The Practical, Powerful Ways to Use AI in Real Estate Right Now

A hunch, checked overnight

So Andrew did something that would have been impossible two years ago and expensive as recently as last year: in an evening, with no coding background, he built a little tool that pulled the Yelp rating of every Sweetgreen location in the country and charted the average over time.

The verdict? He was wrong. The ratings had actually bottomed out a year and a half earlier and had been climbing ever since. His gut had almost talked him out of a decision for no good reason. He wrote up the whole thing here, and it's worth your time.

Here's why that matters to you, even though you flip houses in Virginia and have never thought about a salad chain's Yelp reviews. There are companies that sell exactly the data Andrew built that night. They charge in the low five figures a year to start. Andrew replicated a slice of it in an evening, for basically nothing, by asking an AI tool the right question in plain English.

Sit with that. A tool that cost tens of thousands of dollars a year, or that simply did not exist, is now something a curious person can build in an afternoon. No computer science degree. No hiring a developer. Just a question and a little patience.

I'm not writing this to turn you into a programmer. I'm writing it because I want you to see the move, because the move works just as well on a rehab in Chesterfield as it does on a stock.

We're a lending shop, and we've made this move twice

We fund fix-and-flip and BRRRR deals for real estate investors across Virginia, and we write our own checks. We are not a software company. And even in a business like ours, this shift has changed how we work in two big ways.

The first is underwriting. When a deal comes in, there's a mountain of work between "here's my property" and "here's your money" — pulling comps, sanity-checking the ARV, stress-testing the rehab budget, reading the story behind the numbers. We now lean on AI to do the heavy, repetitive parts of that read faster and more consistently than a person can at 9 p.m. with a stack of PDFs. I'm not going to pretend a machine makes the final call — a human still decides, because lending is a judgment business. But the grunt work that used to take days now takes hours. That's the entire reason an operator can call us and have a real answer, and often a term sheet, inside a day instead of waiting a week.

The second is stranger, and it's the one that really shows what's possible. Part of how we grow is raising money for our lending fund from private investors, and one group we wanted to reach is people who hold retirement savings in self-directed IRAs. A lot of folks don't realize they can invest their IRA into a private lending fund like ours and let the interest compound without the annual tax drag. The trouble is there's no clean, cheap list of those people sitting on a shelf. The usual answer is to rent an expensive database and hope. Instead, we built our own way to find them from public information — a way to sift through an enormous pile of records and surface the households that actually fit. It turned up thousands of them. That project would have been a five-figure line item and a month of somebody's life a couple of years ago. We did it as a side project. (If you're curious about that side of what we do, it lives at harvey-capital.com.)

Neither of those required us to become engineers. They required us to start with a real problem we already had and ask, out loud, "could I build something to solve this?" The answer, over and over lately, is yes.

Here's the move

Strip away the specifics and the pattern is always the same. You have a hunch, or a haystack.

The hunch is the Andrew Walker version: you believe something about your market and you want to know if it's actually true before you bet money on it. The haystack is the version where the information exists but it's scattered across a hundred ugly places and nobody has ever bothered to pull it together, because doing it by hand would take forever.

Two years ago, both of those ended the same way: you trusted your gut, or you paid a lot of money, or you just did without. Today there's a third option. You describe what you want in plain English, and you build the thing.

The trick — and this is the part people miss — is that you don't build a tool for the sake of building a tool. You start with a question you genuinely have. The question comes first. The tool is just how you answer it.

So let me hand you a pile of questions. These are the kinds of things I'd be building if I were flipping houses full time. Some are simple. A couple are borderline ridiculous. That's the point.

Questions you could answer this week

Prove your own hunches

• You feel like Chesterfield has gotten too competitive. Build something that tracks the spread between opening bids and final sale prices at trustee auctions over time. Is competition really up, or did you just lose three you wanted?
• You think ARVs in a certain zip are getting toppy. Pull every sale there for the last two years and chart the price per square foot. Is the trend actually rolling over, or is that one ugly comp living rent-free in your head?
• Rehab costs feel out of control. Track the price of lumber, drywall, and a handful of appliances every week and build your own materials index, so "it feels expensive" turns into a number you can actually see.

Find the deals nobody else sees

• Watch the local probate filings and match them against property records to surface heirs who just inherited a house they don't want and don't live in.
• Pull the city's code-violation and condemnation lists and flag the owners who are clearly under pressure.
• Cross-reference eviction filings against long-held rentals to find the tired landlord who's finally ready to be done.
• Scrape the MLS listings that expired without selling — motivated sellers with a bruised ego and a house that's still there.
• Point something at Marketplace and Craigslist that watches for "as-is," "handyman special," and "must sell" and texts you the second one shows up.

See your market the way an institution does, for free

• Track days-on-market and the sold-to-list spread by zip so you know exactly where you have negotiating leverage this month.
• Chart rent growth by neighborhood over time so your BRRRR refinance numbers stop being a guess.
• Watch who's actually buying at auction through public deed records and build yourself a leaderboard of the biggest local players, then reverse-engineer what they're buying.

Move faster on the ones you win

• Feed it the listing photos and get a rough rehab scope and budget range back in a minute, before you ever walk it.
• Hand it a listing link and let it spit back your maximum allowable offer on your own rules, automatically.
• Check which contractors actually pulled permits and finished on schedule — all public record — before you hire one.

And the one that still feels like science fiction

• Point a tool at your five target zip codes and have it alert you the moment a price drops, a new listing appears, a new auction gets posted, or a renovation permit gets pulled that fits your buy box. A personal, always-on deal radar for your exact corner of the market. That used to be a data subscription that cost more than a used car every year. Now it's an afternoon and some patience.

What's actually scarce now

None of this requires you to write code. I want to be clear about that because it's the thing that stops most people before they start. Andrew doesn't code. I don't code. We describe what we want, we look at what comes back, we say "close, but not quite," and we nudge it until it's useful. The tools are rough around the edges. They're wrong sometimes. You still have to bring judgment. But rough and useful beats polished and imaginary, and the cost of trying went to almost nothing.

The scarce thing used to be the tool. Now the tool is cheap. The scarce thing is the question — the specific, real, "I've always wondered" question that only you have, because only you are standing where you're standing, looking at the deals you're looking at.

So that's my ask. Pay attention to the next time you catch yourself saying "I wish I could see," or "I have a feeling that," or "if only somebody had already pulled all of this into one place." That sentence is the whole thing. That's the raw material. Everything after it is just asking.