Use this whenYou’re running AI training or a pilot and want to find the people who can carry it into their own teams.

Why it exists

Training gets people started, but the people who make AI useful inside a team are usually subject matter experts who already know the work. They don’t announce themselves. Without a deliberate search, you find the loud ones and the technical ones, and miss the quiet person who knows the job better than anyone.

The framework

  1. Step 1SpotWho is already leaning in?
  2. Step 2StartWhat does a champion do first?
  3. Step 3GrowWhere does it lead?
  4. Step 4KeepWhat makes it sustainable?
  1. Spot

    Who is already leaning in?

    Every session is also a search. Watch for the people who come back with questions, already use AI at home, or build something without being asked.

  2. Start

    What does a champion do first?

    Give them one real problem, a mentor, and a building kit, with a place to show what they made. A small win in the first weeks matters more than a plan.

  3. Grow

    Where does it lead?

    Move them from personal shortcuts to shared tools, and eventually into builder work on execution projects. Many of the best come from non-technical roles.

  4. Keep

    What makes it sustainable?

    Champions need time on their calendar, recognition, and a manager who supports it. Without them, the role is extra work on top of a full day.

How to use it

  1. Use training sessions, councils, and show-and-tells as a search. Note who comes back.
  2. Invite likely champions and ask what they would work on, rather than assigning them a topic.
  3. Pair each with a mentor, give them a building kit, and agree on one real problem.
  4. Have them show what they built, and bring in the people around them.
  5. Make the time and the recognition explicit with their managers.

Signals to watch for

  • They come back with questions after the session ends.
  • They already use AI at home, and find the introductory material too basic.
  • They build something without being asked.
  • They explain it to their peers in plain language.
  • They know the work deeply enough to say what a good result looks like.

A typical first 30 days

  • Week one: meet a mentor, get a building kit, and name one real problem from their own work.
  • Weeks two and three: build the smallest useful version, and test it on real work.
  • Week four: show it to the group, say what it saved, what it costs to check, and who else could use it.
  • This is a typical shape to adjust, not a fixed program.

Keeping them

  • Time: protect it on the calendar, so the work is not done after hours.
  • Recognition: show what they built to leaders, and put it in the tool registry under their name.
  • Manager support: their manager agrees to the time and sees the result.

In practice

At Rivian I built AI councils and a champions program to find people who combined subject matter expertise with an aptitude for AI. Many came from non-technical roles and knew the work better than anyone. With mentoring and building kits, they moved from personal shortcuts to shared tools that others depend on. Some went on to build execution work.

Where it breaks down

Don’t pick only the loud or the technical. Quiet experts are easy to miss, and they often matter most. Champions burn out when the role is extra work on top of a full job, so it fails without protected time. And a champion without a real problem to solve is a title, not a builder.

Try it with AI

Help me find AI champions in my organization.

Ask me one question at a time about our training sessions, our teams, and who has stood out so far. Then help me look for these signals: comes back with questions, already uses AI at home, builds without being asked, explains it to peers, knows the work deeply.

Then draft a first 30 days for one likely champion: a mentor, a building kit, one real problem, and a show-and-tell. Finally tell me what time, recognition, and manager support they will need to keep going, and who I might be overlooking.

Related