Harvey workflow-first legal AI

Harvey is turning implementation from a model building project into a workflow rollout motion, which makes premium legal AI much easier to scale. Instead of retraining a custom legal model for each new use case, Harvey can package repeatable flows for drafting, research, and review, then keep the expensive human layer focused on onboarding, change management, and getting lawyers to use the product often enough for renewals.

• The product shift is concrete. Harvey moved away from its proprietary fine tuned model after frontier models outperformed it, and now chains different models and tools for different steps, like document analysis, legal research, and drafting.
  1 sacra 2 sacra
• The service layer still matters. Roughly 10% of the team is dedicated to ex lawyers in forward deployed customer success roles, which keeps Harvey close to firm workflows without requiring the slower, more custom work of training bespoke models.
  1 sacra 4 sacra
• This puts Harvey in a broader legal AI race where the winners are owning repeatable workflows, not just raw model quality. Legora went straight to off the shelf models, Spellbook focuses on contract work inside familiar tools, and Clio is bundling AI into software it already owns for smaller firms.
  2 sacra 3 sacra 5 sacra 6 sacra

Going forward, the highest value legal AI companies will look less like model labs and more like workflow companies with domain specific services attached. Harvey is well positioned if it keeps turning bespoke legal tasks into standard product modules, while reserving its human touch for adoption, governance, and expansion into more daily legal work.