Five levels of defensibility in legal AI — read by 500K+ on LinkedIn.
Chat interfaces on foundation-model APIs. When a model does most of the work for free, the moat is shallow.
Legal workflows, playbooks and structured outputs. The gap between raw API and finished product is closing fast.
Platforms that get smarter as more clients use them, learning patterns no foundation model can see.
Products so embedded in daily practice that replacing them is unthinkable. The moat is operational gravity.
Not selling software to law firms — becoming one. When intelligence is a commodity, the scarce resource is trust.
Levels 1–4 are converging toward infrastructure. Level 5 is the only level where value compounds rather than erodes.
Read the full V1 essay on SubstackIf you are here, you have not started.
One lawyer as chief of staff: leading multiple agents, designing the loops, auditing the outcomes.
The system learns from your work — inside the agent, client and team boundaries.
A unified AI operating layer the whole practice runs on.
AI handles intake, research and the first draft. Lawyers own trust, judgment and accountability.
Start with one workflow: build a data layer that learns from it, set clear boundaries, and put someone in charge of making it work.
Read the full V2 essay on Substack