You remember the kickoff. The vendor on stage. The slick demo where the tool wrote a listing description in nine seconds and the room actually gasped. The email from leadership with the word “future” in it. “Revolutionary.” “Groundbreaking.” Buzzword, buzzword, buzzword. The best new thing you never knew you needed and, if we’re being honest, never asked for. Maybe a leaderboard. Maybe swag.
It reminds me of Ralphie coming down the stairs in the pink bunny pajamas. Somebody who loved him very much picked them out. He never asked for them, he didn’t want them, and he was never going to wear them again after that moment.
Now pull up the login report for that tool and look at week seven.
Nobody announces month two. There’s no email that says, “we’ve quietly stopped.” The subscription keeps renewing, the logo stays on the recruiting deck, and the agents go back to doing exactly what they did in March. The rollout didn’t fail. It just never got attached to anything.
I have trained more than 50,000 agents, and I’ve been a mile deep in training and strategizing AI for brokerages, businesses and associations since it went mainstream in 2022. I’ve watched this happen from inside the training room, not from a vendor deck. The month-two stall is not a technology problem, and it is not an agent problem. It is an order-of-operations problem. Most brokerages roll out AI backwards.
The rollout was a press release with a login attached
Here is the real talk. A lot of AI rollouts are not built for agents. They are built for the public, for the recruiting page, and, dare I say it, for the stock price. The goal is to look like the brokerage of the future by Friday. So the sequence goes: Announce, demo, distribute logins, hope.
That sequence produces what I call Big AI Tool energy. It’s real, it’s loud and it lasts about three weeks. Big AI Tool energy carries an agent exactly as far as their first frustrating result. Then they ask the three questions every rollout has to answer and almost none do: What does this actually do? How does it help me specifically? Why would I open it every single day?
If those questions go unanswered, “no” is the easiest path for an agent to choose. Think about who you’re asking. Their business is going well. They have a system that works for them. So the first bit of frustration with the new tool, the first result that doesn’t sound like them, and the math is done in about four seconds: I don’t really have a problem this thing is solving. I don’t need it.
Not a dramatic no. A quiet one. The tab just stops getting opened. Their old way still works, and their old way doesn’t require learning anything at 9:40 on a Tuesday between a showing and a school pickup.
The numbers back this up. In the National Association of Realtors’ 2025 Technology Survey, only 20 percent of members said they use AI daily, and 46 percent said AI has had no noticeable impact on their business. Read that second number again. Almost half. That is not a room full of people who tried it and hated it. That is a room full of people who were handed a login and a demo and nothing else.
It’s not just us. An MIT report found roughly 95 percent of corporate AI pilots delivered no measurable return, and the researchers blamed a learning gap, not the models. Same disease. Training that stops at “here’s how it works” and never gets to “here’s the job you hand it.”
Training was the last department in the room
True story. I was at a company that was rolling out an amazing new AI feature on its website. Concierge feel, beautiful design, completely on brand. The project team and the owner had been meeting about it for months. The people building it had real expertise in branding and marketing. Not training.
Guess which department got brought in last. Training. By then the tool was finished, integrated into the website, and somewhere between five and six figures had already been spent. And the brief to us was, more or less, we made this beautiful toy for our agents, now go tell thousands of them why it’s good for them.
Nobody asked for feedback before it was built. Nobody said, test this out and tell us how to make it better. And here’s the thing that still gets me. The training department is the one that’s on the front lines with agents every single day. We hear, see and feel every bottleneck in their business. We know exactly where AI could help an agent deliver a better concierge experience, because agents tell us, out loud, in every session. That knowledge was sitting in the building the whole time. It was invited in after the paint dried.
Think about it like this. You painted the room. You made it beautiful. You made sure it was on brand. You just never asked the customer what her favorite color was or whether the room needed painting at all.
So I stood in front of a lot of agents and explained a tool that had been built for the website, not for their Tuesday. You already know how month two went.
Meet with the coach before you buy the sprinter anything
Picture a young sprinter who’s breaking records. Don’t walk up to her with a new training program and tell her she needs it. She doesn’t believe she has a problem, and she’s not wrong. Meet with her coach first. The coach watches every race and tells you she’s a half step slow out of the blocks. There is one tool that helps with that one part of her process. You give her that. It helps her every single time she races, and she’ll use it every day without being told, because it fixes something she actually feels.
The coach is your training department. The blocks are the bottleneck agents already know they have. The one tool for that one part of the process is what a rollout is supposed to be. Everything else is a vitamin.
The brokerages that get past month two flip the order
Here is what the brokerages that make it past month two actually do differently, in the order they do it. None of this is fancy. All of it is the opposite of a launch party.
First, they start with the bottleneck, not the build. Before anyone designs anything, the people who talk to agents every day are in the room. Training, coaching, the managers who hear the same complaint 15 times a week. The question is not “what could AI do for us.” The question is “where do our agents get stuck every single day.” That answer picks the tool. Not the other way around.
Second, they train a group before they announce anything. Not the whole company. Twenty agents, a floor, one office. Real training, meaning the tool is pointed at something those agents already do every day. Not “here’s what AI can do.” Here’s your Tuesday. Here’s the buyer follow-up you write 11 times a week. Here’s how you hand that job to the tool so it comes back sounding like you. A new tool plus a new habit is two things to learn, and agents have time for about half of one.
Third, they build the ongoing system before the first login goes out. Everybody says this part. Very few brokerages budget for it, staff it or put it on a calendar, and it is the part that decides month two. Training is not an event. It’s a stack. A short session, a week of using it, a check-in where people bring the ugly results and fix them together, another layer on top. Stacked learning. Every week the skill gets a little taller and the excuse gets a little smaller. If your rollout plan has a launch date and no calendar after the launch date, you don’t have a rollout. You have a party.
Fourth, they let the trained group become the rollout. This is where the whole thing turns. Once 20 agents are getting results in their own voice on tasks they actually do, you don’t need a vendor on stage. You have evangelists. Agents believe agents. The listing that got seven showings because the description finally sounded like the neighborhood is worth more than any keynote, and it spreads at the copier, in the car and in the group text where the real culture of a brokerage lives.
Fifth, they announce, and then they keep going. Now the announcement has proof, people who’ll vouch for it, and a training calendar already running. It’s the same email the backwards brokerage sent in week one. The difference is this one is true. And it doesn’t end there, because new agents join and models change. The brokerages that stay past month two treat AI training the way they treat contract training. Ongoing, expected, part of the job. Nobody rolls out a purchase contract with a launch party and then never mentions it again.
Here is the play
Let me simplify this. Before you buy, build or announce anything, get the people who hear the bottlenecks into the first meeting, not the last one. Then answer one question for every agent: What is the one job they do every day that this tool is going to take over, and who is going to show them how? If you can’t answer it, you’re not ready to roll out. You’re ready to write a press release.
Pick the group. Train them on the Tuesday task, not the demo. Put the weekly check-ins on the calendar before the first login. Let the results travel by word of mouth. Then, and only then, announce.
Done in that order, the same rollout that would have died in month two makes the brokerage look great, for real this time. Agents deliver better service because the tool sounds like them. Recruiting gets easier because the training is a real benefit, not a logo. Leadership gets the dashboard it wanted. Everyone wins, and nobody had to gasp at a demo.
The tech is artificial. The results are real. But only if somebody teaches people how to get them.
P.S. If you are in month two right now, you don’t have to start over. Pull 20 agents, ask them where they get stuck, pick one daily task and run four weekly sessions on it. Don’t announce it. Let them talk. You’ll know inside a month whether you have a tool problem or a training problem, and I already know which one I’d bet on.
To learn more visit https://jmanai.com/.
