TwinLadder Weekly

Issue #18 | October 2025

Editor's Note

This is the fifth time we have covered Harvey in this newsletter. I considered skipping it — another funding round, another valuation headline. But $8 billion demands a different conversation than valuation math.

Here is what struck me this week. I was at a conference in Brussels where a managing partner of a well-regarded 80-lawyer firm told me, with genuine anxiety, that she "still does not know what her AI strategy is." The same week, Harvey confirmed an $8 billion valuation after raising $150 million from Andreessen Horowitz. Three rounds, $750 million, in a single year.

These two facts — one company worth $8 billion, and a capable managing partner unable to define her firm's approach — capture the real story of legal AI in 2025. It is not a technology story. It is a gap story. And for those of us working in Europe, where the majority of firms look more like that 80-lawyer Brussels practice than like an AmLaw 50 powerhouse, the gap feels more like a chasm.

[HIGH CONFIDENCE]

The $8 Billion Gap

Let us dispense with the valuation analysis. Harvey trades at 41-80x revenue depending on which ARR estimate you use. Mature SaaS companies trade at 5-10x. That is either visionary pricing or venture capital exuberance, and the honest answer is that nobody knows yet which. I am not going to pretend otherwise, and neither should you.

The more important number is this: 50 of the top Am Law 100 firms are Harvey customers. So are KKR, Bayer, Comcast, and Deutsche Telekom. 500+ customers across 54 countries. Weekly active users grew 4x year-over-year. Monthly queries grew 5.5x. Active files stored went from 268,000 to 9.75 million. These are not vanity metrics. They indicate genuine usage at scale.

Harvey's Position	The Rest of the Market
$8B valuation (41-80x revenue)	Mature legal SaaS trades at 5-10x
$1,200/seat/month enterprise pricing	CoCounsel: $110-400/month; ChatGPT: $20-25/month
50+ AmLaw 100 firms as customers	~400,000 law firms worldwide have no comparable option
$750M raised in a single year (2025)	Most legal tech startups raise $5-20M total
54 countries, 500+ customers	European adoption concentrated in Magic Circle and top-tier

But here is what those numbers actually tell us: the legal AI market has priced itself for the top of the profession and left everyone else watching.

Harvey at $1,200 per seat per month. CoCounsel at $110-400. General-purpose AI at $20-25 but without legal specialisation. For the 80-lawyer firm in Brussels, there is no obvious path. Too sophisticated for ChatGPT. Too small for Harvey's enterprise model. Too practical to wait.

I have spent twenty years advising firms on technology decisions, and this is the first time I have seen a market where the leading product is genuinely excellent and genuinely inaccessible to the majority of practitioners who need it. Harvey proved that elite firms will pay premium prices for legal AI — reaching $100M ARR in just three years. They did not prove that this model serves the profession.

The European lens matters here, and it is consistently missing from the conversation. The American lens dominates this discourse. In Europe, the dynamics are different. Our firms tend to be smaller. Our margins are tighter. Our regulatory environment, with the EU AI Act entering full enforcement in August 2026, adds compliance costs that American firms do not face. A $8 billion American legal AI company is impressive. It is also irrelevant to most of the lawyers I know.

That is not Harvey's fault. They built what the market rewarded. The venture capital thesis — articulated by Marc Andreessen himself — is that legal is a $700B+ industry with low technology penetration, and the winner-take-most dynamics of enterprise software favour dominant players. Maybe. But the legal profession is not one market. It is thousands of markets, stratified by size, jurisdiction, specialisation, and client base. Harvey is winning in one of them.

The question for the rest of us is not "how do I compete with Harvey?" It is "how do I build AI capability that serves my clients at my scale?" And that question, worth far more than $8 billion to the profession, has no well-funded startup racing to answer it.

The Article 4 dimension that nobody in Silicon Valley is discussing. The EU AI Act's Article 4 requires documented AI literacy for anyone deploying AI systems. This obligation entered force in February 2025. It applies to the managing partner in Brussels as much as to Clifford Chance. But here is the irony: the firms that can most easily comply — the ones with training budgets, innovation teams, and dedicated AI governance — are the same firms that can afford Harvey. The firms that struggle with compliance are the ones that also struggle with access. Article 4 creates a mandate without creating a pathway for the majority to meet it. That is where TwinLadder's work begins.

I want to be fair to Harvey's founders. Winston Weinberg, a former securities and antitrust litigator, and Gabriel Pereyra, a research scientist from DeepMind and Meta, built something real. Multi-model architecture integrating OpenAI, Anthropic, and Google. Purpose-built features for litigation, transactional work, and regulatory compliance. SOC 2 security posture. This is not a ChatGPT wrapper. It is genuine enterprise legal infrastructure.

But that is precisely the point. The most capable legal AI platform in existence was built for — and priced for — organisations that already have every advantage.

I have seen this pattern before, and it does not end the way investors expect. When I was at A.T. Kearney, we were hired by EDS — the company Ross Perot built, the company that invented IT outsourcing — to catalogue their innovations. They had built the first roboticised factory for General Motors, created OnStar, pushed boundaries that nobody else touched. By the time we arrived, almost everyone who had built those things was gone. EDS had grown to 100,000 employees by codifying everything into neat operational categories — mainframe, midframe, desktop — optimised for delivery and margins. There was no place left for the people who did work that did not fit the categories. When clients came asking "what makes you special?", the answer was: nothing anymore. They had become a commodity competing on price.

Harvey's $8 billion is a bet that optimising what elite firms already do is worth venture-scale returns. Perhaps. But the firms that most need AI to compete are the ones least able to afford it. That is not a market failure. It is a market working exactly as designed for venture returns. The question is whether it works for the profession.

The Competence Question

Consider this scenario. A mid-market litigation firm receives an RFP from a financial services client. The RFP includes a new section: "Describe your firm's AI capabilities, governance framework, and how AI tools are integrated into your service delivery."

Your partners look at each other. You use ChatGPT informally. A few associates have Claude subscriptions. Nobody has thought about governance. There is no policy, no framework, no story to tell.

You lose the RFP. Not because your legal work is inferior — it is not. But because the client has been told by Harvey-equipped competitors that AI capability is a differentiator, and you could not demonstrate yours.

This is not hypothetical. In-house departments are already including AI questions in outside counsel guidelines. I have seen three such RFPs in the past six months, from European financial institutions and multinationals based in Frankfurt, Amsterdam, and London. The questions are getting more specific: not just "do you use AI?" but "what is your AI governance framework?" and "how do you ensure quality control for AI-assisted work product?" Some are now asking about Article 4 compliance directly.

Your competence is not just about whether you can use AI tools. It is about whether you can articulate your approach — or explain, credibly, why your approach differs. A firm that says "we have evaluated AI tools, determined that for our practice areas the verification burden currently exceeds the efficiency gains, and instead invest in structured knowledge management" has a position. A firm that says nothing has a problem. Silence is not a strategy.

What To Do

Write your AI story. Even if it is modest, document what tools you use, how you verify outputs, and what governance you have in place. One page is enough. Having nothing is a liability — both for RFPs and, increasingly, for Article 4 compliance.
Stop benchmarking against Harvey. Your practice does not need an $8 billion platform. Identify the three workflows where AI would save the most time and find tools that address those specifically, at your price point. A EUR 25/month Claude subscription with a well-designed verification protocol may outperform an unused enterprise licence.
Talk to your clients. Ask what they expect regarding AI use in their matters. Proactive disclosure builds trust. Reactive disclosure creates suspicion. European clients, aware of the AI Act, are increasingly expecting this conversation.
Budget for 2026 now. AI competence costs will increase. Whether it is CLE, tool subscriptions, or training time, plan the investment rather than scrambling when it becomes mandatory. For European firms, Article 4 compliance is not future planning — it is current obligation.
Build your AI governance framework this quarter. Document your approach to AI tool selection, usage policies, verification procedures, and staff training. This framework serves triple duty: client confidence, regulatory compliance, and internal quality assurance. You do not need Harvey to have governance. You need governance regardless of which tools you use.

Quick Reads

Harvey confirms $8B valuation after $150M Andreessen Horowitz-led round — third major round of 2025.
Stanford HAI testing found 17-58% hallucination rates even in purpose-built legal AI tools, underscoring that premium pricing does not eliminate verification obligations.
Harvey reached $100M ARR in August 2025, three years after launch — the fastest revenue growth in legal technology history.
The Bloomberg Law analysis asks the $8 billion question directly: can AI startup match its hype? The answer depends on whether the market they serve is the profession or just the top of it.

One Question

If the most valuable legal technology company in history serves primarily the top 50 firms, who is building for the other 400,000?

TwinLadder Weekly | Issue #18 | October 2025

Compliance is the floor. Competence is the mission.