Legora CEO, Max Junestrand: $7M ARR in a Day | Harvey vs Legora: Is a Legal AI Winner Takes All?

Summary

• Max Junestrand’s core market call: legal AI is “totally a winner takes all — like all SaaS. Number one will grab 90% and number two to number ten will share the remaining 10%.” His case for Legora’s momentum against Harvey: a Bloomberg infographic showing Legora as the most deployed genAI tool in the UK’s top 200 law firms outside Microsoft Copilot, clients up from 50 to 750 in a year, headcount from 30 to 300 — “it doesn’t really matter who was first, it matters who’s best.”
• The enterprise model-share signal in the title: Legora was only OpenAI through 2023 and most of 2024, and is now “pretty diehard Anthropic” — the switch came around Sonnet 3 or 3.5. His 24-month ranking: Claude or Gemini on top, OpenAI third (“no, we’re not going to throw Grok in there at all”), because “Anthropic is going more enterprise and OpenAI is going more B2C.” The loyalty caveat: “We will be very promiscuous — and that’s a clip.”
• On capability: Opus 4.5 in coding — “you should just coin it AGI and focus on optimizing the cost.” GPT-3.5 “felt like managing an employee who wasn’t very intelligent”; Opus 4.5 “is like a VP. You give it: here’s the thing I want. Go execute.” His read on Harvey’s early mistake: fine-tuning models in 2023 was wasted effort — “we should be building boats, and when the tide rises, all of our products just get better” — since 80% of the value is “building normal software” at the application layer.
• The Harvey-won-the-US, Legora-won-Europe narrative: “I think one of those statements are true.” Legora went from zero people on the ground in the US at the start of the year to 50 now, with the US already its biggest single market by revenue and set to pass all of Europe by end of Q1 — after a beachhead heuristic of signing two AmLaw 200 firms (White & Case and Goodwin Procter) from Europe before opening a single US office.
• The growth stat of the episode: $7M of ARR added in a single day in December 2025 — more than 2023 and 2024 combined — after doubling every quarter for six straight quarters; 200 by year-end is “Definitely. Definitively.” But he’s honest that everyone is “treating this as an extended pilot and an option on AI — like a call option”: one-to-three-year contracts, no five-year deals, and NRR comparisons vs Harvey’s claimed 98%/178% are premature — “it’s up to 2026 to determine where those real numbers will be.”
• Rare pricing candor: per-seat is “optimal for the buyer, I don’t think that’s optimal for us” — heavy users rack up unsustainable LLM costs — and seat pricing in three years is “absolutely not” surviving. Margins are “okay… not SaaS margins,” but the ceiling is high because AI work is “priced against what would I pay a lawyer to go out and actually do this work,” not against other software. For now: “land-grab time,” not margin-optimization time.
• The end-market call: law firms enter PE-fueled consolidation — “I don’t think there’s going to be an AmLaw 200. I think it’s going to be an AmLaw 20, or maybe AmLaw 12” — with big law and small law winning, mid-law squeezed, fewer junior lawyers, and billable-hour billing changing much slower than expected. Harry’s pushback that labor displacement shows up in figures within 12–24 months gets a partial concession: “on the total level, yeah, probably.”

Deep dive

1. Legal AI is winner-take-all — “there’s no number two”

• Max’s framing of the category: “It’s totally a winner takes all — like all SaaS. Number one will grab 90% and number two to number ten will share the remaining 10%.” The operational consequence: “You got to run like hell. You got to win. There’s no number two… There’s only winning and everything else is losing.”
• Why this isn’t Uber vs Lyft: in ride-hailing “there was no product differentiation — the products were pretty much the same.” In legal AI, product differentiation really matters, and the buildable surface is “this universe of legal technology that just has never been built” — pre-genAI legal tech founders took life-changing acquisition offers at single-digit-million revenue, with “no unicorn outcomes.”
• His counter to Harvey’s brand lead: a Bloomberg infographic showing Legora as the most deployed genAI tool in the UK’s top 200 law firms outside Microsoft Copilot, with Harvey second — “it doesn’t really matter who was first, it matters who’s best. Google was not the name you thought of back when AltaVista was the biggest web browser.”
• The momentum proof: 30 to 300 headcount in exactly 12 months, 50 to 750 clients, won through bake-offs where firms are “basically playing VC” — buying not a tool but “an outcome today… and a vision of what an AI-enabled legal team can look like.”

2. From only OpenAI to “pretty diehard Anthropic”

• The migration: Legora ran only OpenAI through 2023 and most of 2024; it is now “majority using Anthropic,” with the switch made around “Sonnet 3… or 3.5.” The logic: models started demanding different prompting, so “just pick a good model and double down on it and build all the application around it.”
• The structural split he sees — and Harry endorses with a “100%”: “Anthropic is going more enterprise and OpenAI is going more B2C” — “and we’re an enterprise-class type of system, thus we should benefit more from their models.” OpenAI’s drift toward letting users fine-tune models is a journey “so far I don’t have a lot of reason to believe in.”
• His 24-month model ranking: Claude or Gemini on top, the tiebreak being whether context window becomes decisive — so far it isn’t, because Legora “built so much architecture around handling a lack of context window.” Then OpenAI. Grok: “No, we’re not going to throw Grok in there at all.”
• But no sentimentality: “We will be very promiscuous — and that’s a clip.” Clients entrusted Legora as their AI partner, so “if Gemini is better, we will switch immediately… provided that the eval is better” — with users able to pin specific models for deterministic workflows.

3. Application layer beats fine-tuning — and Opus 4.5 “is like a VP”

• Asked what Harvey did wrong, Max names fine-tuning models in 2023: with general models improving so fast, “we should be building boats — and when the tide rises, all of our products just get better.” He concedes the thesis was partly forced — three engineers and $50K of angel funding can’t spend $3M fine-tuning — “but it happened to be right,” and “I doubt that’s going to be the difference maker today.”
• His value pyramid: models at the base, Legora’s “legal interpretation” of them above — but “80% comes from building normal software,” enterprise-grade scaffolding. At the top, clients doing differentiated things: firms are “vibe coding tools internally” and building MCP servers, because law firms used to compete on expertise, recruiting and rates but “increasingly you’re competing on tech.”
• On whether models are plateauing: “No — Opus 4.5 is awesome… it’s gotten to the point in coding where you should just coin it AGI and focus on optimizing the cost.” The contrast: GPT-3.5 “felt like managing an employee who wasn’t very intelligent,” whereas “Opus 4.5 is like a VP. You give it: here’s the thing I want. Go execute. And it just does it.”
• The paradigm he’s porting into law from Claude Code and the new “co-work” tool: give the Legora agent every ecosystem tool plus client MCP servers, let it plan and execute — “pretty much the way that a partner would work with a senior associate,” a new bottom layer of the firm hierarchy available 24/7. Partners already give the same task to an associate and Legora simultaneously, “and very often the quality is pretty good — and that has real implications.” On Jason Lemkin’s 24/7-inference-this-year call: “we don’t have any task in Legora that would take 12 hours to run yet,” but the start-it-before-bed world is coming “for sure.”

4. “Harvey won the US, Legora won Europe” — “one of those statements are true”

• Against the US-VC consensus: Legora began the year with zero US boots on the ground and now has 50 people, a Manhattan office opening this week set to have 150 people, three more US offices coming this year — and the US is already its biggest single market by revenue, set to exceed all of Europe (including the Nordics, where “we basically work with all the big firms”) by end of Q1.
• The entry heuristic, mindful of European launches that failed (“I think Klarna tried a couple of times before it really worked” — likely Klarna): sign and serve two AmLaw 200 firms from Europe first. He landed White & Case, a white-shoe Wall Street firm, and Goodwin Procter, “both in the top 20,” via demos, flights and “always competitive pilots” — without investing a ton in US marketing or B2C content.
• A structural edge he flags for founders: two-week US termination periods vs roughly three months in Sweden. When you double every quarter, “the minute I know that I need somebody, if they wait a quarter, we’re a different company.”
• On the Americans-work-harder trope: “It’s a bit of bullshit” — Legora seeded its US culture from Europe. Though he admits New York “is always up. It’s always awake. It’s my tempo,” against a “little sleepy” Stockholm — and he’s “de facto living on a plane,” 200 travel days last year.

5. $7M of ARR in one day — and the six-month refusal to sell

• The headline number: in a single day in December 2025, Legora added $7M of ARR — “more than what we did in 2023 and 2024 combined” — after doubling “every single quarter for the last six quarters.” Will they hit 200 by year-end? “Definitely. Definitively. Otherwise… you can shame me.” His real target is default status: “If you’re a lawyer and you do serious legal work, you’re on Legora. It’s like Figma.”
• On matching Harvey’s claimed 98% logo retention and 178% NRR: “for both those numbers, yes” — but he calls NRR unfair to cite when growth dwarfs the 2024 renewal base: “it’s up to 2026 to determine where those real numbers will be.”
• The honest caveat on stickiness: every buyer “is still treating this as an extended pilot and an option on AI — like a call option.” One-to-three-year contracts, not five — “many of these firms have been around for 200 years… two years is a blink.”
• The gutsiest decision: at his first board meeting — 12 employees, fresh off $10M from Benchmark and $25M from Redpoint at $150M a month later — he announced Legora would not sell at all for six months to rebuild infrastructure, because “we only have one shot with these lawyers… if it doesn’t work, they’re not going to come back.” By October 1, 2024 they could onboard 1,000 lawyers a day; “if we had continued to push, we would have just churned everything.”

6. Per-seat pricing is wrong — and he says so out loud

• Rare candor about his own revenue model: per-seat pricing is “optimal for the buyer. I don’t think that’s optimal for us… I actually don’t think it’s the right pricing model” — individual users can rack up LLM costs that become “unsustainable on a per-user basis.” Seat pricing in three years? “Absolutely not.” The pivot to consumption comes “when our clients are ready to buy on consumption,” not when Legora is.
• Meanwhile, usage expansion is “incredibly useful for retention, not for revenue optimization” — more usage “actually costs a lot more.” Pressed on margins: “We have okay margins… it’s not SaaS margins” — but he believes they get there, because the ceiling isn’t software comps: “you’re being priced against what would I pay a lawyer to go out and actually do this work.”
• Why not optimize model routing for margin now, Lovable-style? They do some, but “I don’t think we’re in the model-margin-optimization time yet.” Harry’s phrase, adopted on the spot: “You’re in the land-grab time.”

7. Product discipline: delete the code, make three bets, sell shovels

• Two admitted mistakes: the summer-2023 v1 — click-and-point use cases with no agent or chat, “clearly the wrong direction,” all code deleted after YC acceptance — and 2024’s sprawl, when 12 engineers built “six or seven different things” while Max was out selling, producing “a Frankenstein monster.” The fix was the October/November 2024 “Leya product manifesto” (Legora was still called Leya): kill five or six products, hard-commit to the agent, assistant, tabular review and Word add-in as “the best basket money could buy.”
• On copycats — one competitor shipped a feature literally named “tabular review”: fine, but in competitive pilots “the user starts to rip them apart. That’s where you see one product is a Rolls-Royce and the other is maybe a Volvo.” (Harry’s aside: Harvey internals reportedly call Max “their CPO.”)
• On AI-native law firms à la Crosby: not a winning strategy — the low-complexity wedge (NDAs, MSAs) gets crowded fast, big law already does NDAs free to win “the expensive private equity work,” and “as soon as AI can do a task, it will do that task.” He’d rather be “the shovel seller to all the world’s amazingly talented lawyers.” Vertical specialists like Solve (AI for patent lawyers — a $485B market, per Harry) can win, or become spokes a central hub like Legora pings: “why don’t we just ping Solve Intelligence and say, come write us this patent?”

8. AmLaw 200 becomes AmLaw 12 — fewer juniors, but the billable hour survives

• The structural call: law firms enter “a quite significant consolidation period,” with private equity now wanting in — “I don’t think there’s going to be an AmLaw 200. I think it’s going to be an AmLaw 20, or maybe AmLaw 12,” though not a Big Four, because of regulation and time.
• The mechanism: bread-and-butter M&A work is undifferentiated at an equilibrium price — say £100K. The first AI-leveraged firm to offer that at a lower price with the same quality, or perhaps higher speed, “breaks the equilibrium — everybody has to move.” It becomes a market-share game where big law wins (moats, brand, data), small law can do really well (personal relationships), and mid-law gets squeezed — “you throw AI into the mix and it will make it even more competitive.”
• Yes, fewer junior lawyers and trainees — but bigger firms overall. He already sees firms not filling vacancies while growing revenue, “so it’s a higher profit” flowing to partnerships. His engineer analogy — “we’re hiring more engineers although they’re writing more code” — draws Harry’s pushback: labor displacement will exceed expectations and show up in labor figures within 12–24 months. Max half-concedes: that’s exactly the window when “AI gets pretty good at completing end-to-end tasks very deterministically within our vertical” — a single firm can grow the pie, but “on the total level, yeah, probably.”
• The billable hour: not dead. “Billing will move much slower than both you and I would think” — clients themselves demand the breakdown — with fixed fees creeping in by practice area and “a sort of ephemeral billable hour above that.”

9. Missionaries, mercenaries, and picking partners

• The biggest current challenge isn’t competition, it’s doubling from 300 to 600 in two quarters while keeping “the ambition, integrity, teamwork and just raw grit that got us here.” He still interviews everyone — “I try to create missionaries, not mercenaries” — and stokes competition deliberately, down to micro-levels: signing deals on New Year’s Eve, a live sales dashboard at the Christmas mulled-wine party. His own arc: he once celebrated a McKinsey offer “by buying a bag of peanuts”; this year Legora learned to celebrate — because feeling the wins makes you feel the losses, and “you need to see the world for what it is.”
• Team structure as thesis: all engineering in Stockholm, organized in pods, and 10% of engineering/product/design are YC founders — “there’s a YC founder running part of the product and they just run like hell on their thing.”
• The YC advice he lives by — “build a good business and it’s very easy to fundraise.” Benchmark was “the lowest price out of any seed fund,” and another firm offered a much higher preemptive term sheet — but he took the discount for the partner (likely Chetan at Benchmark): “I met within two weeks probably 80 partners. Nobody knew my space better than him… he’s taken three companies public.” Next company, one investor to bring: “Chetan. That’s easy.” Best advice ever received, from YC and Joel at likely Sana Labs: “take the check with Benchmark over anyone else.”
• What he wishes he’d known: the psychic cost of total focus — two and a half years thinking about nothing else, “like running a mega sprint as part of a marathon. And I love it” — but knowing it going in “would have made me better at handling the parts of my life that I couldn’t focus as much on.”
• He was super nervous meeting likely Daniel Ek from Spotify. Growing up, he drew inspiration from likely Niklas, Daniel, and Sebastian and their successful Swedish tech companies, and now admires Alex and Gustav, who are taking over from Daniel. He doesn’t have daily idols; he draws inspiration from good examples and runs at them.