
Tribal knowledge in 2022 said you had to fine-tune a model to compete. Legora couldn’t afford to, so they didn’t.
“Our view was partly because we didn't have enough money and partly because we truly believe that to be right. The models will keep improving, right? Like I'm sure Sam is going to come up on stage and say that later today.”
Bloomberg reportedly spent millions building a proprietary law model on the fine-tuning theory. Legora didn’t have the capital to compete on that front, so the constraint became the strategy: bet on the frontier moving, and build the delivery layer on top of it instead of fighting to own the model outright.
Even early ChatGPT was borderline unusable for law, since it had no private conversations and no European data hosting. Junestrand’s actual first sales pitch was that Legora was like ChatGPT, but compliant in Europe.
He calls it a weak pitch now. It worked anyway, because the model kept getting better underneath it while the compliance wrapper held the customer relationship in place.
6. The Real Skill Is Knowing How to Eval Models
Model choice used to be simple. 1 or 2 options were good enough to run Legora. That’s no longer true, and Junestrand thinks that’s a good thing.
“One of the core IPs and muscles that I encourage as many of you as possible to build is the ability to eval new models and to eval new use cases, because that is the superpower that then allows you to route things effectively.”
Different jobs want different models. Customer support optimizes for latency and cost, since token spend has to stay tiny relative to the value of the ticket. Complex litigation wants the opposite: throw the most capable model available at the problem, because the cost of intelligence is trivial next to the cost of a lawyer’s time.
Legora’s users split into 2 camps. One wants to loop expensive frontier models on hard problems regardless of cost. The other wants cheaper open-source access.
Legora hired lawyers early specifically to build eval sets against genuine use cases, and only made its internal benchmark, Legora Bench, public in the days before this talk, after 3 years of keeping it closed. The surprising finding: Grok performed among the best models for cost given performance, and Legora didn’t even have a data processing agreement in place to offer it to customers yet.
7. Why Legora Moved From Reactive Agents to Proactive Ones
For 3 years, Legora worked the way most AI products still work: give it a prompt, it does the thing, it stops.
“We are basically connecting Legora to different pieces of context, and when it gets a trigger, it will start to do something. That means that if the sales team gets a contract, it will get routed to a Legora agent. The Legora agent will start working on it in parallel.”
No instruction required. A contract lands, an agent picks it up, works it, and either executes it or escalates to a human lawyer depending on what it finds. Connect an entire data room to an agent and it starts organizing the room and drafting the due diligence report on its own.
The framing Junestrand uses for the payoff is blunt: 1 lawyer producing the output of a team of 10. That’s not a feature update. It’s a claim about what headcount at a law firm should even mean going forward.
Building the eval-and-agent muscle before you need it:
▫️ Brian Armstrong Runs 1,200 AI Agents at Coinbase
▫️ Satya Nadella Said Every Agent Needs Its Own Computer
▫️ Google Just Made Agents 3x Cheaper to Run
8. What Is the Y Intercept Hiring Mistake
Early hiring at Legora chased resumes with fancy logos on them. They called the pattern “Y intercept.”
“If you think about somebody's skill curve, it might start out really high, but if they don't have a good trajectory upwards, they are going to have a really hard time working in a company that is scaling exponentially.”
A joke about someone whose skill curve starts high but goes nowhere. They stopped hiring for the starting point and started hiring for slope.
Their top seller today is 23, sold over $10 million worth of Legora, and walked in straight out of university with no sales background at all.
Junestrand, an engineer by instinct, admits the team overweighted product and technology early on and underweighted people. Building a product and building a company turned out to be genuinely different skills, and the trajectory bet on hiring is the one he credits most directly with the culture that still lets 750-plus people move like a startup.
Rebuilding a hiring filter around slope instead of pedigree:
▫️ The Free AI Hiring Kit Every Startup Founder Needs
▫️ Free AI Hiring Kit for Founders
▫️ The Salary That Disappeared: 6 Roles AI Owns Completely Now
9. What Startups Can Do That Microsoft Can't
Microsoft’s Copilot launch scared Legora badly. Every lawyer already lived in Word and Outlook, and the fear made sense at the time.
“I remember when that came out and we were like, oh my God, like we're so screwed. You know, every lawyer already works in the Word and Outlook and now they're just going to use copilots, but turns out like it didn't work. And the ability for us then to still build a lot of value was huge.”
It didn’t work, and Legora kept the advantage that actually mattered: speed of iteration against direct customer feedback.
Junestrand still drops individual customer complaints straight into the product Slack channel and asks how fast the team can turn it around, even now, at a company approaching 1,000 employees. He calls it a bit strange to do at that scale. He does it anyway.
“Big companies protect you from discomfort. Small teams don't.”
10. Why This CEO Re-Qualifies for His Own Job Every Quarter
Growing from $1 million to $100 million in ARR in 18 months means the job changes underneath you constantly.
“And I think you really need to put the company first way ahead of your own ego in a way. And I tell this to the executive team. I need to re-qualify for the job as CEO of Legora. Every quarter, right? It's a new company, new challenges. As do they. Like running a sales team when you are 1 million ARR is very different from 150 million ARR.”
Junestrand describes the growth as a series of mini-games, each requiring skills the last one didn’t. Executives joining from companies that grew “triple, triple, double, double” over 4 years now have to compress that same growth into 1 year at Legora. A 40% year-on-year growth rate, fine for most SaaS companies, isn’t enough here anymore.
The requirement isn’t heroics. It’s admitting that competence at the last stage doesn’t transfer automatically to the next one, and building the habit of checking that assumption on a fixed schedule instead of waiting for a crisis to reveal the gap.
The Learning Speed Playbook
The willingness to learn faster than everyone else beats every credential you don’t have.
▫️ Founders: Stop waiting for domain expertise to feel earned. Log hours with the people who live in the problem, pay for their time if that’s what it takes, and treat the learning curve as the actual roadmap.
▫️ Investors: Team background is a weak signal in fast-moving categories. Learning velocity, evidenced by how fast a team closed a specific knowledge gap after a failure, is a stronger one.
▫️ Operators: Build the eval muscle now. The pool of usable models isn’t settling into 1 or 2 winners, it’s fragmenting by use case, and routing between them correctly is turning into its own discipline.
▫️ Everyone else: Re-qualifying for your own job on a fixed schedule isn’t a CEO-only habit. If your role has changed shape recently, run the same quarterly gut check.
The 5 Principles to Steal
Bet on the frontier, not the fine-tune. If your product’s value depends on a model that isn’t good enough yet, build the delivery layer and wait for the model to catch up rather than trying to own it yourself.
Freeze sales before you freeze trust. In markets where one bad first impression is permanent, slowing down on purpose beats scaling into a reputation you can’t undo.
Vote once, then write a manifesto. Democratic product decisions work at 3 people. They stop working well before 25. Write down the filter and stop re-litigating it every sprint.
Hire for slope, not for altitude. A high starting point with a flat trajectory loses to a low starting point climbing fast, every time, at a company actually trying to scale exponentially.
Re-qualify for your own job on a schedule. Don’t wait for a bad quarter to notice the role changed under you.
If this breakdown saved you the 60 minutes of watching it yourself, send it to one founder or investor who needs it.
If domain expertise was never your actual gate either
▫️ Career Zigzaggers Are Winning in AI: Why Non-Linear Paths Create Alpha
▫️ Founder-Market Fit: The #1 VC Filter You Didn’t Know Was Judging You
▫️ You’re Not a Fraud. You’re Undervalued: Why Reverse Imposter Syndrome Hits Founders Hard
If you’re trying to figure out where the AI moat actually sits
▫️ Nobody Cares About the Model Now. It’s About the Type of Moat
▫️ Why Companies Invest in Open-Source Tech and Research
▫️ Marc Andreessen: The AI Moat Is Not the Model
If you’re scaling past your first $1M
▫️ How Lovable Hit $400M ARR in 14 Months With 146 People and Almost Zero Paid Ads
▫️ Jensen Huang: 10 Lessons From the CEO Building the Most Important Company in History
▫️ You Won’t Believe What These 12 CEOs Actually Wrote to Their Employees
Building without domain expertise is one thing; grading the output is where it returns. Stanford's RegLab team ran the first preregistered evaluation of AI legal research tools, published in the Journal of Empirical Legal Studies, and measured hallucination rates of 17 percent for Lexis+ AI and 33 percent for Westlaw AI-Assisted Research across 202 queries hand-scored by legal experts. Both products had marketed the opposite. Nobody without legal training could have scored those answers. So the founder lesson holds on the build side and inverts on the buy side.
The no-domain-expertise rule holds for law, but I'd separate two things he's merging: not knowing the industry, and not knowing the customer's workflow.
Legora closed that gap by hiring lawyers early to build eval sets. Most founders skip that step, call it "learning fast," and end up shipping features nobody asked for.
Learning speed is the leading indicator only when it's directed at the buyer's actual day, not just the market's vocabulary.