The More Capable AI Gets, the More Licenses Look Like Moats

Building a competitor prototype with Claude might take an afternoon. The difficulty in replicating a company that requires a license isn’t in the code.

SIP’s tweet lays out a path: choose a regulated business; if you’re not a lawyer, find a lawyer as a partner; you handle marketing, growth, and AI work, while the lawyer runs the company. The tweet gives the example that in some countries only lawyers can own law firms. That way, competitors can’t build the same thing with Claude in a day.

The point of this approach is access. AI lowers execution costs—and lowers competitors’ execution costs too. Licenses, compliance liability, and relationships with licensed operators push the cost of replication back up. For developers, beyond model capabilities, industry access and relationship capital are worth watching.

In a reply, Gaurav Bhowmick adds another layer: a license is a moat, and so is trust. Some people only let people they know into their homes; that kind of trust may become more valuable over time. He didn’t answer which is more durable.

Dylan Normandin’s example is closer to reality. He runs a physical therapy (PT) business in California, where regulation is heavy and margins are thin. His approach is to be AI-native, own the full stack himself, and keep operating costs low. Regulation both protects and restricts; a license doesn’t guarantee profitability.

This approach isn’t meant to be copied wholesale. Partnership structure, scope of practice, and liability allocation all need to be confirmed by local lawyers and compliance professionals. AI can write marketing copy, qualify leads, and run automation, but it can’t replace a licensed person’s signature.

If you’re looking for an AI startup direction, ask the reverse question: which businesses are bottlenecked by licensing and trust, not by writing code? AI’s value may be higher there—and harder to replicate overnight.