Go first, out loud.
Most leaders who want their team to use AI start with an announcement. There is a memo, a kickoff meeting or a slide in the all-hands deck, and people read it, nod and keep working the way they did the week before. What moves a team faster is watching the person in charge use the thing, badly at first, where everyone can see it.
People copy what they watch.
Teams take their cues from what the boss does on an ordinary Tuesday. If the director says AI matters and then writes every board report by hand, the staff draw the obvious conclusion that this is something for other people. If the same director forwards a draft with a note that says "the first version came from AI, I rewrote the middle," the message lands differently, and nobody had to schedule a training.
There is also a permission problem underneath. In a lot of organizations, people already use AI on the side and keep quiet about it, because they aren't sure it's allowed or they worry it looks like cutting corners. A leader who uses it openly settles that question faster than a policy would. The staff can see that it's allowed, and they get a rough picture of what good use looks like.
The clumsy attempts are the useful part.
Leaders tend to wait until they're good at something before they show it. With AI that instinct works against them. A polished demo tells the team that the leader is skilled, while a messy one tells them the tool takes practice and that practice is expected.
Say the head of a school district office asks AI to summarize a long state guidance document, and the summary leaves out the one deadline that matters. Telling the staff about that at the next meeting is worth more than any perfect example. They see that the output needs checking and what checking looks like in practice. It also gets easier for the next person to mention their own miss, because the most senior person in the room already did.
A leader's bad first attempt, shared out loud, gives everyone else permission to have one.
The same goes for the small wins that took three tries. If it took you four rounds of back and forth to get a usable onboarding checklist out of AI, say that. People who gave up after one bad answer will recognize themselves, and some of them will try again.
Out loud means specific.
"I've been using AI a lot" doesn't help anyone, because there is nothing in it a person can copy. What helps is the task, what you asked for, what came back and what you changed. "I gave it last quarter's intake numbers and asked for the questions the board was likely to raise. Most were obvious, but one I hadn't thought of" is something a program manager can try that afternoon with their own numbers.
Being specific also means being honest about limits. If you tried to have AI draft a sensitive personnel letter and decided it wasn't the right tool for that job, tell people, and tell them why. Teams learn where the edges are from a leader's actual decisions much faster than from a list of rules in a shared drive.
Where leaders slip.
The most common slip is going first once. A leader shares one example at a kickoff, then goes quiet for two months, and the team reads the silence as the real message. Going first has to look like a habit, so small and frequent beats big and rare.
Another is turning it into a performance. When every mention of AI is a story about how much time the leader saved, people hear a quota coming. Mixing in the misses, and the tasks you decided to keep for yourself, keeps the tone closer to "this is how I work now" and further from "this is how you will be measured."
The last one is asking the team to share their AI use before the leader has shared any. Coming first from the top, that request reads as monitoring, and the people most nervous about it will say the least. Go first for a few weeks, then ask, and expect the early answers to be cautious.
Five ways to go first this month.
None of these needs a budget or a new tool. Pick the ones that fit how you already run your week, and do them where your team will notice.
- Put one AI-assisted draft in front of the team and label it. A meeting agenda, a reply to a vendor or a first pass at a monthly report all work. Say which parts came from AI and what you changed before it went out.
- Tell one story about a miss. At a staff meeting, spend two minutes on something AI got wrong for you and how you caught it. Keep it plain, since the goal is to make checking sound routine.
- Run a real request live. During a meeting, turn a messy set of notes into action items, or ask for a cleaner version of an intake form, and let the room see the first answer before you fix it.
- Share one request worth reusing. When something works, post the exact wording in the team channel with a line on what it's for, so other people can start from it and adjust it to their own work.
- Name one thing you won't hand to AI, and say why. It might be performance reviews, a condolence note or anything holding a client's private records. Drawing a line in public makes the rest of your encouragement easier to believe.
Then do them again next month. The second round is usually when people start bringing their own examples to the meeting without being asked.
