Rolling out AI to a team without losing the room

I have watched a lot of brokerages launch AI training the same way. All-office meeting, an impressive demo, a tool everyone gets access to, real enthusiasm in the room. Then nothing. Six weeks later usage is concentrated in the four people who would have figured it out on their own, and leadership concludes the team is resistant.

The team is not resistant. The rollout answered a question nobody asked. The demo showed what the tool can do. What most of the room wanted to know was whether this is being introduced to help them or to eventually replace them, and nobody addressed it, so they filled in the answer themselves.

Name the fear in the first ten minutes

Managing 300 sales managers taught me something that applies exactly here: the thing nobody says out loud governs the meeting anyway. In an AI rollout the unsaid thing is job security, and it is worse for your most experienced people, who are watching a technology arrive that appears to flatten the value of twenty years of pattern recognition.

So say it first. Not as reassurance theater, as a real position: here is what we believe this changes, here is what we believe it does not, here is what it means for how we evaluate you. If your honest answer includes something uncomfortable, say that too. People handle an uncomfortable truth far better than a cheerful non-answer, and they will detect the non-answer instantly.

Adoption is a trust problem wearing a technology costume. Solve the trust problem and the training takes half as long.

The sequence that works

1. Pick one workflow, not a platform. Not "we are adopting AI." Something like "listing descriptions now start from a draft." One workflow gives you a clean before and after, and it lets someone succeed in a week.
2. Recruit five volunteers, and make them representative. The temptation is to pick your five most tech-forward agents. That is the mistake, because the room will discount the result. Include at least two skeptics and one veteran with real standing. If the skeptic reports back that it saved her three hours, that is worth more than any demo you can run.
3. Run it for thirty days with a defined measure. Hours saved, drafts produced, response time. Something specific enough to be wrong about.
4. Report the failures publicly. Where it hallucinated, where it wasted time, where the old way was better. A rollout that only reports wins is a sales pitch, and your team knows what a sales pitch sounds like.
5. Let the pilot group teach the next group. Peer to peer, in small rooms. Nobody has ever learned a workflow from a company-wide webinar.
6. Only then set an expectation. Once it is proven internally, by their own people, you can say this is now how we do it. Doing that in the opposite order is how you get compliance without adoption.

Set guardrails before you need them

Write the rules down before anyone gets burned, and keep them short enough that people remember them. At minimum: no client data in tools you have not vetted, every factual claim verified by a human before it goes out, fair housing and advertising language reviewed the same as always, and a named person people can ask when they are unsure.

One page. If your policy runs to twelve pages, nobody reads it and you have created the illusion of governance rather than governance.

Learn to read the two kinds of resistance

They look identical in a meeting and they need opposite responses.

• Fear resistance sounds like dismissal. "This is a fad." "Clients want a real person." It usually comes from strong producers who have the most to lose from any change in the rules. What works is a private conversation, not a better argument in public, and a role for them in the pilot that uses their standing.
• Competence resistance sounds like agreement. Enthusiastic nodding, no usage. These are people who do not want to be the one asking the basic question in front of colleagues. What works is small groups, screen sharing, and someone visibly making mistakes first.

Treating the second group as though they are the first is how good managers accidentally alienate the people who were willing.

What you are actually building

The tools will change. Whatever you standardize on this year will look dated within eighteen months, and if the value of your rollout was the tool, you will run this whole exercise again from zero.

What survives is the capability: a team that can evaluate something new, pilot it honestly, report what did not work, and adopt it without a crisis. That is a management asset, and it is the real deliverable. The listing descriptions are just the first thing you practice it on.