Harvey, the artificial intelligence company that builds tools for large law firms, raised $80 million at a reported $715 million valuation, according to TechCrunch, in a round that values a company built on top of other companies' models.

The financing is early evidence that the first durable AI winners may be vertical companies rather than the labs that build the underlying models. Harvey's product wraps OpenAI's models in the specific workflows, document formats and citation practices that law requires, and its customers are firms that would not adopt a general-purpose assistant for work that carries legal liability.

The platform risk

A vertical AI company that depends on a single model provider carries a platform risk that the general labs do not. Harvey's defence is the workflow integration and the legal-specific training data, both of which are hard to replicate but neither of which fully insulates it from a model provider that decides to compete. The vertical thesis will be tested the first time a lab launches a competing product.