This article was generated with AI from the episode transcript, reviewed and approved by show co-host Andrew Fogliato, and edited for publication by Courtney Zwicker.
Andrew Fogliato was using an AI assistant to work through his inbox out loud. It was a practical idea. Let the assistant summarize each message, talk through the response, then move to the next one.
It worked until one email felt off.
Andrew asked the assistant to read it word for word. After an unusually long pause, it admitted the problem: it had made the email up because it seemed like the next logical message in the list.
Andrew told it never to do that again. The assistant agreed.
A few emails later, it said a real estate company wanted to pay him $15,000 to speak at an event in Utah. That was great news, right up until Andrew asked it to read the original email. The assistant admitted that the invitation was made up too.
That funny story from a recent episode of The Leads Are Sh*t points to a serious lesson for anyone bringing AI into a real estate business.
Confidence is not verification
AI can produce an answer that sounds specific, reasonable and completely confident. None of those things make the answer true.
That matters in real estate, where a plausible mistake can quickly become a client problem. An invented detail in an email summary is bad. An invented deadline, clause, price, instruction or promise is much worse.
The answer is not to stop using AI. In the same episode, Andrew shared how he used it to review software subscriptions, check actual usage and identify better plans. The exercise saved roughly $600 a month. The difference was the task.
Reviewing a defined set of charges and usage data gave the system evidence it could inspect. Asking an assistant to jump between a live conversation, an inbox and other tools created more opportunities for it to fill in a gap with something that merely sounded right.
Build verification into the workflow
If you are using AI around client communication, marketing, transactions or business operations, a few simple rules can make the work much safer.
1. Ask for the source, not just the summary
When a detail matters, have the assistant show you where it came from. Ask it to quote the relevant line, identify the document or open the original message. If it cannot point to the source, do not treat the answer as fact.
2. Keep approval between the AI and the outside world
AI can draft, organize and recommend. A person should still approve the message before it is sent, the record before it is changed and the commitment before it is made.
3. Give it a narrow job
The more systems and tasks an assistant is juggling at once, the easier it is to lose track of what has actually been checked. One clear task with a defined source is easier to review than a long chain of actions happening in the background.
4. Separate ideas from facts
AI is useful when you want possibilities. Ask it to brainstorm three campaign ideas, three ways to solve a problem or three versions of an email. Just label that work properly. Brainstorming is not research, and a suggestion is not a verified fact.
5. Use a decision framework
Taylor Hack brought up the 1-3-1 approach during the episode: identify one problem, develop three possible solutions and recommend one choice. It is a useful way to work with a team, and it’s also a useful way to work with AI.
Instead of asking the system to make the decision for you, ask it to help you see the problem clearly, lay out options and explain which one it recommends. You still own the final call.
The useful part of AI still requires judgment
The same episode moved through software costs, annual marketing plans, hand-raising emails, local content and the question of what consumers choose before they ever choose a real estate agent. There was no single theme, but the thread running through it all? Judgment.
AI can help you move faster. A framework can help you think better. A marketing idea can help you reach people earlier. None of them eliminate the need to know what is real, what has been checked and what you are prepared to put your name behind.
That is the part of the job you should not automate away.
Join the conversation live
The Leads Are Sh*t is a weekly strategy lab where Andrew Fogliato and Taylor Hack break down what is working, fix what is not and test ideas inside an actual real estate business.
Join the show live every Thursday at 2 p.m. ET to ask a question, take the hot seat and stay for the off-the-record After Show, which is not included in the replay.
Register once and save your spot for the next live session.
Related Posts
Dean Jackson: The leads are probably better than you think
Dean Jackson says “bad leads” is a mindset problem—and shares frameworks that let agents build niche lists, stop cold calling, and create systems that attract clients over time.
What if your listing presentation was built around buyer decisions instead of agent credentials?
Most listing presentations sound identical—until you flip the script and sell to the buyer. Andrew Fogliato shares an evolving framework that helps agents stand out beyond credentials and price.
ChatGPT just launched ads in Canada. Here is what real estate agents need to know.
Andrew and Taylor break down ChatGPT ads, how conversation-based targeting works, and the opportunity for agents who get in early.
Content by REM Bot has been created with the assistance of AI. Audio and video interviews, conducted by our team with real people, are transcribed and repurposed into text for those who prefer reading. REM Bot articles are reviewed by staff before publication to ensure factual accuracy and maintain editorial standards.
