Harvey, the AI platform for law firms, just launched an agent builder — a tool that lets firms create custom AI agents to run specific legal workflows autonomously. Legora, its nearest rival, is now valued at $5.55 billion after its latest round, has 800 law firm customers, and is expanding into the US from its Scandinavian base. Both companies are on nearly identical revenue trajectories. Both are racing to be the platform that defines what AI-assisted legal practice looks like. And both are making a quiet but significant transition: from tools that help lawyers do research faster to systems that can run parts of a legal matter on their own.
The research assistant phase of legal AI is largely behind us. The early value proposition — ask an AI to summarise case law, draft a first pass at a contract clause, or surface relevant precedents — proved real and adoption spread faster than most of the legal industry expected. Harvey surpassed $100 million in annual recurring revenue, a milestone it hit ahead of schedule, and its customer base now includes some of the largest firms in the world. Legora's 800 law firm figure represents a meaningful fraction of the global legal services market for a company that was barely known outside Europe eighteen months ago.
What both companies are now building is different in kind. Harvey's agent builder lets firms define a workflow — say, the standard due diligence process for an M&A transaction, or the intake and triage steps for a litigation matter — and assign AI agents to execute it. The agents can pull documents, run searches, draft outputs, and flag items for human review without a lawyer directing each step. Legora's platform is structured similarly, with an emphasis on collaborative workflows where AI and human lawyers work in parallel rather than in sequence.
The implications for legal practice are meaningful and contested within the industry. Law firms bill by the hour, and a technology that makes their work faster is, on a simple analysis, a technology that reduces their revenue. The standard counterargument is that efficiency allows firms to take on more matters, expand into work that was previously uneconomical, and compete for clients who would otherwise go to in-house teams. That argument has historical precedent — every previous wave of legal technology, from Westlaw to e-discovery software, was greeted with similar concerns and ultimately proved net-positive for firm revenues at the top end while commoditising the bottom.
What's different this time is the scope. Previous legal technology automated discrete tasks — document review, citation checking — while leaving the judgment-intensive work entirely to humans. The agent systems Harvey and Legora are building are attempting to automate entire process flows, including some that involve genuine legal judgment: flagging unusual contract terms, assessing litigation risk, identifying the relevant standard of care in a negligence matter. These are not binary tasks with objectively correct answers. They are the tasks for which lawyers charge premium rates and bear professional responsibility. Whether AI can do them reliably enough to satisfy malpractice standards is the question the industry is still working out — in practice, through deployment, as much as through deliberate policy.
The race between Harvey and Legora is also a bet on geography. Harvey is US-centric and pushing into Europe. Legora is European and pushing into the US. Legal practice is highly jurisdiction-specific, and the firm that builds the better model for each jurisdiction's specific documents, precedents, and professional standards will have a durable advantage. That suggests the endgame is not one winner but a landscape of deeply specialised platforms — which is, perhaps, an appropriate outcome for an industry that has always been fragmented by jurisdiction, specialisation, and culture.