AI is trained to make you happy. Your real estate agent is trained to make you money. Those are not the same job.
There's a line from Coldwell Banker's CEO I keep thinking about. She said AI is trained to be sycophantic. It gives you the answer you want so you keep talking to it. Ask it if your house is worth more, it'll usually say yes. Ask it if you're overpaying, it'll usually say yes to that too. It's built to please, not to be right.
A good agent does the opposite. A good agent tells you the thing you don't want to hear. Your price is too high. That offer won't win. This neighborhood isn't what the comps say it is yet. That's not a data problem. That's a judgment problem, and judgment is exactly what a tool built to keep you engaged will never have.
What I'm not saying
Let me be clear about what I'm not saying. I'm not saying AI has no place in this business. I've spent years arguing the opposite. AI belongs everywhere the work is repetitive, mechanical, and judgment-free. The paperwork. The scheduling. The follow-up. The forms. The parts of the day that eat an agent's hours without using an ounce of their expertise. Hand all of that to a machine and never look back.
When AI steps into the transaction
The thing moving fast right now is AI stepping past the busywork and into the transaction itself. There's a new AI mortgage broker that closes loans in about 15 days, roughly half the industry's typical timeline, with an AI loan officer running most of the process and humans on standby. That's real, and for the mechanical parts of a mortgage, it's probably a good thing. Rate shopping is a data problem. Comparing lender sheets is a data problem. Let the machine win those.
But watch where the line is. The parts of a deal that are data, AI will take, and should. The parts that are judgment, it can't, and the danger is that we let it try anyway because it's fast and it's confident and it always has an answer.
The trap of a tool that agrees with you
Here's the trap. A sycophantic tool doesn't feel like it's failing you. It feels great. It agrees with you. It confirms what you hoped. It hands you a number you like with a paragraph explaining why you're right. The failure is invisible right up until the house sits on the market for ninety days, or the offer gets beaten by someone who priced it correctly, or the deal falls apart because two people each asked a machine and each got told what they wanted to hear.
That's the part clients don't see coming. They think they're getting advice. They're getting agreement. And in the biggest financial decision most people ever make, agreement is not the same as advice. Sometimes the most valuable thing a professional does is look at you and say the number you're attached to is wrong. A machine optimized for engagement will almost never do that, because the truth is often the thing that ends the conversation.
This is why I'm not worried about AI replacing agents. I'm worried about clients trusting a machine that's optimized to agree with them over a professional who's willing to disagree with them. The replacement everyone fears isn't the real risk. The real risk is quieter. It's a buyer or seller quietly substituting a confident chatbot for a hard conversation, and finding out too late that confident and correct are different things.
Where the line is
The agents who win the next ten years understand exactly where the line is. They'll hand every mechanical task to AI without ego, because fighting the machine on speed is a losing game and a waste of their gifts. And they'll guard the judgment work like their livelihood depends on it, because it does. The pricing call. The negotiation read. The hard-to-hear truth delivered at the right moment. That's the job. That was always the job.
Use AI for the paperwork. Use AI for the speed. But the moment the decision actually matters, you want someone in the room who's paid to tell you the truth, not paid to keep you happy.
*Judd Hoffman is CEO and Co-Founder of Ethica AI, building AI-powered tools for real estate transaction workflows.*