For Sale By AI:
The Post-Mortem.
I let an AI make every decision in the sale of my home in Arlington, Virginia, from the list price to the contract. Here is how that went, and what it says about the 2.5% everyone still pays an agent.
I sold my house in June 2026 at full list price, $1,785,000, with no agent commission paid on either side, buyer's or seller's. That left about 5.5% of the price, roughly $98,000, with the buyer and me instead of with agents. An AI made every real decision in the sale: the price, the comparables, the staging, the contractors, the marketing, the words on this page. I did the physical work and signed the papers. As far as I can tell, it's the most expensive home anyone has sold this way.
- AI can already make the judgment calls an agent charges 2.5% for. It priced, staged, and negotiated as well as any agent I've worked with.
- AI did analysis no agent would bother with. It also worked on records no agent can see: my own email, receipts, and closing file. The listing's strongest selling points came from combining the two.
- Physical AI is the missing piece. Someone still has to host the open houses, check the contractors' work, sweep the dead flies off the windowsills, and guard the house. I did all of it, and I lay awake when buyers passed.
- The best contractors are invisible online. Review sites favor whoever pays for ads. Permit records show whose work passes inspection.
- Agents are afraid, and some will use their networks against you. Selling this way threatens their income. Instead of arguing the merits, some put up obstacles, and one told me that selling my own house was illegal. It isn't.
- The buyer-side commission gets overlooked. Everyone argues about the listing side, but the buyer side no longer has to go to a middleman. In my sale it stayed with the buyer and me.
Data and AI are what I do for a living. When it came time to sell the house, I wanted to know whether the tools I use all day could run the biggest transaction most people ever make.
I used AI for the whole sale: the list price, the comps, the staging, which fixtures to pull, who to hire, what the photographer should shoot, what the ads should say, and the tone of every reply to every tire-kicker who emailed. The AI made the decisions, and I did the physical work.
A 2.5% listing commission on a house like this is real money. But mostly I wanted to answer one question.
When an AI runs the sale instead of an agent, can the market tell the difference?
As far as I can tell, it can't.
01 AI handled the judgment calls
Most of what an agent charges 2.5% for is judgment, and in my experience AI now makes those calls well and for almost nothing.
Think about what you're paying an agent for: the call on what to list at, the instinct to stage the basement as a bedroom and not an office, the eye to swap the dated brass and keep the hardwoods, the read on how to answer a lowball, the sense of which contingency to fight. My favorite example was the sump pump. The AI told me to leave it out of the listing. It's a good thing to own, but to a nervous buyer the word whispers "this basement floods." Calls like that come from pattern-matching across thousands of past deals, which is exactly what these models are good at.
Pricing was where the AI was least sure of itself. I had it backtest its own method against closed sales in the neighborhood, and on a house this unusual the error bars were wide. It gave me a range with its reasoning, and I chose the low end to sell faster. Both serious offers came in at list or above, so the range held up. Signing was still my decision.
02 AI did analysis no person would do by hand
No agent was going to pull the county's GIS records for all 458 single-family parcels in Aurora Highlands, test which lots touch two streets to sort corners from interiors, filter for size, and prove with a map that my house is one of only 22 non-corner homes in the neighborhood on a 9,000-plus square foot triple lot. Nobody does that for a commission. The AI did it in an afternoon, and the result became the marketing: the map on the homepage, the "1 of 22" line in the ads, and the strongest argument for the price.
The neighborhood was platted in 25-by-120-foot strips, so reaching 9,000 square feet takes three of them. Every one of the 22 is three original 1920s lots stitched together, mine included: Lots 5, 6, and 7 of Block 20. I'd owned the house for six years and never knew that.
It did the same kind of digging in my own records, which no agent could have seen. The "$80,000 in receipted improvements" on my listing wasn't an estimate. The AI went back through years of my email, Costco orders, and store receipts and rebuilt the invoice trail line by line: the Trane HVAC system, the refinished floors, the egress window that made the basement a legal bedroom, down to the garbage disposal. It read the House Location Survey the title company sent me six years earlier, a flat image scan with no text layer, and pulled the exact platting off the page.
If I were advising another seller, I'd start there. Even if a brokerage hires a thousand agents, none of them can search your Gmail. You hold those records, and AI can work through them alongside the public record and produce claims a buyer can check.
03 Physical AI is still missing
No model could climb a ladder, crimp a wire, or drive to Home Depot at 9 p.m. because the thing I needed wasn't the thing I bought. None could move the hose when the sprinkler timer misfired at 6 a.m. I let in the stager, the photographer, the egress crew, and the floor guys, walked the punch list, and did it again the next day. I did hundreds of those things over a few weeks, with a newborn in the house.
The labor was obvious. It took me longer to see how much of the job is just being there, and how much of an agent's time goes to exactly that.
Open houses need someone to unlock the door at noon, greet strangers until five, and lock up after the last one leaves. Contractors' work needs checking in person. The AI picked the crews and read their permit histories, but it couldn't stand in the room and see whether the paint line was clean or the trim was caulked. The house has to look its best before every showing, with the lights on, the blinds set, and the dead flies swept off the windowsills. And it has to be protected. A listing publishes the address, the photos, and the showing schedule to anyone who looks, which makes the house a target for damage, theft, and squatters.
All of that recurs, and all of it needs a person in the house. Today's AI can't do any of it. I don't think most homeowners should do this yet. A robot that could do the work, walk a punch list, clean before a showing, and keep watch at night would change my answer. That robot doesn't exist yet, though it's closer than I expected when I started.
04 Finding contractors in the permit records
Yelp, Angi, Thumbtack, and Houzz reward contractors who advertise and chase reviews. The best trades I found were booked solid through word of mouth and barely showed up online.
I had the AI build a tool to find them. It pulled Arlington County's entire building-permit history, about 18,000 records, and Virginia's full contractor-license file, roughly 50,000 more, and ranked contractors by whether their permits passed final inspection, which a review can't fake. It also ranked contractors with strong permit records and few or no online reviews higher on purpose, since a thin Yelp page often means someone is too busy working to chase stars.
The best contractor in the data nearly got missed. He had a near-perfect inspection history over more than a decade, but his 28-year-old master license was filed under his own name instead of his company's, and the county had misspelled his business on half his permits. The tool's first pass ranked him mediocre. His record put him among the best in the county, and my own search had almost skipped him. The public record told me more than the reviews did.
I later turned the tool into a public site, ContractorReceipts.com, built from over 6 million permits across five markets, Arlington included. It ranks contractors on permits closed out and license history, shows which firms professional property owners keep rehiring, and links every number to the public record behind it. No contractor can pay to appear there. Where the site doesn't reach yet, I'd ignore star ratings, get several quotes, and pay attention to how clearly each contractor writes back and how much thought went into the quote.
05 How the agents reacted
I held a brokers' open house. The AI built the guest list by pulling every agent who'd sold a home in my zip code, ranking them by volume, and writing each one a note that opened with their own numbers. One began: "Zillow shows you with 118 sales in 22202, which is on a different planet from most of my invite list." About 250 agents got a personal invitation. The AI expected three to ten percent of them to come, so I ordered Chick-fil-A for thirty. Two came. I ate the leftovers, more or less alone, for the better part of a month.
Some of the replies were condescending. One agent, at a national brokerage that says it's leading the charge on AI in real estate, answered my invitation with what he called unsolicited advice. He opened by noting kindly that he could see I was on paternity leave, then told me that what I was doing was illegal, because "only a licensed real estate broker can assist a buyer or seller in the purchase or sale of a home for compensation." He was describing me selling my own house. In the same email he conceded that "a seller can sell their own property on their own," adding that it "typically ends with them eventually hiring an agent after being unsuccessful." To his credit, his next email pointed out a real problem with my MLS listing, and I fixed it.
Another agent asked why I'd left photos out of the invitation, a fair point, and added, "Surprised AI didn't give you these tips!" When I replied that I'd pay a standard 2 to 2.5% to a buyer's agent, he wrote back, "'Standard' is not less than 2.5%, fyi."
I don't hold it against them. If someone ran a public experiment to show that the core of my job could be automated, I'd be prickly too. Their income depends on the old way of doing this.
06 Comparing the two offers
I had two serious offers. One came in about $10,000 over asking and carried a 3% buyer-agent commission. The other came in at list with no buyer-side commission. The first looked better on paper. After the commission, the second left me with more, because three percent of $1.8 million is a lot more than ten thousand dollars. I took the second.
Sellers look at the price first. The difference between the price and what you keep is usually a commission, often one the seller pays on behalf of a buyer they never met.
07 Where the buyer-side commission goes
In a normal sale, when a buyer walks in without their own agent, the listing agent usually keeps the buyer-side commission or steers it to a friend who "represents" the buyer in name only. The commission is already built into the deal, and the unrepresented buyer rarely sees any of it.
In my sale there was no listing agent to take it. The house closed at full list with no buyer-side commission paid to anyone, so that 3% stayed with the buyer and me. The buyer didn't finance someone else's fee into a mortgage, and I kept my full asking price.
This is one sale, so treat what follows as a guess. Flat-fee MLS services are already cutting into the listing-side commission, and I expect the buyer side to go next. As buyers use AI to evaluate homes, read inspections, and check contracts, more of them will skip having an agent, and the roughly 3% that now goes to a buyer's agent almost automatically will be up for negotiation. That money can go to the buyer as a lower price, to the seller as higher proceeds, or to whatever middleman is still around. For forty years it has gone to the middleman by default. I doubt that lasts another ten. The industry collects something like a hundred billion dollars a year in commissions, so a lot rides on it.
08 What the industry still controls
With the judgment calls handled, I ran into two things a homeowner still can't easily get around: the MLS and the purchase contract.
The MLS is the database every buyer's agent and listing site pulls from, and only licensed members can post to it. For decades that meant you listed with an agent or buyers never saw your house. That's no longer true. A flat-fee service will put your home on the MLS for a few hundred dollars with no agent and no commission. I used Beycome, and Houzeo and others do the same. The portals syndicate from the MLS, so my house showed up on Zillow, Redfin, Realtor.com, and every agent's search looking exactly like an agent-listed home. The access rules haven't changed, but getting onto the MLS now costs about three hundred bucks instead of two and a half percent.
The contract caught me off guard. The standard purchase contract belongs to the National Association of Realtors and its state and local affiliates, and it's licensed only to members, so as a for-sale-by-owner seller I couldn't use it. I had to commission a neutral purchase agreement through a title company. That cost time and money, and a one-off contract carries more risk of a mistake than a form everyone already knows. AI can't fix a licensing restriction. Of everything in the sale, that's the barrier that gave me the most pause, though I don't expect it to last forever either.
09 The stress didn't go away
I assumed that handing the decisions to AI would take most of the stress with them. Every buyer who walked through and passed got to me anyway, and so did quiet showings and days with no interest. I kept wondering whether I was wrong, whether anyone wanted the house, and whether the confident data guy had mispriced his own house in front of everyone he knows. The comps and the model both said the price was right and that most buyers pass. I worried anyway.
What I think now
AI handled the pricing, staging, and marketing of a $1.785 million house as well as any agent I've worked with, and the house closed at full list with no buyer-side commission. I wasn't the first to try this. A homeowner in Florida and a reporter in upstate New York did it before me, but both of their houses sold for under a million dollars. As far as I can tell, this is the most expensive home an owner has sold without a listing agent while AI made the decisions.
AI can make the judgment calls, but I still had to do the physical work and worry about the sale. Fully automating it would take a robot that can host an open house, and sellers who don't care whether the neighbors think they botched it. In my sale the buyer-side commission stayed with my family and the buyer, and I expect more sellers to keep it as these tools spread.
I wouldn't recommend this to everyone. It was a lot of work with a newborn in the house, and I ate a frankly upsetting amount of cold Chick-fil-A. I'd still do it again. Nobody in the market seemed to miss the agent, and I kept the commission.
Stuart