Short answer
An AI champion is a staff member given time to help colleagues use AI tools well: answering questions, maintaining the team’s prompt library, logging what works and what fails, and running short show-and-tell sessions. Choose someone respected who does the work, not the loudest enthusiast, agree a time allowance in writing, keep policy and tool approval with management, and train them over six weeks on real workflows.
- Choose a respected practitioner with good judgement, not the most enthusiastic person.
- Agree a time allowance in writing, such as two hours a week for six weeks.
- The champion supports colleagues; management still owns policy and tool approval.
- Train on the team’s real workflows, with weekly practice and a use-case log.
AI use spreads fastest when someone inside the team shows the way. Written for managers and owners of teams of roughly 10 to 50 people who would rather not hire a specialist, this lesson ends with one internal AI champion appointed, with a written role card, a protected time allowance and a six-week plan built on your real work. Allow 40 minutes to plan; the programme runs for six weeks alongside normal duties.
Why peers teach AI better than experts
People adopt tools they see colleagues using on familiar work. TRA’s 2025 research for MBIE found that SMEs find success stories more convincing from peers who have walked the same path than from technology evangelists, and that confidence is built through small wins with real-world relevance (TRA for MBIE). The same research warns that training pitched too advanced intimidates people and training pitched too basic frustrates them.
An internal champion solves both problems. They know the work, they speak the team’s language, and they can pitch help at the right level for each person.
Choosing the right person
| Criterion | Look for | Watch out for |
|---|---|---|
| Credibility | Does the core work and is respected for it | Someone outside the work, such as IT or admin, who colleagues will not ask |
| Judgement | Notices when an output is wrong and says so | The loudest enthusiast who treats every output as impressive |
| Communication | Explains plainly, patient with beginners | Jargon, or impatience with slower colleagues |
| Capacity | Can genuinely give two hours a week | Someone already overloaded who will do it at night |
| Position | Peer level, so people admit what they do not know | A manager whose staff may hide their struggles |
Then agree the time allowance and the change to their duties with them in writing. Employment New Zealand lists changing an employee’s duties as a form of workplace change (Employment New Zealand), so treat the role as something agreed, not assigned.
The six-week plan
| Week | Focus | Output |
|---|---|---|
| 1 | Foundations: the first four lessons in this pathway, the AI policy, the approved tool | Notes on three things that surprised them |
| 2 | Two real workflows from their own team | A context pack and a working prompt for each |
| 3 | First 30-minute office hours | A use-case log with every question asked |
| 4 | Turn the best prompts into library entries | Three tested entries in the shared prompt library |
| 5 | A 20-minute show-and-tell with a real before and after | One colleague who tries it the next day |
| 6 | Review with their manager | Issues log, next three workflows, adjusted time allowance |
Exercise: set up your champion
- Score two or three candidates. Use the selection table. Pick the strongest on judgement and credibility, not enthusiasm.
- Draft the role card with AI. Use the prompt in the worked example, then edit it until every line is something you will actually support.
- Agree the role in person. Talk it through with the candidate, confirm the time allowance, and record it in writing.
- Book the six weeks. Put office hours, the show-and-tell and the week-six review in calendars now.
- Set up the two logs. A use-case log (task, tool, result, time saved) and an issues log (what went wrong, what was done). A shared spreadsheet is enough.
- Tell the team. A short message from the manager explaining who the champion is, what they can help with, and what they are not responsible for.
Worked example: a Timaru freight company
Illustrative example: a 30-person freight and logistics company in Timaru wants dispatch, accounts and customer service staff using its approved AI tool. The operations manager picks a senior dispatcher respected across the yard and the office, and drafts a role card:
Draft a one-page role card for an internal AI champion.
Business: 30-person freight and logistics company in Timaru, New Zealand.
Teams: dispatch, customer service, accounts, yard.
Champion: a senior dispatcher. Time allowance: 2 hours a week for 6 weeks,
then 1 hour a week, taken from their dispatch rostering duties.
Approved tool: our business AI plan only. The operations manager owns the AI policy.
Use these headings: Purpose; Time allowance; Responsibilities; Not responsible for;
Support they will get; How we will know it is working.
Under "Not responsible for", be explicit that the champion does not approve tools,
change policy, or fix colleagues' work for them.
Plain New Zealand English, under 350 words.
Example output (illustrative and abridged):
Not responsible for
- Approving new AI tools or changing the AI policy. Send those questions to the
operations manager.
- Checking or correcting colleagues' AI-assisted work. Each person stays responsible
for what they send.
- IT support, logins or account problems.
How we will know it is working
- At least 10 colleagues have used the approved tool on a real task by week 6.
- Three tested prompts are in the shared library.
- Every reported problem is in the issues log with an outcome.
The “Not responsible for” section is the one most teams skip, and the one that protects the champion.
Where champion programmes go wrong
- The unpaid help desk. Without a time allowance and clear limits, the champion fixes everyone’s work and burns out. Protect the hours.
- Enthusiasm over judgement. A champion who praises every output teaches the team not to check. Choose the sceptic who still sees value.
- Policy by default. If nobody owns policy, the champion ends up deciding what is allowed. Keep that with management, as set out in creating a team AI use policy.
- A single point of failure. If the champion leaves, everything leaves with them. Keep logs and the library in shared locations, and consider a second champion after three months.
- Measuring activity, not use. Office hours held is not the goal; colleagues doing real work better is. The lesson on measuring adoption after training shows how to check.
The AI training needs assessment helps the champion see where colleagues are starting from, and the AI learning plan for teams gives a structure for the months after week six. If you would rather give the champion a running start, our AI training for business teams can train them alongside the team; the comparison of workshops versus self-paced courses helps you choose.
Check your set-up
- I scored candidates on credibility, judgement, communication, capacity and position.
- The role card says what the champion is and is not responsible for.
- The time allowance is agreed in writing and taken from existing duties.
- Office hours, show-and-tell and the week-six review are in calendars.
- The use-case log and issues log exist in a shared location.
- The team has been told who the champion is and what they can help with.
Next lesson: Rolling out AI to a small team covers the first 30 days for everyone else.
Sources
- From daunting to doable: positioning AI in reach for SMEs, TRA for MBIE, July 2025
- The basics of workplace change, Employment New Zealand, last modified 13 November 2025
Written by the Levy Street team in Wellington. Last updated 19 October 2026.