The Marketplaces of the Future

From Zillow to Airbnb to LoopNet, some of the best real estate technology businesses to emerge in recent decades have been marketplace models. By connecting buyers and sellers in a notoriously fragmented and opaque industry, these marketplaces have proven durable and remarkably profitable.

But they’re also notoriously hard to build, requiring buyers and sellers to come together in the right numbers at the right time. Opportunities to build great marketplaces tend to emerge at times of major technological change or dislocation when market players are figuring out new ways to discover each other.

As the real estate industry enters its fifth year of elevated interest rates and AI increasingly reshapes how the industry operates, it feels as if we’re on the cusp of such a moment. There has never been a better time to build a new marketplace.

Today’s letter coves six marketplace models that are likely to emerge over the next decade. It also covers four oft-proposed models that I believe are unlikely to gain real traction, and why.

Marketplace #1: The Agent Marketplace

Increasingly, "technology" in real estate means autonomous AI agents. 

But it makes little sense for every operator to build its own agents from scratch. Buying a package of prebuilt agents off the shelf is equally problematic in a world where every operator runs its own workflows and every asset is unique. An underwriting agent trained on a merchant builder's deal flow is useless to a core-plus fund; a leasing agent tuned for Class A multifamily fails the moment it touches a value-add portfolio.

The opening is a marketplace that adapts libraries of prebuilt agents to an operator's specific data model and workflow. This could look something like Outcome or Runy, companies doing exactly that today. Platforms like Kore already run this playbook outside of real estate, and Venn's Zapier-like build-your-own-workflow tool could be an opening to the agent marketplace from within it.

The winner here won't sell agents but rather the translation layer between a generic agent and a particular operator's messy, idiosyncratic reality.