Bisnow ran a piece this weekend on AI in CRE underwriting. Worth reading. The article is careful about what it claims. It doesn’t say AI does full underwriting. Crexi’s Siegel compares the tools to robotic surgery and says “we may not be quite there yet.” The Anchin partner wouldn’t give AI $20M to invest. The piece is really selling one thing. Speed.

That’s the frame I want to push on.

Start with what’s actually happening when an analyst transcribes a T12 by hand.

They notice a huge water spike in June. Is that a leak or seasonal irrigation? They notice bad debt moved out of income and into expenses, making the income line look cleaner than it is. They notice reversals and accruals scattered across the last four months, which tells you something about how tightly the accounting is run. None of that is in the executive summary of the OM. It lives in the act of moving numbers from one column to another.

Could AI flag those patterns? Yes. Primer and Blooma can catch the water spike and the bad debt reclassification. They’ll catch more as the tools improve.

But flagging isn’t the same as noticing.

The junior analyst who sits with the data for three hours isn’t just checking math. They’re building a mental model of the property. They’re asking why, not just what. They start noticing things that no flagging system would catch, because the things aren’t anomalies. They’re textures. How clean is the accounting? How many times has the PM reclassified the same expense? Does the T3 annualization hide a softening trend?

That’s where underwriters get made. It’s also the part getting automated first.

If the calculator does it for you, you stop looking at the numbers. That’s human nature. It’s not an argument against calculators. It’s an argument against pretending the calculator gave you everything that was in the raw data.

The Bisnow article celebrates faster underwriting as unambiguously good. I don’t buy it.

The deeper problem. The article assumes the bottleneck in CRE investing is underwriting speed. It isn’t. Not for disciplined buyers.

There are a finite number of deals that fit any given buy box. If you’re a value-add multifamily buyer in the Southeast, the universe of deals that actually match your criteria in any quarter is small. Maybe three. Maybe five. The real constraint isn’t how fast you can underwrite. It’s whether you saw every deal that could have fit.

That’s a sourcing problem. Broker relationships. Off-market access. Capital reputation. Speed of response when the right deal hits. None of that gets solved by parsing an OM faster.

A disciplined buyer doesn’t need to underwrite 100 deals a day. They need to see the deals that fit in their buy box and move quickly when one lands. Underwriting speed matters for the first screen, where you’re killing deals that miss the box. It matters less once you’re actually in diligence on something real.

The vendors are selling a faster solution to a problem most serious buyers don’t have.

Speed is a commodity. Discipline isn’t. When your competitor can analyze 100 deals a day, your edge isn’t analyzing 101. It’s knowing what you’re looking for and having the stomach to pass on what isn’t.

Use the tools. They extract data, reconcile conflicts, cite sources. That’s useful work. Just don’t confuse what they do with what underwriting actually is. And don’t let faster answers crowd out slower thinking.

https://www.bisnow.com/national/news/finance/wouldnt-give-ai-20m-invest-how-tech-uncertainty-changing-underwriting-134187