1Pick the person, don't recruit them.
You are looking for someone already doing this without being asked. The signals are small: they bring you things you did not request, they mention using AI for something outside work, they have opinions about which tool is better. Enthusiasm in a meeting does not count. Doing it on a Sunday does.
If nobody comes to mind, you do not have your person yet, and no amount of training will manufacture one. Wait and watch.
2Name what comes off their plate.
"Find some time for this" is not room. It means do it on top of everything else, which means at night, which means it only happens if they are the kind of person who does things at night.
Room is specific. A recurring meeting they stop attending. A report someone else picks up. A project that slips a month. Write down the actual thing. If you cannot name it, you have not made room. Two hours a week is enough to start. It is small enough that you do not need to ask anyone's permission for it.
3Tell them the time is theirs.
Most people hide time they save, because finishing early has historically earned them more work. If you free somebody up and then quietly fill the gap, they learn the lesson fast and you never get the hours back.
So say it out loud. The time is theirs to build with, and using it well will not cost them.
4Ask what they tried, not what they finished.
The first thing will probably not work. That is the normal shape of this, not a sign you picked wrong.
Set a check-in at thirty days and ask what they tried and what broke. If you ask what they shipped, you will get a status update and they will stop taking risks. If you ask what they tried, you will find out whether the room is working.
What to expect
Weeks one and two: not much. They are figuring out what is even possible, which is most of the work and none of the output.
Around week four: something rough that solves a problem only they had.
After that: somebody else sees it and asks for their version. That is the moment it stops being a personal project.
Copy this and send it
Hey. I want you to spend two hours a week building things with AI. Not on anything I've assigned. On whatever problem you keep running into. I'm taking [[ the thing ]] off your plate to make the room. That time is yours and using it well isn't going to cost you. Most of what you try won't work, and that's fine. Let's talk in a month about what you tried, not what you finished.
Keep going
You might like these
Give your whole team the same context.
Put your business knowledge in one place so your whole team can use it with AI.
Read guide →Stop re-explaining yourself to AI.
Set your business context once and get better answers from any AI tool.
Read guide →The Execution Gap.
Why good processes fall apart between documentation and what happens each day.
Read guide →