Harvey AI Raised at $11B. Here's What That Means for Small Law Firms.

Harvey just raised $200 million at an $11 billion valuation. Two months earlier, it was worth $8 billion. Before that, $5 billion in June. And $3 billion in February 2025. In one year, the company went from "promising legal AI startup" to one of the most aggressively valued private companies in enterprise software.

The round was led by Sequoia and Singapore's GIC. The numbers behind it: $190 million in annual recurring revenue, 1,000 customers, 100,000 lawyers on the platform. A majority of AmLaw 100 firms now use Harvey. Corporate legal teams at companies like Comcast and HSBC are signing on.

This is real traction. And it matters far beyond BigLaw.

Who Harvey actually serves

Harvey was built for the top of the market. Its first clients were firms like Allen & Overy, Latham & Watkins and PwC. Its use cases centre on M&A due diligence, complex contract analysis and regulatory compliance. The kind of work where partners bill at $500 or more per hour and a single deal can generate millions in fees.

Harvey does not publish pricing. Industry observers estimate per-lawyer costs in the range of $1,000 or more per month, with annual commitments and minimum seat requirements. There is no public pricing page. You talk to a sales team.

The LexisNexis partnership, announced in June 2025, added another layer. Harvey became the first generative AI platform with full access to LexisNexis primary law content and Shepard's Citations. For firms already paying for both Harvey and Lexis, dual licences could lift per-lawyer costs by 15 to 25 percent.

None of this is designed for a 12-person firm in Madrid or a 30-lawyer practice in Rotterdam.

The size gap in legal AI

Here is the structural problem. Over 90 percent of law firms have fewer than 50 lawyers. Solo practitioners account for about 40 percent of all firms. Firms with fewer than six attorneys make up more than 75 percent of the total.

But AI adoption tracks directly with firm size. According to the American Bar Association, firms with 51 or more lawyers report generative AI adoption rates of about 39 percent. Firms with 50 or fewer? Roughly half that, at around 20 percent.

The gap is not about willingness. A 2024 survey found that 58 percent of corporate legal departments expect their outside counsel to use AI, and 70 percent of clients either want their lawyers using AI or are open to it. Demand exists. The tools do not match.

The price tag of firm-ready AI systems is the primary obstacle. Smaller firms experiment with ChatGPT or free tools for drafting and research. But they cannot deploy those in client-facing workflows with privileged data. As one industry report put it: most attorneys at small firms do not trust general-purpose LLMs with client data, and redacting sensitive material adds enough friction to defeat the purpose.

What $11 billion does not solve

Harvey's valuation reflects a bet on enterprise legal becoming an AI-native industry. As venture investor Liam Bryce noted, "$190M ARR from 1,000 customers shows enterprise legal is paying top dollar for workflow integration. Professional services AI is moving from automation to embedded intelligence."

That is true. And it is a story about a very specific segment of the market.

There are meaningful questions about whether the valuation itself is sustainable. Critics point to potential ghost users: 100,000 lawyers licensed across 1,000 firms, but unclear how many use the tool daily. Others question whether $190 million in ARR is truly recurring, given per-document pricing models at some firms. Some firms are starting to build their own in-house AI tools rather than pay for a third-party platform.

But the more relevant question for the 90-plus percent of firms that Harvey does not serve is: what are your options?

The EU problem

For European firms, the question gets sharper. GDPR sets strict requirements for processing personal data. The EU AI Act, with most core compliance obligations mandatory by August 2026, adds another layer of regulation. Fines can reach 7 percent of global revenue for prohibited AI violations and 3 percent for high-risk non-compliance.

Harvey runs on top of models from OpenAI, Anthropic and Google. Its infrastructure was built for the US market first. European firms processing client data need clear answers about data residency, processing locations and controller/processor relationships. These are not abstract compliance questions. They are operational requirements that determine whether you can legally use a tool with client files.

Microsoft has expanded its EU Data Boundary for Copilot, ensuring European interactions stay on EU servers. But Harvey has not made comparable public commitments for the European mid-market. When your firm processes privileged legal data under GDPR constraints, "talk to our enterprise sales team" is not an answer that works for a 20-person practice.

What mid-market firms actually need

The work is the same at every firm size. Contract review. Document classification. Client intake processing. Due diligence on smaller transactions. Extracting terms from lease agreements, insurance policies or regulatory filings. Assembling reports from structured data. These tasks follow predictable patterns and consume disproportionate associate time at firms of every size.

What differs is the delivery model. A 25-person firm does not need a platform that orchestrates multiple LLMs across practice areas with enterprise SSO and custom workflows for 500 users. It needs a focused agent that handles one specific document workflow, deployed on infrastructure that meets EU data requirements, at a cost that reflects its actual scale.

The technology exists. GPT-4, Claude and Gemini can all handle legal document analysis at a level that would have been impossible two years ago. The missing piece is not the model. It is the implementation: connecting an LLM to a firm's specific document types, integrating with existing tools and deploying on compliant infrastructure.

Right-sized AI for the rest of the market

At Gradion, we build document automation agents for EU professional services firms. Not platforms. Not annual enterprise contracts. Specific agents for specific workflows.

A pilot takes 10 days. We identify one document workflow, build an agent that handles it and deploy it on Azure with full EU data residency. After the pilot, a 90-day roadmap covers the next workflows worth automating.

The economics are different from Harvey's model by design. A mid-market firm does not need to commit to six-figure annual contracts on a platform it might not fully use. It needs to prove that one workflow works, see the time savings and decide what comes next based on real results.

Harvey's $11 billion valuation confirms something important: AI is becoming infrastructure for legal work, not an experiment. But that headline is about the top of the market. For the firms that make up the other 90 percent, the opportunity is just as real. The tools just need to match the scale.

If your firm handles repetitive document work and you want to test what automation actually looks like, start a 10-day pilot with Gradion.