The AI had never stepped inside the house. It had not climbed onto the roof, tested the electrical system, examined the plumbing or spoken with a contractor. Yet it confidently told my buyer that the property needed approximately $250,000 in repairs.
He believed it — and canceled the contract.
I have been a Realtor since 2000 and a real estate technology instructor for more than two decades. I use artificial intelligence regularly, teach real estate professionals how to use it and believe it will fundamentally change our industry.
That is precisely why this experience concerned me.
AI can be extremely helpful. But when consumers mistake a chatbot’s polished response for qualified professional advice, the consequences can be expensive.
The $250K estimate
I was representing a 70-year-old attorney whom I will call George. He was searching for a home with a budget of up to $2 million. We found a property he liked, negotiated a contract and hired a home inspector. The inspection took approximately three hours.
George could not attend the inspector’s verbal summary, so I recorded it and uploaded the video to YouTube using an unlisted link. This allowed him to watch the complete explanation and privately share it with family members, friends or advisors.
I also sent him the approximately 75-page written inspection report.
We planned to speak the following day to prepare Arizona’s Buyer’s Inspection Notice and Seller’s Response, commonly known as the BINSR. This is when buyers typically request repairs, ask for a credit or decide whether to move forward.
When we spoke, George told me he wanted to cancel.
I asked whether he would consider first submitting a repair request. The property had been on the market for some time, and I believed the seller might be willing to address many of the concerns.
George said no. He had provided the inspection information to an AI tool, and it estimated that the repairs would cost approximately $250,000.
I was stunned. I had reviewed hundreds of inspection reports and negotiated hundreds of inspection responses. Nothing in this report appeared remotely close to a quarter-million dollars in repairs.
I explained that an AI tool could not physically inspect the property, determine the actual scope of work or accurately account for local labor and material costs.
George still wanted to cancel. So we did.
What the contractors found
The listing agent and seller were understandably disappointed. After learning why George had withdrawn, the seller brought in two licensed contractors to evaluate the inspection items.
- One estimate came in at approximately $17,000.
- The other was approximately $19,000.
- The AI estimate was off by more than $230,000.
By then, however, George and I had already moved on. We found another home, negotiated another contract and paid approximately $750 for a second inspection. Once again, I recorded the inspector’s verbal explanation so George could review it carefully.
A few days later, George consulted his AI advisor again. This time, it estimated that the property needed approximately $150,000 in repairs.
I reminded him what had happened with the first home. Two licensed contractors had evaluated the actual work and reached numbers nowhere near the AI-generated estimate.
Then I gave him a choice.
“You can trust an AI tool that has never seen the property,” I told him, “or you can trust the inspectors, contractors and professionals who have.”
Then I stopped talking.
For nearly a minute, George sat silently and looked at me. Finally, he said, “I guess I’ll go with your answer, Jeff.”
We brought in two licensed contractors. Their estimates came in at approximately $15,000 and $18,500.
We asked the out-of-state seller to complete the repairs or provide George with a $20,000 credit. The seller agreed to the credit, and the transaction moved forward.
AI is changing the agent-client relationship
The lesson is not that buyers, sellers or agents should avoid AI.
I use it. I teach it. I believe it can help real estate professionals work faster, communicate more effectively and serve clients better. But AI must be used for the right tasks.
A chatbot can summarize an inspection report, organize concerns into categories and help a buyer prepare questions for an inspector or contractor.
What it cannot do is physically evaluate a property and produce a dependable repair estimate based solely on a written report.
This experience taught me five important lessons:
- AI output can sound more certain than it really is. A confident answer is not necessarily an accurate one.
- Consumers may place enormous trust in AI. Agents should not assume their clients understand its limitations.
- Inspection reports require context. A lengthy report does not automatically mean a home has catastrophic problems.
- Repair estimates should come from qualified professionals. Inspectors identify concerns; licensed contractors determine scope and cost.
- Agents must learn how to challenge AI without dismissing it. Simply telling a client that AI is wrong may not be enough. We must explain why and provide credible human evidence.
The first AI estimate cost George a house he might otherwise have purchased. The second nearly did the same.
The agents who succeed in the AI era will not be the ones who reject the technology. They will be the ones who know when to use it, when to question it and when to insist that a qualified human remain in the loop.
Jeffrey “Mr. Tech” Raskin is President at Tech4RE | HomeSmart. Get connected on Facebook and LinkedIn.
