Define what AI means before your employees do.
Silence isn't neutral. When the subject is whether your job exists in two years, people don't wait for the official version. They write their own.
Silence isn't neutral. One of the leaders we spoke with this summer put it more precisely than I could have. Inside a strong culture, her engineers were anxious about their relevance under AI, and the absence of a visible people plan was being read as a statement about what the organization values.
Read that again. Not "we don't have a plan yet." A statement about what the organization values.
That's how it works in every organization I've been part of. Leaders think of a pending decision as a blank. Employees experience it as a message. And when the subject is whether your job exists in two years, people don't wait for the official version. They write their own, from whatever evidence is lying around. A headline. A vendor demo. A colleague who just automated half their week. A leader who uses the tools constantly and hasn't said what that means for anyone else.
The version they write is almost always worse than the truth. It's also stickier, because it got there first.
A few things I'd tell any leader about to roll out AI, drawn from what we've heard this year.
Say what you know and what you don't
"We don't know yet how this changes your role, and here's when we'll know more" is a plan. Not a great one. But it beats silence by a mile, because it tells people the question is being held by someone.
Put the people plan next to the technology plan
In a strong culture the workforce can't separate them, so don't try. If the AI announcement doesn't contain a sentence about what happens to the people whose work changes, that sentence gets written for you.
Decide the two things that actually need deciding
Most organizations don't need a comprehensive AI policy. They need a position on data and a position on disclosure. What can go into these tools, and what do we tell the people we serve about AI's role in the work. Decide those two, say them out loud, and trust your people with the rest. When we ran this conversation with our own team, one of our most capable practitioners argued against over-regulation, and he was right. One bad actor, or the fear of one, shouldn't ruin AI for everyone else.
Show, don't announce
Enthusiasm from leadership is not modeling. People need to see someone they respect use the tool on real work, including the moments it fails. Until they do, the message is positive but unanchored, with nothing concrete to calibrate against.
Listen before you define
The organizations that got this right this year found out what their people already believed before writing the official version. Because the official version has to answer the one they'd already written, and you can't answer a story you haven't heard.
Define it first. Or it gets defined for you, and not by you.