Legora CEO Max Junestrand on Legal AI, Growth and Competition

Legora cofounder and CEO Max Junestrand joins Harry Stebbings to discuss legal AI, model providers, enterprise adoption, competition and law-firm employment. Junestrand speaks for a commercial provider, while Stebbings discloses investments relevant to several examples; company metrics, competitive claims and forecasts are not independently verified in the interview.

Junestrand describes Legora as a platform for legal work, with tools for tasks such as reviewing documents in due diligence and helping draft litigation briefs. He says time spent, messages and actions are important product metrics. These are descriptions of intended use and company priorities, not independent measures of legal accuracy or professional outcomes.

Stebbings asks why Harvey is often associated with legal AI. Junestrand disputes the premise, citing a report he says ranked Legora ahead of Harvey among deployed generative AI tools at leading UK firms outside Microsoft Copilot. The conversation does not independently inspect the report or establish overall market share.

The CEO says law firms commonly compare vendors in pilots and choose a partner based on both current outcomes and future plans. He says Legora grew from roughly 30 to 300 employees and from about 50 to 750 clients in a year. The figures are company-reported, and client counts do not specify contract value or usage.

Junestrand describes a legal-engineering team of former practitioners helping clients implement and adopt the product. He cites White & Case as an example and says a firm-wide rollout requires training across offices, practice areas and seniority. His view is that hands-on implementation matters when a tool changes how an enterprise works.

Asked what he learned from competition, Junestrand says Legora chose to build around improving general models rather than devote its early resources to fine-tuning. He notes that the company began with three engineers and $50,000 in angel funding. His assessment of a competitor’s approach is a commercial opinion, not proof of its costs or effectiveness.

He praises Claude Opus 4.5 and uses the term AGI for its performance on some coding tasks. This is his subjective characterization of a tool, not an established scientific finding that the model has general intelligence or reliably performs every task.

Junestrand says Legora initially used OpenAI models and later switched predominantly to Anthropic, before Opus 4.5. He says adapting prompts and building enterprise software around a chosen model required concentrated effort. His team also uses both Claude Code and Cursor; those choices reflect the company’s experience at the recording date.

He argues that applications need interfaces, safeguards and workflow software around models, and that clients may add their own differentiated tools. He estimates much of the work lies in ordinary software engineering. That proportion is his judgment, not a measured attribution of product value.

Junestrand says the company should change model providers when evaluations show a better fit or price, and could let users select a model for specific workflows. Asked to rank models two years ahead, he guesses Anthropic or Gemini may lead for his work and places OpenAI behind them. This is a conditional forecast, not an objective ranking or proof of future performance.

On continuous inference, he says Legora did not yet have tasks taking twelve hours, though he expects more work to run while users are away. He describes agents that plan tasks and interact with tools, and says some partners compare output with work assigned to associates. His observations do not establish accuracy or replace legal review.

Stebbings raises the perception that Harvey has won the US while Legora has won Europe. Junestrand disputes the US conclusion and says the United States became Legora’s largest single country by revenue, though Europe collectively remained larger. He predicts the US may exceed Europe by the end of the next quarter; that is a forecast, not reported achieved revenue.

He says Legora tested US demand from Europe by serving two large firms before opening a local office, then hired roughly 50 people in the country. He expects the office to expand further. Client names, employee counts and future hiring plans are company statements, not evidence that the broader market has been won.

Junestrand says shorter US notice periods helped the company recruit quickly compared with Sweden, where he describes longer typical periods. He advises some enterprise founders to validate customer demand before committing heavily to expansion, while noting that the strategy may differ for consumer businesses. His comparison is an operational observation rather than legal advice.

He recalls telling the board after early financing that Legora would pause sales for six months to rebuild infrastructure and improve reliability and scalability. He says the company delayed onboarding while it prepared to handle more lawyers. The CEO presents this as a deliberate tradeoff; the interview does not independently establish the counterfactual churn he believes it avoided.

Junestrand says legal AI purchases often remain extended trials, with one- to three-year contracts that are short relative to the age of major law firms. Stebbings cites competitor retention figures and asks for a comparison. Junestrand says it is too early to interpret Legora’s net revenue retention meaningfully because much growth comes from new customers.

He says Legora added $7 million of annual recurring revenue in one day in December 2025, more than its total in the two prior years combined. ARR annualizes recurring contracts and is not the cash or recognized revenue earned that day. He says 2026 renewals will be needed to establish meaningful retention figures; the interview does not independently audit the claim.

Legora charges per seat, which Junestrand says is easier for buyers to budget even though heavy usage can increase model costs. He expects a move toward consumption pricing when customers are ready, but does not know the timing. His prediction of a change within three years is not a confirmed pricing plan.

He describes margins as acceptable but below traditional SaaS levels and expects them to improve over time. He argues the product could eventually be priced against legal work completed rather than a narrow software tool. These are the vendor’s commercial claims and forecasts, not proven savings or profitability.

Junestrand says model selection can help manage costs but considers the company’s current phase a push to gain customers rather than maximize margins. He identifies maintaining teamwork, ambition and execution while growing from 300 toward a planned 600 employees as a major challenge. The hiring target is prospective, not a completed headcount.

He says he interviews candidates and tries to recruit people committed to the mission, while describing late working hours and team momentum. He believes celebration and competition motivate employees. These are his management preferences, not evidence that long hours prevent burnout or suit every worker.

Asked whether competitors copied Legora, Junestrand avoids confirming the specific allegation and says superficially similar products can differ in competitive pilots. The interview does not verify copying or establish which product performs better. He emphasizes both improving existing features and making new product bets.

He recalls that the first 2023 product was organized around fixed use cases without an agent or chat and says the team deleted that code after entering Y Combinator. The company built its own early agent architecture, later rebuilding it several times. He now prefers choosing suitable third-party components where appropriate.

He says a small engineering team tried to build too many things at once in 2024. A subsequent product manifesto narrowed its focus to an agent or assistant, tabular review and a Word add-in, and the team removed other work. He views the combined suite as more valuable for customers than separate products; that is the CEO’s assessment.

Asked whether the market will resemble ride-hailing or cloud infrastructure, Junestrand argues that legal products can differentiate more deeply through functionality. He predicts a highly concentrated market and suggests the leader could take 90%. The percentage is his competitive thesis, not observed market-share data or a certainty.

Stebbings discloses an investment in a patent-focused AI company and asks about specialist tools and AI-native law firms. Junestrand prefers supplying software to existing lawyers over owning both the service and software layers, while recognizing that specialists may create value. His views reflect Legora’s business model and are not evidence that a named rival will fail.

He says a central platform might eventually call specialist products for work such as patent drafting, while acknowledging uncertainty about how the ecosystem develops. This hypothetical integration does not establish a partnership or guarantee legal quality. Stebbings’ investment interest is relevant to the example.

Junestrand clarifies that the US team is around 50 people, not 150, and predicts it could reach 150 before summer. He rejects a simple claim that US workers inherently work harder than Europeans and credits colleagues who helped establish Legora’s culture. Future hiring is a plan, not a completed result.

He says much of the engineering team is in Stockholm and describes organizing product work into autonomous teams, including employees with founder experience. His management approach favors delegation and accountability for a product area. The claimed composition of the team is company-reported.

Asked for revenue, Junestrand repeats the $7 million single-day ARR addition and says the business doubled each quarter for six quarters, without disclosing a current total. He predicts reaching $200 million by year-end and discusses a possible higher target. These are company claims and forward-looking goals; ARR is not recognized revenue and no result is guaranteed.

On law-firm structure, Junestrand predicts consolidation and says technology could help larger firms compete on price and speed. He sketches a possible shift from many leading firms to a much smaller group, but the figures are scenarios, not announced mergers or a forecast supported by an independent model.

He expects fewer people may be needed on an individual transaction and ultimately acknowledges that there may be fewer junior lawyers and trainees. He also imagines that firms could handle more transactions. These are competing possibilities and employment predictions, not evidence that any specific firm has cut positions because of Legora.

Junestrand says large firms may benefit from brand and data, small firms from personal relationships, while midsize firms may face price pressure. He expects more people to write code as engineering productivity grows; Stebbings is more concerned about near-term labor displacement. Neither prediction establishes future employment levels.

He expects billable hours to change slowly because clients often want detailed breakdowns, even if fixed fees grow in some work. This is an assessment of billing practices, not a legal or financial rule. The interview does not establish the pace at which firms will change pricing.

Asked about raising three rounds without a deck, Junestrand credits the strength of the underlying business and describes choosing Benchmark partner Chetan Puttagunta despite a lower seed valuation than another offer. His relationship with the investor is a relevant interest; the financing story is his account rather than proof that every company can raise similarly.

He says he expects some narrow AI products to struggle and sees value in broader platforms, a view consistent with Legora’s strategy. He also reflects on the personal intensity of working on the company for more than two years and says he wishes he had anticipated its effect on the rest of his life.

Junestrand says Swedish technology founders inspired him, but he tries to draw lessons from many sources instead of following one idol. He credits advice from Y Combinator and others to choose his Benchmark partner. The conversation closes with Stebbings noting that he did not invest in Legora.