Ryan Serhant almost lost a fifty million dollar deal last week. The story is short, the lesson is long, and every agent watching the AI conversation should sit with both for a minute.
Did Ryan Serhant really almost lose a $50 million deal because of ChatGPT?
Here's what happened, per CNBC's reporting from June 16. Serhant brokered a deal on a fifty million dollar property. At the last minute, the seller asked ChatGPT whether the price was right. In Serhant's words: "at the last minute the seller uses ChatGPT, asks it, 'Should I sell at this price?' And maybe because of how he asked, whatnot, ChatGPT basically told him no, you should not sell at that price, it's worth more." So the buyer did the same thing. Serhant: "the buyer did the same thing, asking the AI tool from OpenAI if he was overpaying, and ChatGPT told him that, yes, he was paying too much. It gave him comparables that showed why, without context and without actually understanding the property."
Both sides walked into the closing convinced the other was wrong, and the deal almost died on the spot. Serhant salvaged it by sitting both parties down and explaining what AI does and doesn't do inside a real estate transaction. The Instagram post he made about the whole thing has over three million views.
What did Ryan Serhant say AI can and cannot do in real estate?
The line that anchors his explanation is the one I want every agent to write down. In Serhant's own words: "AI can model a market. It can't model a deal."
Serhant went deeper in his post on what AI is missing inside a transaction. He said AI "doesn't know the future, it can't predict the future. It doesn't know intentions, doesn't know emotions, doesn't know what buyers are circling, doesn't know off-market comparables, doesn't understand, fully, replacement costs, and doesn't actually optimize for the deal." Read that list again. That's the entire negotiating environment. That's everything that determines whether a transaction closes. And none of it is in the data the AI is looking at.
What is the difference between an AI market model and a real estate deal?
I have been waiting for the perfect example of the thing I keep trying to explain about AI in real estate, and Serhant handed it to me. The market and the deal are not the same thing. A market is a pattern. It's the average of what has already happened. Square footage. Comps. Days on market. Interest rates. Macro trends. AI is excellent at this, because patterns are exactly what large language models are built to find. Hand a model a thousand transactions and it will tell you, with statistical confidence, what the next one should look like in the aggregate.
A deal is a different animal entirely. A deal is a moment. It's two specific humans, in a specific situation, with specific motivations, at a specific point in their lives. The seller's mother is sick. The buyer's wife loved the kitchen and hated the basement. There's a competing offer the listing agent has not formally surfaced. The buyer's lender pulled a curveball on the rate lock. The seller is moving for a job and the timeline is non-negotiable. Two other buyers walked the property last weekend, and one of them is about to come back with a number that resets the whole conversation.
None of that is in the data. None of it ever will be. AI can give you the spreadsheet. The deal lives in everything the spreadsheet does not capture. The agent's job is the deal. It always was.
Should real estate agents be worried about clients using ChatGPT?
And here is what should excite every real estate agent watching the AI conversation, instead of scaring them. The clients who walk into a transaction with ChatGPT comps in hand are the clients who need the agent in the room the most. Because the comps are real. The judgment about what those comps mean for this specific buyer, this specific seller, at this specific moment, is not in the AI. It is in the agent. The buyer who asks ChatGPT if he is overpaying is the buyer about to lose a deal because his AI advisor has no access to context. The agent who can explain the difference between the model and the moment is the agent who closes that deal.
How should agents respond when clients bring AI advice into a transaction?
I want to be careful here, because the temptation in this industry is going to be to take the Serhant story and use it as ammunition against AI. That would be the wrong read, and it is not the read Serhant himself took. He told CNBC he believes AI is a critical tool for agents, and he launched his own AI-powered workflow platform called S.MPLE. He is not anti-AI. He is pro-judgment. So am I.
AI is not the enemy. AI is the new client. The client now arrives with their own AI advisor whispering in their ear, and the agent's job has expanded to include explaining where the advisor is right and where the advisor is missing the room. That is more complex than the old job, not less. It is also more valuable, because the agents who learn how to do it well will be the ones who matter in the next decade.
This is what I have been writing about for weeks. A few days back I wrote that AI is growing up, that the technology is moving out of the demo phase and into real industries with real consequences. The Serhant story is exactly what that looks like in practice. AI walked into a fifty million dollar real estate deal last week and almost broke it. The deal closed because a human knew the difference between the model and the moment.
There is also a quieter point in this story that deserves attention, and it comes from CNBC's reporting. Kamini Lane, who runs Coldwell Banker Realty, told CNBC that her agents are seeing more clients on both sides use generalized AI tools to price homes and calculate offers. She made a sharp point. In her words: "One of the most important things that agents can see, that ChatGPT, or any other AI tool is not going to know, is what's up and coming. So neighborhoods that are up and coming, design features that are up and coming. Anecdotal data that agents are aggregating through their conversations, that is something that no AI tool is ever going to be able to aggregate in the same way that a real estate professional can."
She is right. That information lives nowhere on the internet for the AI to scrape. It only lives in the agent. And it is the part of the value the agent provides that has actually grown more important since AI arrived, not less. Because everyone now has the spreadsheet. Almost nobody has the room. The agents who can speak credibly about the spreadsheet AND the room are the ones whose phones keep ringing.
So here is what I would put on every agent's wall this week. Serhant's line. "AI can model a market. It can't model a deal." Your job has always been the deal. The clients showing up with AI advice are not your competition. They are your opportunity to demonstrate exactly what you do that the model cannot.
Ryan Serhant almost lost fifty million dollars last week because two humans asked an AI what to do. The deal closed because a third human walked in and explained the difference between a market and a moment. Credit where it is due, that framing is his, and it is the cleanest summary of where AI fits in real estate I have seen anyone put on the record this year. The whole story is on CNBC's site, and I would tell every agent reading this to go read the original piece.
*Judd Hoffman is CEO and Co-Founder of Ethica AI, building AI-powered tools for real estate transaction workflows.*