How to Pick AI Tool Champions on Your Team

How to pick AI tool champions on your team comes down to one counterintuitive call: skip the person already evangelising the tool in the team chat, and recruit the one who pushed back hardest when it was proposed. That instinct feels backwards. It works anyway.

A champion's real job is getting a dozen sceptical colleagues to trust a new way of working. The person who needed zero convincing is rarely the person other people believe — what does a software champion do, at bottom, is vouch for a tool to people who trust their judgement more than a rollout email.

This guide covers who to pick, how many you need per team, what to hand them to do each week, and how the model changes as headcount grows. It picks up after the "should we roll this out" conversation covered in change management for AI email adoption.

Should Champions Be Early Adopters or Sceptics?#

The obvious pick is your most enthusiastic user — the person who signed up before the rollout was announced and now brings the tool up unprompted. Enthusiasm is real signal, but it isn't the signal that changes anyone else's mind. Early adopters convince other early adopters; they rarely move the middle of the team, who assume the loud fan just likes new things.

A better pick is the person who raised the sharpest objection when you first floated the idea — the one who asked what happens when the agent misreads a client's tone, or whether a draft can be stopped mid-send. If you gave them a real answer and they came around, they now carry something an enthusiast can't: they already ran the objection your other sceptics haven't voiced yet, and they know the answer holds up.

This maps onto Prosci's ADKAR model for change. Awareness and Desire get built top-down, by you and by training. Reinforcement is what keeps the change from reverting once the novelty wears off, and reinforcement is a peer job, not a manager job — a converted sceptic reinforces the change every time a colleague asks "does this actually work" and means it.

This isn't just change-management folklore, either. NIST's AI Risk Management Framework lists a person closely reviewing an AI system's output as one of the more effective ways to build warranted trust in it — which is functionally what a champion does every time a colleague watches them approve or edit an AI-drafted reply in real time.

• Respected for judgement, not tenure — people ask their opinion before making a call
• Raised a real objection early and accepted a real answer, not a talking-to
• Does the same kind of email work as the team they'll influence
• Has two to three spare hours a week — not the person already drowning
• Isn't already on the IT or ops team; a champion who's "supposed to like this" carries less weight

How to Recruit and Set Up Champions, Step by Step#

Once you know the profile you're recruiting for, the setup itself is mechanical. Here's the sequence that gets a champion functional inside two weeks, not two months.

1. 1

   Ask, don't appoint

   Approach candidates privately and ask if they'd take the role. Being volun-told reads as a demotion to helpdesk duty, and it shows in how they talk about the tool afterward.

2. 2

   Give them a real training session

   Thirty to forty-five minutes, hands-on, in their own mailbox — not a slide deck. A champion training plan that skips hands-on time produces someone who can repeat the pitch but can't answer "why did it file this under the wrong label."

3. 3

   Hand them a one-page cheat sheet, not the full manual

   The five things people actually ask in week one: how to undo a send, how to check what the agent drafted before it went out, how to turn a feature off. Point to the full docs for everything else.

4. 4

   Put a name on the incentive, even a small one

   Incentives for internal champions rarely need to be cash — a line in the next review, early access to new features, or public credit when the rollout goes well is usually enough. What kills the role is no incentive at all: real, unpaid extra work resents itself.

5. 5

   Set a standing fifteen-minute weekly sync

   Not a status meeting — a chance for the champion to flag the same question coming from three different people, which usually means a training gap, not three separate problems.

6. 6

   Connect champions to each other, not just to you

   Past about three champions, a peer champion network for AI adoption — a shared channel where they compare notes — surfaces a recurring question faster than routing every one of them through you.

How Many Champions You Need, by Team Size#

The right number of champions scales with team size, not headcount alone — a champion can realistically stay close to fifteen to twenty people before the weekly sync stops being personal and starts being a broadcast. Below that ratio, one champion works. Above it, the role needs to split or become a small network.

What to Do When a Champion Isn't Working#

The most common failure isn't outright refusal — it's a champion who says yes, does the training, and then goes quiet. That usually means the role never had a real weekly task, only a title. Fix it by giving them one concrete thing to check each week, such as how many drafts their closest colleagues approved unedited, rather than a vague "help people out" mandate.

The second failure is a champion who becomes the only person anyone asks, including for problems that have nothing to do with the rollout. If tickets about a forgotten password start landing in their inbox, that's a sign the escalation path was never actually communicated to the team — not a sign the champion is failing at the job.

The third is a champion who quietly stops trying because nobody mentioned the rollout again after the kickoff. Reinforcement fades fastest for the person doing unpaid extra work, so the fix is usually not a new champion — it's the incentive from step four above, applied a month later than it should have been.

If a champion genuinely isn't working after a fair attempt at all three fixes, replace them without ceremony. The role only functions with someone the team already trusts, and forcing a bad fit to continue costs more credibility than a quiet swap.

A Faster Way to Get the Same Adoption#

Champions solve a real problem: people trust a colleague's word over a rollout memo. What they can't solve alone is the ongoing coaching load — a champion who has to check five colleagues' drafts every week just to know the tool is working is doing manual quality control by hand, indefinitely.

AI Emaily's Copilot mode gives every user that visibility directly, without routing it through a champion. Every AI-drafted reply sits for approval before it sends, and the audit log shows exactly what the agent proposed and what a person changed — the same trust-building your champion does in a weekly sync, just visible to everyone at once instead of relayed secondhand. Champions still matter for the human side of adoption; they just stop being the only source of proof the tool works. We build AI Emaily.