Harvey Improves Legal Drafts by Remembering How the Partner Likes the Numbering
Richer context and remembered preferences are part of Harvey’s latest drafting work. The substance still needs a lawyer, even when the indentation behaves.
Legal Tech · · 6 views

A useful legal assistant must remember two different things: the matter being worked on and the way the reviewing lawyer wants it presented.
Harvey's OpenAI case study describes stronger drafting with GPT-6 Astra through richer context, including preferences and formatting requirements. Professional judgment remains part of the workflow.
That sounds mundane beside grand promises about transforming law. It is also the sort of improvement that can prevent a qualified professional from spending another billable interval correcting list styles.
I take presentation seriously. The numbering should not become the main subject of a document drafted to address something more expensive.
But a well-organized error is still an error. Remembering that the partner likes a particular layout cannot establish the soundness of the argument within it. A review that only admires the layout has stopped early.
The useful result is a draft that lets the lawyer reach the substantive questions sooner. There will still be comments in the margin. With luck, more of them can concern the client's problem and fewer can concern the tab key.
Based on: Harvey turns legal context into stronger drafts with GPT-6 Astra, OpenAI, September 23, 2026.