Why champion networks often fade
Many organisations start an AI champions network. A call goes out for volunteers, a channel is created, a launch session is held, and for a few weeks the channel is busy. Then the champions’ day jobs reassert themselves, the examples shared at launch go out of date, colleagues bring questions the champions are not sure they are allowed to answer, and the network quietly stops.
The idea is sound. The failures are practical: champions were chosen for enthusiasm rather than credibility, nobody protected their time, they were asked to teach generic tips rather than their own team’s work, there was no route for the questions they could not answer, and no one owned keeping the material current.
This programme is designed around those failure points.
What the programme involves
A role with a definition. Before anyone is selected, the sponsor and I write down what a champion does (teach colleagues one or two workflows, answer first-line questions, keep their library entries current), what they do not do (approve data use, configure the tool, judge colleagues), and the time allowance their manager agrees to.
Selection against a rubric. Managers nominate; the rubric decides. Illustrative criteria: respected by peers for the quality of their own work; does a recurring task that suits AI assistance; shows good judgement about confidential information; willing to maintain material for at least six months; manager has agreed the time.
Workflows from their own work. Each champion redesigns one recurring task from their team with me: the inputs, which data may go into the approved tool, what the tool is asked to do, and what the reviewer checks. They use it on live work before they teach it, so they teach from experience.
Teach-back. Champions first rehearse teaching their workflow to other champions, with structured feedback. Then they teach their own colleagues while I observe, and we debrief afterwards. Teaching once under observation does more for confidence than any amount of slide material.
Escalation. Champions will be asked things they should not answer alone: “Can I put this client contract in?” “The output was wrong and went to a customer.” “Can we connect it to our CRM?” The escalation map gives each question type a destination in IT, data protection or management, agreed with those teams.
The signature deliverable, illustrated
You receive a champion selection rubric, teach-back exercises, escalation map and maintained workflow library. Illustrative extract from an escalation map:
| Question or event | Champion handles? | Goes to | Expected response |
|---|---|---|---|
| How do I phrase this request better? | Yes | — | On the spot |
| May I use client personal data in the tool? | No | Data protection lead | Within two working days |
| AI-assisted output sent with an error | Log and inform | Team manager, then data protection if data was involved | Same day |
| Request to connect the tool to another system | No | IT service desk | Normal change process |
Illustrative example. It shows the format, not a client’s map.
How acceptance is judged
The sponsor and programme lead accept the programme when the cohort has been selected against the rubric, each champion has a documented workflow in use in their team, each has delivered an observed teach-back to colleagues, the escalation map has been agreed by the teams it routes to, and the library has an owner and review dates. Repeat use of champions’ workflows in their teams is measured at the end and handed to the programme lead to keep measuring.
Ownership and handover
The internal programme lead owns the community, the library and the next cohort. Champions own their library entries. I hand over the rubric, the teach-back exercises, the escalation map and a short guide to running the next cohort without me; the first cohort usually helps teach the second.
When to choose something else
If adoption has stalled for reasons nobody yet understands, diagnose first with adoption rescue. If no one is available to run the programme internally, a fractional AI enablement lead can hold that role while a successor develops. If one department needs intensive help on its own workflows, see practical AI enablement for business teams.