Enablement programs, including champion programs, aren’t new. I didn’t invent them. Even OpenAI and GitHub both publish activation playbooks for them. An in-house facilitator bringing your AI enthusiasts together is a great way to cross-pollinate use cases, share techniques, and build community. But neither of those references give you facilitation guides or assets to run the program. They’re good ideas. But relying on one person to build technical, change management and train-the-trainer skills that stick is a lot to ask.

No matter how strong your vision or how detailed your plan, the first cohort teaches you things planning can’t. And with the sessions already on the calendar, you learn them while the program is running.

What I learned from running my first AI Champions cohort

A client brought me in for a 12-month program: build a community of 24 internal AI Champions who could influence how AI was used in their teams from the inside. The premise: they were already SMEs who understood how work gets done on their teams, and were AI experts, but they didn’t know how to turn those two things into team-level influence or capability in their peers.

We documented what it meant to be an AI Champion and asked people to self-nominate. We curated slightly, to make sure every part of the business was represented and each team only had one champion. Then with the champions selected, we kicked off a survey to understand their existing skillset and what they wanted to learn. It revealed our champions weren’t AI experts (yet). They were beginners with a desire to learn and influence how AI is used in their teams.

Did you catch that first mistake? The application process was thin. We made sure there was only one individual per team, but we didn’t assess skill or job level. That meant our cohort weren’t generally AI experts, nor were they senior enough to influence the way their team works.

The second mistake was baked into the format. Half of every 60-minute session was reserved for champions to share learnings and build community, a design built for the experts I thought we’d have, with techniques and wins ready to trade. That wouldn’t work for beginners. Instead we expanded to 90 minutes, most of it spent teaching skills and frameworks to build capability fast.

The third mistake was the role documentation: too high level, never updated once we realized our champions’ actual skill level, and never written side-by-side with the other essential partner in team-level AI capability, the people manager. Different visions of the role surfaced while the program was already in-flight, and I had to redo not just planned lessons, but homework I had already assigned (ouch).

I quickly pivoted: champions could still create value by modeling good AI use in their own work. That bought time to develop their expertise and figure out when to re-introduce training, documentation, and experimentation on team workflows.

I worked closely with my day-to-day partner, the VP L&D, to document the champion role explicitly, and especially how it partnered with the people manager. Besides creating alignment with our executive sponsor the document gave us a lens to evaluate programming decisions. It allowed us to discuss what skills were achievable and which tradeoffs were worth making within a 12-month program, with the commitment champions could honestly give.

And it surfaced a bigger finding: people managers needed their own training program. Their role in bringing AI into the organization intentionally, rather than piecemeal, is even more critical than the champions’. In the meantime, instead of generic email blasts managers would likely delete, I encouraged 1-1 discussions between champions and managers, with specific goals and agendas in mind.

Within 30-60 days of teaching a new skill I was hearing first-hand accounts of champions applying it to their work: the notebooks they created after I demoed what notebooks are useful for, the competitive analysis built from a project brief after a session on creating persistent context for ongoing projects.

Six months into the program became an inflection point. We re-confirmed the role with champions and asked what barriers were getting in their way (a story for another issue). And now that they had core AI skills, I had each champion build an individualized AI Action Plan: pick the projects for the next six months with their manager, and I would teach based on the additional skills needed.

Although I had guided them to create a plan based on the work they do as individuals, most of the projects they’re planning will impact their teammates as well: reusable assets, notebooks for shared projects, or an agent. I already knew they were engaged, over 80% have attended nearly every session, but this is work outside our monthly sessions, requiring the skills they’ve been taught. They are building the capability the program intended to build and I am thrilled.

The goal was always the same: build AI skills and integrate AI into the work the team does. What it looks like to get there changed. Twice. How you respond to the challenges that come your way is what determines if the program is ultimately successful, or if the first cohort becomes the last.

Looking forward to future cohorts

As we enter the final few sessions and I see evidence of the program working, the bigger realization for me is that this program shouldn’t look the exact same for every company or every cohort. Its shape will flex based on the seniority of its participants, their existing AI skills, and, most importantly, how champions partner with their managers in the transformation process.

That partnership document is now a public resource: The AI Partnership Map, six domains showing where a champion’s responsibility ends and a manager’s begins. If you’re building a program, start there.

-Mariena