AI Champions Program: How to Pick Champions and Keep Them Going
Champions keep AI in daily work after training ends, if you pick them for coverage and credibility and give them time. A scorecard, an 8-week plan and what to measure.
Pick the people for your AI champions program by the workflows they cover, the trust colleagues already place in them and their curiosity. Then agree their time with their managers before anyone is announced. Enthusiasm helps. But a keen volunteer with no protected hours and no part in a real workflow can slide into the role of unpaid help desk. Once the launch is over, they can drift away. Below is the scorecard we recommend for choosing champions, an 8-week plan for their first two months, and the measures that show whether the program is working.
What an AI champion does, and what they should not do
An AI champion is a colleague inside a business team who keeps approved AI tools in real use after formal training ends. They show peers how the tool fits their tasks, collect what is not working, and look after the shared prompts, templates and skills their team relies on. Some organisations call the group an AI champions network; the name matters less than the job.
Microsoft's adoption guidance draws the line clearly: champions are "not an IT support function but business representatives" (Microsoft Learn, service adoption module). We agree, and we take it one step further: write the split down, because without it, sign-in problems can land on the champion's desk.
| Request | Who handles it |
|---|---|
| "How do I use the assistant for our monthly variance commentary?" | Champion |
| "Our shared prompt for tender summaries gets the dates wrong" | Champion, who fixes it or raises it with the program lead |
| "I cannot sign in" or "I need a licence" | IT service desk |
| "May I paste this customer file into the tool?" | The data rules in your AI policy, then the policy owner |
| "The tool gave a customer a wrong answer" | The incident route in your AI policy |
Put this table into a one-page champion charter, signed by the champion, their manager and the program lead. It tells colleagues where each kind of question goes.
How to choose AI champions: a selection scorecard
Score each nominee 0, 1 or 2 on six criteria. Ask the line manager and one peer to score separately, then compare the two.
| Criterion | What a 2 looks like | Why it matters |
|---|---|---|
| Role coverage | Does the core work of a workflow in scope, for example drafting credit memos or answering supplier queries | Champions teach from real tasks; someone outside the workflow teaches theory |
| Credibility | Peers already ask them for help on work questions; respected for accuracy, not for rank | People copy colleagues they trust |
| Curiosity | Has tried the approved tool on their own work and can describe one thing that failed | Curiosity that admits failure gives honest guidance |
| Judgement on data | Knows what is confidential in their team and follows the AI policy without reminders | Champions set the example for what goes into the tool |
| Manager backing | Manager agrees in writing to protected time | Without it the role becomes after-hours work |
| Language reach | Can explain the tool in the language the team works in: Arabic, English or both | Guidance has to match how the team writes and speaks |
Two rules sit on top of the score. First, cover workflows, not the org chart: at least one champion for each workflow and team in scope, and two where a team works shifts or across sites. Second, do not pick only the loudest early adopter. Someone who scores 2 on enthusiasm and 0 on credibility gives good demos but changes little. Mix seniority too: a respected senior officer can move a team further than a new graduate who is quick with tools.
An illustrative example: a retail bank in Amman rolls out an AI assistant to credit, compliance and branch operations. It receives fourteen volunteers, ten of them from IT and digital teams. After scoring, it keeps three of the ten and adds people who write credit memos and handle branch complaints, because those are the workflows in scope.
The 8-week plan: what champions do and what managers give
Run the first cohort over 8 weeks in four two-week blocks. Each block pairs what champions do with what their managers and the program lead give in return. We recommend a fixed allowance of protected hours each week, written into the champion's objectives. Agree the figure with each manager rather than expecting the role to fit into spare minutes.
| Weeks | Champions do | Managers and the program lead give |
|---|---|---|
| 1-2: Set up | Complete training built on their own tasks; choose two or three workflows with their manager; record how each runs today (time, steps, common errors) as the baseline | Signed charter and time allowance; access to the approved tool; a champions channel in Teams, Slack or the tool your organisation uses; a named program lead |
| 3-4: Build | Build and test shared prompts, templates or skills for the chosen workflows; try each one with two colleagues; log what fails | A fortnightly champions call; a reviewer from risk or compliance for anything that touches personal or customer data; time to test without targets |
| 5-6: Spread | Hold short office hours for their team; sit with colleagues on live tasks; collect questions and blockers in the channel | The manager attends one session; blockers answered within the week; first public recognition, such as a mention in the team meeting or internal newsletter |
| 7-8: Hand over and measure | Write a one-page guide for each workflow; compare results against the week 1-2 baseline; present findings to the sponsor | A review with the sponsor that decides what to keep, fix or stop; the champion's contribution recorded in their performance review; a date for the next cohort |
Training and an Arabic-language example
The training in weeks 1-2 works best when it is built from each role's real tasks, as we set out in our guide to role-based AI training for employees.
A second illustrative example: a semi-government entity in Riyadh introducing an AI assistant for Arabic correspondence picks champions from its correspondence, legal and customer service units. Because the work is in Arabic, the champions build their shared prompts and guides in Arabic from the start. The legal reviewer joins the week 3-4 calls to check templates that quote regulations.
How to keep AI champions going after training
The 8 weeks are the easy part. What keeps a champions network alive is a steady rhythm and visible return on the time people give.
- Keep a monthly meeting. Microsoft's guidance on champion programs recommends regular monthly meetings and says consistency is key (Microsoft Learn, Teams champions program).
- Make recognition count. Thanks in a newsletter helps; a line in the annual objectives and the performance review helps more. The same Microsoft page notes that employees often appreciate recognition rather than monetary gain.
- Act on what champions report. If a champion raises the same blocker three months running and nothing changes, expect them to stop raising it. Show which decisions their reports changed.
- Set a term. Appoint champions for 6 or 12 months, with a named successor.
- Watch for help-desk drift. Ask champions each month how their hours were spent. If access and sign-in problems take a large share, fix the service desk route instead of asking champions to absorb it.
If the rollout itself has stalled, champions alone will not restart it. Our article on why AI rollouts stall covers the causes that sit above the champion level, such as unclear workflows and no owner inside the business.
What to measure in an AI champions program
Measure three layers, and do not stop at the first one.
- Use. Compare active users in teams that have a champion with teams that do not. Vendor dashboards give you the raw numbers. The Microsoft Copilot Dashboard in Viva Insights covers readiness, adoption, impact and sentiment metrics. Its readiness, adoption and impact pages show the previous 28 days (Microsoft Learn, page dated 29 September 2026). Claude Team and Enterprise plans include usage analytics such as weekly active members, adoption level and estimated time saved. Plan owners can see them and, on Enterprise, so can admins (Anthropic Help Center). Check what your own plan shows before you promise a report.
- Workflow results. For each workflow a champion owns, compare time per task, rework and error rates with the baseline recorded in weeks 1-2.
- Program health. How many champions are still active, and how many hours they spend and on what. How many questions they answered, how many items they escalated, and how many shared prompts and templates are in use.
Treat dashboard numbers with care. An active user opened the tool; that does not show the work got better. A vendor's time-saved figure is an estimate built on the vendor's assumptions, so set it beside your own baseline before you quote it to the board.
How to start your AI champions program
Start with one cohort tied to the workflows you are already rolling out, rather than a company-wide call for volunteers. Score the nominees, sign the charters, then run the 8 weeks and decide at the end what to keep.
Our AI Champions Program in the Academy Lab turns early adopters into champions who maintain workflows and support colleagues. It pairs well with team training built on real tasks. If licences are already bought but barely used, our 30-Day Adoption Rescue sets champions and adoption metrics as part of the fix. Not sure where to begin? The free AI Exposure Scan is a 45-minute conversation that ends in a written next step.
Frequently asked questions
How many AI champions does an organisation need?
Plan by workflow, not by headcount. We recommend at least one champion for each workflow and team in scope, and two where a team works shifts or across several sites, so there is cover during leave. Start with the workflows in your current rollout and add champions as new teams join, rather than recruiting across the whole organisation on day one.
How much time should an AI champion get?
Agree a fixed number of protected hours each week with the champion's manager and write it into their objectives before the program starts. The exact figure depends on the size of the team and the number of workflows they look after. What matters is that the time is agreed in advance; a role that has to fit into spare minutes tends to fade after launch.
Should AI champions be paid extra?
Extra pay is a decision for each organisation. Microsoft's guidance for champion programs notes that employees often appreciate recognition rather than monetary gain. Whatever you decide, make the role visible: record it in the performance review, thank champions publicly, give them early access to new features, and act on what they report so they can see their time changes decisions.
What is the difference between an AI champion and the IT help desk?
An AI champion is a business colleague who helps peers use approved AI in their own work, looks after shared prompts and templates, and reports what is not working. The IT help desk handles access, licences, sign-in and technical faults. Keeping the two apart, in a written charter, stops champions becoming an unpaid help desk.
How do you measure whether an AI champions program works?
Measure three layers. First, use: active users in teams with a champion compared with teams without one. Second, workflow results: time per task, rework and errors compared with a baseline recorded at the start. Third, program health: how many champions are still active and how they spend their time. Usage counts alone do not show that work improved.
Get new insights by email
One email when we publish something useful. No spam; leave any time.
Services in this article
Related lab: Academy Lab ↗
