Startup build · Legal · Solo and small-firm attorneys · Technical co-founder
NextChair AI
Legal research and drafts that the attorney checks and signs off
6 stages
from intake to attorney sign-off
50 states
plus DC and 13 federal circuits mapped
3 versions
rebuilt as we learned
Before
Solo attorneys and small firms don't have associates to research and draft. General AI tools make up citations, which is a nonstarter in court.
After
An attorney opens a matter and gets researched, drafted work built on real case law, a critique of its weak spots, and a verification checklist. Nothing is final until the attorney signs off.
What we built
Memos, motions, briefs, demand letters, and contracts
Research against real case law, scored for authority, recency, and jurisdiction
A built-in adversarial review that argues against the draft
Required attorney verification, with an audit trail
Five specialized agents (intake, research, drafting, adversarial review, and verification) orchestrated with LangGraph, using Claude for reasoning and Gemini for parsing.
Several layers of anti-hallucination checks verify citations before anything reaches the attorney.
Python backend, a React workspace with live streaming, and Convex for state. Three major architecture rewrites.
