
How I built a law firm with AI agents with Helen Fan
Fringe Legal Presents Bots @ Work
Why most firms remain at levels one and two
They discuss how many companies stop at chat or workflow integrations and why that limits defensibility.
Helen Fan has been building an AI-native law firm from scratch, in public, for 50 days. Not a demo. Not a prototype. A real practice with AI agents handling legal strategy and research, and she's documenting every stumble along the way.
In this episode, we get into what that actually looks like: the agents, the arguments, the security concerns, and the hard questions about whether AI-native firms and traditional law firms are on a collision course or just running separate races.
In this episode:
- How Helen built OpenClaw Law LLP with two AI agents, Morgan and Cleo, and why agent-to-agent argument reports matter more than most people realise
- The practical pain of open-source agent frameworks: stability issues, setup overhead, and the security surface that opens up the moment you connect an agent to real systems
- The Legal AI Value Stack — five levels of AI maturity in law, and why most firms are still stuck at the bottom two
- How Big Law and AI-native firms are competing on entirely different timelines, and why the boutique model might be the one that actually moves
- What Helen is telling firms just starting out: start with mindset, build the orchestration layer, and don't skip workflow integration
Timestamps:
00:00 — Helen's OpenClaw Law LLP experiment: what an AI-native law firm looks like in practice
02:20 — Agent-to-agent communication and argument reports: why they reduce hallucinations
04:10 — Stability, troubleshooting, and the real cost of open-source frameworks
07:00 — The Legal AI Value Stack: five levels of AI maturity and where the moats actually are
12:00 — Why most firms are still at level one or two
16:00 — Proprietary data and the scaling wall
19:00 — Big Law vs AI-native firms: speed, trust, and structural barriers
23:00 — Guardrails, verification, and building an orchestration layer that holds
34:00 — Vendor moats, platform plays, and what M&A in legal AI actually looks like
39:00 — What makes a law firm genuinely AI-native, not just AI-curious
43:00 — Where to start: mindset, workflow, and infrastructure
44:50 — Final thoughts on the pace of change and what to watch next
Resources:
- Helen Fan on LinkedIn — follow her 100 Days of AI Law experiment as it unfolds
- The Legal AI Value Stack — Helen's five-stage framework, published on her Substack, Helen's Legal AI Lab
- OpenClaw Framework — the open-source multi-agent system Helen built OpenClaw Law LLP on
- Claude AI — the AI tool Helen uses for legal and strategic work
- AI-Native Law Firm Index — a running list of AI-native law firms referenced in the episode


