The incumbent stack is sharecroppers vs. landowners now.

Here's the structural problem with the way most CRE technology gets sold to brokerages: the data lives in someone else's house.

When you pay CoStar, or any of the other big incumbent platforms, you're not buying software. You're renting access to a database. The data they have on your market, the comps your firm needs, the property research your associates depend on: it's not yours. Stop paying and it disappears. Try to export it and you find out about the contractual handcuffs.

Sharecroppers, not landowners

It's a sharecropping arrangement. You do the work, touring, negotiating, closing, and a third party owns the soil you're working on. They take a meaningful percentage of your overhead in exchange. They decide what tools you have and what tools you don't. They decide what the data costs next year.

The broker who closed a deal at 1500 Market last quarter doesn't own the comp record from that deal. CoStar does. If that broker wants to use that comp to pitch a new client next year, they're paying for the privilege of accessing their own work.

AI-native flips the leverage

What's changed in the last 18 months is that the AI-native CRE stack, DealDesk and a handful of others, treats the broker's data as the broker's data. The lease abstracts you generate are yours. The comps your team curates are yours. The pipeline activity, the LOIs, the email threads, the deal context, all yours. Export anytime, in any format, no contractual lock-in.

The model isn't "we're the database vendor and you're our subscriber." The model is "we're the working software, you own the data, we charge for the tools that operate on it." That's a fundamentally different power dynamic, and it's the dynamic that every modern SaaS category eventually settles on. CRE is just late.

Why now

The reason this transition wasn't possible until recently is that the data was hard to generate. You couldn't replace CoStar's comps database without a way to ingest comps at scale, and ingesting comps at scale required either (a) thousands of researchers manually transcribing deals or (b) a model that could read a deal closing email and structure it. (a) was the incumbent's moat. (b) is what 4.x-class models unlocked.

Once you can structure your firm's deal flow automatically, the math changes. Every closed deal becomes a comp. Every abstracted lease becomes a benchmark. Every LOI becomes a data point on what the market actually pays. After a year, your firm has a comps database that's specifically tuned to the deals your brokers actually do. That's more valuable than a generic database that covers the whole country at low resolution.

What we'd watch for

If you're a head of brokerage evaluating CRE tech right now, the test is simple. Ask each vendor: "if I cancel, what happens to the data my team generated using your tool?" If the answer is anything other than "you keep all of it, in any format you want, no charge," you're being asked to sharecrop. The deal will look fine on year one. It'll look terrible on year five.