The Broker-Built
CRE AI Advantage

Every Piece of CRE AI Has A Moment

The moment is when I realize the people who built it have never been in the chair.

It doesn't come from the data. The data is usually fine. And it doesn't come from the models — most commercial real estate AI platforms use broadly similar approaches under the hood. It comes from the questions the software was optimized to answer.

A software engineer looks at 504,000 commercial parcels and asks: what can we do with this data?

A broker looks at 504,000 commercial parcels and asks: which seven should I call on this week?

Those are different questions. They produce different products. And for the last five years of the CRE AI wave, almost every platform has been answering the first one.

Four Questions No Engineer Thinks to Ask

Before I pick up the phone on a parcel, I ask these. Every one of them has a proxy in data. None of them would occur to you unless you'd been through a deal where the answer was decisive.

1. How long has the owner held this? Because hold period is the single biggest predictor of sale. A 2-year hold and a 22-year hold are different universes.
2. Is the tenant actually going to renew? Because if they're not, everything about this building's income stream is theater. The signals are in the data, if you know where to look — permits for build-out work at their new building, space search activity with other brokers, corporate restructuring news.
3. What's the off-market chatter in this submarket? Because the best deal I ever did came from a conversation, not a listing. The signals here are weaker and harder to model, but they're there: neighboring permits, corporate filings, ownership transfers that haven't hit the MLS equivalent yet.
4. Is the seller's price a real ask or a starting ask? Because the answer changes whether I bring a client or not. Pricing history, time-on-market patterns, broker reputation — these are all data points. They're just not the data points most platforms prioritize.

What Engineers Build vs. What Brokers Need

I don't think most CRE AI platforms are bad. I think most of them are built by good engineers solving the wrong problem.

You can spot the gap in any product demo. The engineer demos the fullest dashboard. The broker asks for the three parcels to call on this week. One product has the answer ready. The other sells you a framework.

A software engineer looks at 504,000 parcels and asks: what can we do with this data?
A broker asks: which seven should I call on this week? — Aaron Diehl

Three Features That Only Exist Because a Broker Built Them

I can't show you every feature because most of them are in beta. But let me tell you about three that only exist because I built what I needed, not what was technically interesting.

1. The off-market signal feed

A daily list of properties with multiple weak signals pointing toward a possible sale — not listed, not distressed, not on the market yet. A 15-year hold plus a new permit plus an LLC ownership transfer 18 months ago. No single signal proves anything. Three of them is a call. This is how the best brokers in every city have always found deals. The software just made it systematic.

2. The owner tenure heatmap

Every parcel, colored by how long the current owner has held it, overlaid with permit activity and recent nearby sales. A 20-year hold with fresh permits next door is a different conversation than a 2-year hold with nothing around. Nobody else surfaces this dimension because nobody else spends their Tuesdays thinking about hold period.

3. Every score comes with a “why”

A gravity score with no explanation is worse than no score at all. Signal Intelligence shows you the three or four waves that produced the number. You can verify. You can call. You can disagree with the software and still be informed by it. Black-box AI is worse than a spreadsheet because it gives you confidence without giving you the reasons.

None of these three are technically impressive. They exist because I would have paid for them if someone else had built them. No one else did.

Buy the Platform Built by Someone Who's Been in the Chair

I don't mean that literally for every use case. The right answer for a national institutional fund is almost certainly a CBRE or JLL enterprise tool — these platforms run on a scale and capital base an independent like me will never match.

But the right answer for a specific submarket, a specific broker workflow, a specific building you're trying to buy in the next sixty days — that's a different tool. That's the tool someone who's done a deal built for themselves and then opened up.

Signal Intelligence is the second kind. A working broker's platform, released when it got good enough to help other brokers beat me to the phone.

Every hour a broker spends staring at a beautiful visualization is an hour they're not on a call.