From Legal Copilot to Knowledge Superapp

Harvey is trying to turn trust earned in legal into a broader control layer for high stakes office work. Legal is a good starting point because lawyers handle sensitive documents, strict permissions, and source backed outputs, so the product has to get security, auditability, and workflow fit right before it can spread. Harvey has already moved beyond a single legal model into a multi model workflow system, which is the kind of architecture that can travel into adjacent knowledge work.

• The product is becoming more like a work surface than a chatbot. In legal, that means Word add ins, document management integrations with iManage and NetDocuments, and custom agents that fit into existing drafting and review flows, rather than asking lawyers to leave their normal tools.
  1 sacra 5 sacra
• The main lesson from legal is that regulated work rewards context and permissions, not just reasoning. Across vertical AI, durable value comes from wiring in repositories, access controls, templates, tool calls, and human review loops, because public data and generic chat features are quickly copied by horizontal model companies.
  3 sacra 4 sacra
• Harvey is also running into the limits of being a general layer too early. In legal alone, the market is fragmenting into contract tools like Spellbook, search infrastructure like DeepJudge, services led players like Crosby, and incumbents like Clio that own billing, practice management, and now legal research data.
  2 sacra 3 sacra 6 sacra

The path to a knowledge worker super app runs through owning the context layer across more teams, not through a bigger chat box. If Harvey can carry its security model, workflow integrations, and evaluation loops from law into other document heavy functions, it can expand from legal copilot into the operating layer where knowledge work actually gets done.