I've been spending a lot of time recently thinking about what kinds of startups make sense to build right now. Recently I've been thinking a bit about a subset which you might call "AI market makers". A lot of great businesses could be thought of as market makers, meaning, they reduce friction in some market and facilitate exchanges at scale.

I prefer the "market maker" framing to the "two-sided marketplace" framing, specifically because the scaled business requires creating a market where there previously was none by lowering the friction required for some exchange to take place. This seems like a natural framing for AI agents, and the resulting business model is easy to model (just take some percentage from each exchange).

There are already category-defining companies in most of the above markets, and I'm sure most of those companies are thinking about adding some version of an AI agent to their existing platforms, but I think AI agents will open up a number of new markets that were previously not possible, and these new markets will result in generational companies being built. Put differently, if the market already exists and has some dominant brokerage today, there's probably already a bunch of other people thinking about how to make it more efficient by sprinkling some AI on top, but I think there are markets still waiting to be created.

• Selling used items: I've recently been selling a bunch of stuff on Facebook Marketplace, and wish that I could have an AI agent helping me do it. A lot of money changes hands every year for used goods, and it's currently done pretty inefficiently. There are a few points of friction, mostly on the buyer's side, which AI agents could probably solve:
  • Taking photos of my item
  • Curating a listing
  • Identifying a fair market price
  • Fielding customer requests
• Interacting with the government: This is not as clear to me, but the rough shape of the idea is some agent that can interact with government services at scale, then broker access to those services with customers who are willing to pay a bit more for a faster or lower-friction version of that service. It would be cool to train an RL agent to more successfully get permitting approval, for example.

I'll add some more examples as I think about them. If you have ideas that seem to fit this theme and that don't already define the business model of some large, successful company, please send me a note!