How to Build AI Champions and Change Agents Inside Your Organisation

The most common selection error for AI champion networks is choosing the most technically confident employees. Technical confidence helps, but it is not the primary quality that makes an effective champion.

Effective AI champions share three characteristics: peer credibility (they are respected by their colleagues as competent professionals, not just as technology enthusiasts), genuine AI adoption (they are actually using AI in their own work in ways that produce value, not just theoretically supportive of the idea), and communication willingness (they are willing to share their experience openly, answer questions, and spend time helping colleagues, not just use AI privately).

The search for these individuals should focus on: people mentioned by their colleagues as good sources of AI advice, people who are demonstrating AI use in team meetings or shared outputs, and people who have expressed interest in AI adoption without being pushed. Voluntary champions who meet these criteria outperform recruited champions who may lack genuine adoption or peer credibility.

Champions need more support than most organisations provide them.

Advanced training: champions need deeper AI proficiency than their colleagues. They will be asked questions that go beyond the basics, and their credibility depends on being able to answer them. A dedicated champion development programme, running two to four weeks ahead of the broader workforce training, gives champions the depth they need.

Content resources: a regularly updated prompt library, FAQ document, and collection of internal use case examples that champions can share with colleagues without having to create their own materials. Champions who have to create their own resources will gradually stop providing support as the effort exceeds the motivation.

A champion community: a regular (monthly or bi-monthly) meeting of all champions across the organisation, sharing what is working, addressing common questions, and building the peer relationships that sustain champion motivation. Champions who are isolated within their own team lose energy faster than those who feel part of a broader movement.

Recognition: champions are giving time and energy to support the organisation's AI adoption programme on top of their existing role. Visible recognition (from the programme lead, in internal communications, in annual review discussions) is necessary to sustain that commitment over the 12 to 24 months that meaningful AI transformation requires.

What should champions actually do? Clarity about the champion role prevents burnout and ensures the activity adds value.

Core champion activities: running informal 30-minute 'AI coffee sessions' for their team (showing two or three practical use cases, answering questions, collecting feedback), maintaining a local prompt library and sharing updates, answering colleagues' AI questions informally, and feeding usage observations and feedback back to the central programme team.

Champions should not be expected to: solve technical IT problems (that is IT's role), provide legal or compliance advice about AI use, or be accountable for their team's AI adoption rate. Scope creep into these areas reduces champion effectiveness and increases burnout risk.

For Copilot deployments specifically, champions who run the Microsoft-provided Copilot 'lunch and learn' sessions with their own organisation-specific examples (rather than the generic Microsoft content) consistently produce better adoption outcomes. The investment in helping champions customise these sessions to their team's specific work context pays back in adoption rate.

The initial champion network (first wave, typically covering 20-30% of the licenced population's teams) should be evaluated at three months before scaling.

Evaluation criteria: team adoption rates in champion-supported teams versus non-champion teams; champion retention rate (are the original champions still active and motivated?); and quality of champion feedback (are champions providing useful operational intelligence back to the programme team?).

If the champion model is working, the second wave should expand it to the remaining teams, using first-wave champions as mentors and advocates for the second-wave cohort. Peer recruitment is more effective than central recruitment for second-wave champions: asking existing champions to identify strong candidates in teams where no champion yet exists produces better matches than a central call for volunteers.

Champion networks require maintenance. At 12 months, refresh the champion training (new AI capabilities, updated use cases), rotate out champions who have become inactive, and identify champions who have developed enough to take on a more significant role (subject matter lead, CoE contributor, or similar advancement pathway).