Commercial real estate adopted AI faster than anyone predicted. Lease abstraction, underwriting support, market analysis, document summaries. Then, at almost every firm, the same ritual: a human goes back and checks the AI’s work. Line by line.
Propmodo’s Franco Faraudo published a sharp piece on this yesterday, and Dealpath CEO Mike Sroka gave the behavior a name: the verification tax. “If AI is used to summarize a document, most clients still go back and verify that it is correct.” Every hour spent double-checking output is an hour AI was supposed to give back. The adoption numbers look great. The actual gains are smaller than anyone’s admitting.
And here’s the line that matters. Sroka again: “The models are already really good.” The problem isn’t the AI. It’s the foundation it’s standing on.
WHAT’S ACTUALLY HAPPENING
CRE runs on a pile of systems that don’t talk to each other. Property management in one, leases in another, comps from a third, financials in a fourth. My co-founder Chase Garbarino is quoted in the same piece (I may be biased, but he’s right): “Large real estate companies have always had a ton of different tech vendors. The question is how is AI going to sit on top of all of those?”
An AI agent has to cross four data architectures before it can even start the analysis, and one vendor quietly changing a date format breaks the whole chain. So firms verify everything, and the tax compounds.
Faraudo’s endgame is the interesting part. When models get good enough to read any system in any format, the switching costs that protected CRE software vendors for a generation start to erode. If AI can pull the same answer out of any vendor’s database, the lock-in is gone. The system of record stops being a moat.
Good. It was always a bad moat. Trapping your customers isn’t a strategy, it’s a hostage situation with an invoice.
WHY IT MATTERS
When lock-in dies, one thing is left standing: what your data can actually tell you. And most portfolios can’t answer embarrassingly basic behavioral questions.
Try this one: how many guests walked into your buildings last month?
Not badge-holders. Guests. Visitors are the most underrated behavioral signal in CRE. Every registered visitor is a client meeting, an interview, a tour, a prospect experiencing your lobby without a broker narrating it. Across our portfolio, one Park Avenue trophy tower’s visitor traffic more than tripled from February to July. A campus property roughly quadrupled its guest volume over the same stretch. And one of the busiest buildings in the portfolio stayed flat the whole time. Same six months, wildly different trajectories. That’s not a dashboard curiosity. That’s leasing intelligence: buildings that fill with guests are destinations, and destinations renew.
One more: guest traffic across our buildings runs more than twice as heavy on a Wednesday as on a Friday. If your tenants’ clients only ever see the building midweek, midweek is when your building has to perform.
Leesman built the same muscle on the workplace side: over a million structured survey responses, 70 million post-pandemic data points, one schema. That’s why it surfaces what instinct never will. Globally, only about a third of employees are satisfied with office noise levels. No lobby renovation budget catches that. Measurement does.
WHAT TO DO
Stop asking which AI to buy. Start asking what your buildings can tell it.
Pick your behavioral signals: entries, bookings, visitor registrations, service requests. Get them into one structure, portfolio-wide. That work is boring, and it’s the whole ballgame. Sroka said the quiet part: the models are ready. Most portfolios aren’t.
Then re-underwrite your software vendors. If the pitch is stickiness, they’re telling you they plan to win by trapping you. Keep the ones betting on the quality of their analysis instead.
The moat was never the software. It was always the data. AI is about to call the bluff.
