Somewhere in your organisation a leader has said some version of this: "We want everyone experimenting with AI."
Those words give permission. Permission and direction are different messages. Permission says you may. Direction says what for, what good looks like, what stays with a person, and who to ask when you are unsure. A team given only permission will use the tools, because that is what permission invites. Whether the use adds up to anything is left to chance.
I particularly liked one finding from Culture Amp's 2026 AI at Work Benchmark on the gap between being encouraged to use AI and knowing why. 71% of employees say using AI makes them feel more productive, and increases to 93% with power-users. It's cute to be fair, but there is a LOT more a team can do with AI than just feeling more productive!
A test you can run this week
Ask five people on your team to finish this sentence without talking to each other: "We use AI so that ..."
If you get five different answers, or five blanks, the message has not landed, however many times it was said.
What a message with direction contains
The goal. Which outcome the use of AI serves. Faster first drafts, fewer review cycles, more time with clients. Pick the one that matters.
The standard. What good output looks like in this team, and what bad output looks like. Without it, everyone applies a private quality bar.
The boundary. Which decisions and which checks stay with a person.
The route. Where the examples live and who to ask.
This is the same shape as the AI enablement work I do: context, output definition, evaluation and human oversight. The tool matters less than whether these four things have been said out loud. I wrote about the context part earlier in How AI Can Help Your Operations, If You Stop Treating It Like a Mind Reader.
Leaders sometimes hold back because spelling this out feels like micromanaging. A team cannot aim at a goal nobody stated. Saying the goal and the standard is what lets people use their own judgement inside them.
If you asked your team to finish "We use AI so that ..." tomorrow, how many different answers would you get?
Sources: Culture Amp, 2026 AI at Work Benchmark.
