TwinLadder Weekly

Issue #1 | February 2025

---

Editor's Note

I have spent twenty years watching legal technology vendors promise transformation. Most delivered incremental improvement at premium prices. So when Harvey announced its $300 million Series D at a $3 billion valuation, my first instinct was not excitement. It was a question: transformation for whom?

Last month, a managing partner at a 40-lawyer firm in Riga asked me whether his practice was falling behind because he could not afford Harvey. I told him the truth: Harvey was not built for him. That is not a criticism of Harvey. It is a market reality that the funding headlines obscure.

This newsletter exists because the legal AI conversation is dominated by voices selling to the Am Law 50. The rest of us — and that includes most European practitioners — deserve honest analysis, not venture capital press releases. We will cover what works, what does not, and what the gap between the marketing and the reality means for professionals building real practices at real budgets.

---

Harvey's $300M: Who This Money Actually Serves

Harvey's metrics are real. $50M+ ARR with a path to $100M within eight months. 235 customers across 42 countries, up from 40 in early 2024. 4x revenue growth year-over-year. The majority of the top 10 US law firms as customers. Sequoia led the round. Kleiner Perkins and Coatue participated. LexisNexis parent RELX joined. These are serious institutions making serious bets.

CEO Winston Weinberg told Fortune the company is building "the defining professional services AI company." Given the numbers, it is hard to argue with the ambition.

But Harvey's $3 billion valuation on $50M revenue represents a 60x revenue multiple. That is venture capital theatre — a bet on future dominance, not current value. And the estimated pricing tells you who funds that bet.

Harvey Economics	Mid-Market Reality
$1,200 per lawyer per month	Average European mid-market IT budget: $200-400 per lawyer per month total
20-seat minimum, 12-month commitment	A 50-lawyer firm pays $720,000/year
$288,000 minimum annual commitment	Most European firms spend less on all technology combined
60x revenue multiple valuation	Built for firms writing seven-figure technology cheques

Do the maths. A minimum annual commitment of $288,000. A 50-lawyer firm pays $720,000 per year. A large firm at 200 lawyers faces $2.88 million annually. A 15-lawyer insurance defence practice in Riga, or Rotterdam, or Lyon? Not the customer Harvey was built for — and not out of malice. They are simply not the market that justifies those economics.

This matters because the funding narrative conflates two different things. It proves massive demand for legal AI among firms that write seven-figure technology cheques. It says nothing about whether the other 95% of the profession will access these capabilities at sensible prices.

The mid-market landscape in early 2025 offers alternatives. CoCounsel at $110-400 per seat with flexible commitment. Lexis+ AI at $99-250 per feature, though costs compound across features. General-purpose tools like ChatGPT and Claude at $20-25 per user, powerful but lacking legal guardrails. Clio Duo coming soon with practice management integration. None are Harvey-equivalent. You can assemble point solutions or use general-purpose AI with careful prompting. Neither option is elegant. Neither delivers the seamless, purpose-built experience Harvey offers its enterprise clients.

The honest assessment: the technology Harvey is building will eventually become more accessible. That is how every enterprise technology evolves — expensive, then affordable, then commodity. Mainframes became laptops. Enterprise CRM became Salesforce became free tiers. Legal AI will follow the same arc. But "eventually" does not help the firm making purchasing decisions this quarter.

What does help is understanding the distinction Harvey's funding actually illuminates. The Am Law elite are buying capability — they can afford to experiment, fail, iterate, and absorb the cost of tools that do not yet justify their price through productivity alone. The European mid-market needs to buy outcomes. Different purchasing logic, different timelines, and nothing wrong with either. The mistake is pretending they are the same decision.

For European practitioners, there is an additional dimension the American coverage entirely ignores. The EU AI Act entered its first compliance phase on February 2, 2025, with Article 4 requiring AI literacy for all staff deploying AI systems. Harvey's sales team will not help you with that obligation. Neither will Sequoia's investment thesis. The firms that build genuine AI competence — understanding what these tools can and cannot do, developing verification workflows, training their people — will be better positioned to adopt purpose-built tools when the economics make sense. The firms that wait for a vendor to solve everything will wait a long time.

The question is not whether AI will transform your practice. It is whether you will have built the judgment and workflows to deploy it intelligently when the right tools arrive at a price you can afford.

---

The Competence Question

A partner I know recently hired a lateral associate from a firm with full Harvey deployment. The associate's briefs were flawless. Every citation verified. The research was thorough, well-organised, efficiently produced. But when opposing counsel raised an argument the brief had not anticipated, she had no instinct for how to respond. She had never had to develop one. Harvey always gave her a starting point.

This is the training problem hiding inside the adoption story. If your associates learn legal reasoning through AI-mediated research from day one, what happens to the instincts that come from struggling through a problem manually? The fumbling, the wrong turns, the moment when you finally understand why a line of cases matters and how they connect — that is where professional judgment forms.

I am not arguing against AI adoption. I am arguing that firms rushing to deploy tools like Harvey need to think simultaneously about how they are developing the lawyers who use them. When your associate cannot draft a motion without Claude, that is not an adoption success story. It is a training failure. The tool amplified her output without deepening her understanding. She became faster without becoming better.

This is not a problem unique to Harvey or to any particular tool. It is a structural problem with how we think about deployment. Technology that makes a skilled lawyer more productive is valuable. Technology that prevents a junior lawyer from becoming skilled is dangerous. Both can be the same tool. The difference is in how it is introduced and supervised. At a firm in Tallinn last autumn, I watched a senior partner require his associates to draft their research memo by hand before running the same query through an AI tool. The discrepancies — what the human caught that the machine missed, and vice versa — became the basis for a weekly training discussion. That is what thoughtful deployment looks like.

---

What To Do

1. Experiment with general-purpose AI on non-confidential work. ChatGPT and Claude cost $20-25 per month. Use them to build intuition for what these tools can and cannot do. The learning is in the gaps — the moments where the AI confidently produces something wrong. Budget nothing but time. The education is worth more than most vendor demos.

2. Audit your workflows before shopping for tools. Know exactly which tasks consume disproportionate hours. Buy technology second, build processes first. A firm that knows its bottlenecks makes better purchasing decisions than a firm chasing headlines. A 30-lawyer corporate practice in Amsterdam told me they mapped every associate hour for a month before evaluating any AI product — and discovered their biggest time sink was not research but internal communication.

3. Start a verification habit now. Every AI-assisted output gets human review of citations before filing. No exceptions. Stanford's research shows even the best legal AI tools hallucinate roughly one in six queries. That number should frame every conversation about trust.

4. Protect your associates' learning. Designate some work as "AI-free" — not because the tools are not useful, but because struggle is how lawyers develop judgment. The associate who has wrestled with a difficult research problem manually will use AI tools with far more discernment than one who has never had to.

5. Understand your Article 4 obligations. EU AI Act Article 4 took effect February 2, 2025. If your staff deploy or operate AI systems, you need documented AI literacy training. This is not optional. Start by identifying which staff use AI tools and what training they have received. The compliance clock is already running.

---

Quick Reads

• 79% of legal professionals now report using AI tools in some capacity — but most usage is basic summarisation and drafting assistance. The gap between "I use ChatGPT sometimes" and "I have an integrated AI workflow" is enormous. Adoption breadth without sophistication depth.

• LexisNexis parent RELX joined Harvey's round as an investor. Watch for integration announcements that could reshape the research platform landscape and create new competitive dynamics with Thomson Reuters. For European users, the question is whether Lexis+ AI will close the gap with Harvey at a fraction of the price.

• Stanford RegLab's ongoing research documents hallucination rates across legal AI: ~17% for Lexis+ AI, ~34% for Westlaw AI, ~69% for general-purpose GPT-4. Next issue examines what these numbers mean for your daily practice.

• Harvey alternatives for smaller firms — a practical roundup if you are evaluating what is available below the enterprise pricing tier. Most alternatives lack Harvey's depth but offer specific capabilities at one-tenth the cost.

---

One Question

If Harvey's pricing means only the largest firms get the best AI, does legal AI widen the access-to-justice gap — or does the technology inevitably democratise, as every previous enterprise technology has? And while we wait for that democratisation, what are the rest of us supposed to do?

---

TwinLadder Weekly | Issue #1 | February 2025

Helping European professionals build AI competence through honest education.