The Problem Solvers: Max Junestrand at Legora

1. The Oldest Job in the World

Max Junestrand opens with a recognition that puts Legora in historical context: legal work is one of the oldest jobs in human civilization. Interpreting the law, drafting agreements, and structuring how society operates are activities that have existed for millennia. Lawyers have adapted to every major technology shift — from handwriting to printing, from filing cabinets to digital databases. What Junestrand is claiming, and what the evidence from large language models supports, is that the current shift is categorically different.

“No technology before has ever been able to work with language and reasoning like large language models,” Junestrand says. This is the key claim. Previous legal technology improved the logistics of legal work: faster document retrieval, better search, more organized databases. But it could not actually read a contract, understand its context, assess its implications against other documents, or draft new clauses that match a specific lawyer’s voice. Language models can now do all of these things. The nature of what a machine can do in legal work has changed fundamentally, not just incrementally.

This creates both an opportunity and, as Junestrand acknowledges, “a lot of responsibility.” Legal advice shapes outcomes that matter enormously — contracts determine business relationships, legal opinions affect regulatory compliance, court filings influence lives. Getting things wrong in this domain has real consequences for real people. Junestrand is building inside that responsibility rather than minimizing it, and that framing shapes how Legora is designed.

2. Why Claude Models for Legal Work

When Junestrand is asked directly why Legora chose Claude models, his answer is precise: nuance. “A lot of the nuance in drafting and reviewing is very context specific,” he says. “You wanna match the style of the existing document, you might wanna match the style of that lawyer in particular. And there we found the Claude models, especially Opus, extraordinarily good at the sort of fine drafting.”

This is a meaningful distinction from generic AI capability. It is not that Claude is simply better at language tasks in the abstract. It is that legal drafting requires a kind of style-matching and contextual sensitivity that goes beyond factual accuracy. A contract clause drafted for a particular client relationship should feel consistent with the rest of that agreement. A brief written for a specific judge should reflect that judge’s preferences. A memo drafted by a partner should sound like that partner’s voice, not like a generic legal document.

These are subtle requirements that previous generations of language models struggled with. The sensitivity to register, tone, formality level, and stylistic consistency that makes Claude models effective for Legora represents a genuine capability step change — not just in accuracy but in fitness for a specific professional context.

3. Building on a Rising Tide

One of the most practically useful strategic insights Junestrand shares is what he calls building boats on a rising tide. Every three months, Anthropic releases a new model with meaningfully better capabilities. For Legora, this is not a disruption to manage — it is a feature. “The very cool thing for us is that with each new model, new capabilities are unlocked,” he says. “What’s happening right now is every third month you guys come up with something new, and it’s very exciting, and it sort of raises the tide.”

The strategic implication is significant: if you build your product architecture in a way that is tightly coupled to a specific model version, you have to do significant work with every new release to capture the improvements. If instead you build abstractly — designing around what you want the system to do rather than exactly how the current model does it — each new model release automatically improves your product. Legora is designed so that better Claude means better Legora, without requiring Legora to be rebuilt.

This approach requires a particular relationship with the foundation model provider. You need enough insight into where the capabilities are heading to anticipate how to structure your product to benefit from them. Junestrand describes the Legora-Anthropic relationship as involving rich bidirectional information exchange: “this exchange of ideas, what’s working, what’s not working, where do we need to push further, that has been incredible.” The partnership is not just API access — it includes genuine collaboration on what the next generation of capabilities should be.

4. Velocity as the Core Differentiator

Junestrand’s differentiation strategy converges on a single concept: velocity. “I think when I look at differentiation, it’s about how do we make sure that Legora has the highest velocity out of any company in the industry.” This is a different kind of competitive claim than most. He is not saying Legora has the best features today, or the deepest data moat, or the most famous customers. He is saying Legora improves faster than anyone else.

The logic: if you are a law firm or legal team making a bet on an AI platform, you are not just buying a product — you are picking a partner for your AI transformation. Junestrand frames this explicitly: “if you place a bet on us and your firm or legal team goes all in on Legora, we’re going all in on you. You’re picking a software partner, an AI partner.” The commitment goes both ways.

This framing suggests that the decision of which legal AI platform to use is more like choosing a technology partner for a multi-year relationship than purchasing a SaaS subscription. It changes the criteria: the question is not just “which tool works best today” but “which team will move fastest and maintain their lead as the underlying capabilities continue to improve.” For a domain as complex and high-stakes as law, that partnership frame makes sense.

🧠 Check your understanding

5 questions · your answers are saved in this browser only

1. 1. What does Junestrand identify as the key reason large language models are transforming legal work?

   They can store and retrieve legal documents faster than databases No technology before has been able to work with language and reasoning in the way LLMs can They reduce the cost of legal services by automating billing They can access real-time legal databases automatically

   Junestrand specifically says 'no technology before has ever been able to work with language and reasoning like large language models' — distinguishing the current shift from previous legal technology improvements.

2. 2. Why does Legora specifically favor Claude models for legal drafting?

   Claude models are the cheapest to run at scale for document processing Claude has the largest legal training dataset Claude models are extraordinarily good at fine drafting that matches document style and individual lawyer voice Claude models can automatically verify legal citations

   Junestrand says legal work is 'very context specific' — you need to match the style of an existing document or a particular lawyer. Claude models, especially Opus, excel at this kind of nuanced, style-sensitive drafting.

3. 3. What does Junestrand mean by 'building boats on a rising tide'?

   Expanding Legora's services to maritime law Designing the product so that improvements in foundation models automatically improve Legora Waiting for AI to mature before building product features Building a product that works independently of any single AI model

   Junestrand describes building Legora's architecture so that every new Claude model release unlocks new capabilities for Legora automatically — the rising tide of improving models lifts the entire product.

4. 4. How does Junestrand frame the decision for a law firm choosing Legora?

   As a simple SaaS subscription that can be cancelled monthly As selecting the AI tool with the best current feature set As picking a software and AI partner for a long-term committed relationship As a risk-free trial with no switching costs

   Junestrand says 'if you place a bet on us and your firm or legal team goes all in on Legora, we're going all in on you. You're picking a software partner, an AI partner' — framing it as a mutual long-term commitment.

5. 5. What does Junestrand identify as Legora's primary competitive differentiator?

   The largest legal document training dataset The most well-known law firm partnerships Having the highest velocity — improving faster than any competitor in the industry The lowest cost per legal task

   Junestrand explicitly frames differentiation around velocity: 'how do we make sure that Legora has the highest velocity out of any company in the industry' — prioritizing speed of improvement over any static advantage.

Reset quiz

🛠️ Build it yourself

Follow these exact steps to reproduce it yourself · estimated time: ~40 min

Prerequisites

• Access to Claude or another capable language model
• A sample contract or legal document (a lease agreement or service contract works well)

Step 1 — Test style matching on an existing document

Take a two-page section of a real contract. Ask Claude to draft an additional clause “in the style of this document” for a scenario you specify (e.g., an indemnification clause for a software liability scenario). Evaluate how closely the output matches the original document’s tone, formality, and structure.

Step 2 — Simulate a legal review task

Upload a short agreement and ask Claude to identify the top three issues a client-side lawyer would flag. Compare the output against your own reading. Note where the model catches things you missed and where you catch things it missed.

Step 3 — Try a research task

Ask Claude to summarize the key considerations in a specific area of law you are unfamiliar with (e.g., “what are the main issues to consider when negotiating an indemnification clause in a software contract governed by California law”). Evaluate accuracy by checking against any reference material you can find.

Step 4 — Design the “rising tide” architecture

Sketch on paper how you would structure a legal AI product so that improvements in the underlying model automatically improve the product. What would you write once, and what would need to change with each new model?

Step 5 — Reflect on responsibility

Write two paragraphs: (1) one describing the benefits AI could bring to legal work, and (2) one describing the risks if the AI gets something wrong in a high-stakes legal context. What safeguards would you build in?