Short answer
A champions program turns a one-off agentic coding workshop into lasting practice. Select engineers who are respected rather than just enthusiastic, give them a clear mandate and protected time, connect them in a community with shared assets, and measure adoption and quality instead of activity.
On this page
Why workshops alone fade
A good workshop changes how people work for a few weeks. Then deadlines return, the trainer is gone, and questions start piling up: how do we handle this kind of task, why did the agent do that, is it allowed to use this tool for that repository? Without someone nearby to answer, people fall back on old habits or develop private workarounds.
Champions fill that gap. They are engineers inside product teams who have deeper skills with coding agents, a mandate to help colleagues and a direct line to the people who set policies. At neooutput we often help organizations set up such a program as the second phase after initial workshops.
Choosing the right champions
The instinct is to pick the most enthusiastic early adopters. That is part of it, but enthusiasm alone is not enough. Look for a combination:
| Quality | Why it matters |
|---|---|
| Respected by peers for engineering judgement | Colleagues follow advice from people whose code they trust |
| Curious and willing to experiment | They will discover patterns before others |
| Sceptical in a healthy way | They will flag problems instead of overselling |
| Good at explaining | Teaching is the core of the role |
| Spread across teams and locations | Coverage matters more than concentration |
One champion per team of roughly 6 to 12 engineers is a good starting ratio. Include at least one champion from platform or security, because many questions touch their areas.
The mandate
Champions need a clear, written mandate so that their role is understood by managers and colleagues. A typical mandate covers:
- Help colleagues with questions, pairing and reviews of agent-assisted work
- Maintain shared assets such as context file templates, prompt patterns and examples
- Collect feedback on tools, policies and pain points and bring it to the platform team or leadership
- Run short sessions for their team: demos, lunch-and-learns, pairing sprints
- Flag risks they notice, such as unsafe practices or policy gaps
What champions should not be: unpaid support desks, policy enforcers or the only people allowed to use the tools.
Protected time
The most common reason champion programs fail is that champions have no time. Agree on a fixed budget with their managers. Ten to twenty percent of working time is a common range for the first months, decreasing once routines are established. Without explicit time, the role becomes a burden and the best people quietly drop out.
Training the champions
Champions need more depth than the standard workshop provides. A champion track typically includes:
- Advanced use of the organization's approved tools: planning, subagents, headless runs, permissions and hooks
- Context engineering for different repository types; our CLAUDE.md and AGENTS.md workshop is a common building block
- Security and governance, so they can answer "are we allowed to" questions correctly; see the secure agentic coding workshop
- Facilitation skills for running sessions and pairing
- How to measure and report adoption honestly
We usually run this as two to three days spread over several weeks, with real work between sessions.
Rituals that keep the community alive
A champions network needs regular contact to stay effective:
- Bi-weekly champions call (45 minutes): new patterns, problems, policy questions
- Shared channel for quick questions and sharing examples
- Pattern library in a repository: context file templates, task description examples, review checklists
- Quarterly show and tell open to all engineers
- Direct line to decision makers: a platform or engineering lead attends champion calls regularly and takes feedback into tool and policy decisions
Measuring impact
Measure what matters to the organization, not champion activity:
- Share of teams with a maintained context file
- Adoption of approved tools across teams (and reduction of unapproved ones)
- Quality signals on agent-assisted pull requests, such as rework rate and incidents
- Time to answer for common questions, measured informally through the channel
- Engineer satisfaction with support, through short periodic surveys
Activity numbers such as sessions held can be useful for planning but should not be the headline.
A 90-day plan
| Phase | Weeks | Focus |
|---|---|---|
| Set up | 1 to 2 | Select champions, agree mandate and time with managers, kick-off |
| Deepen | 3 to 6 | Champion training, first pattern library, first team sessions |
| Spread | 7 to 10 | Champions run pairing and workshops in their teams, collect feedback |
| Review | 11 to 13 | Measure, adjust mandate and time, decide on expansion |
Champions beyond engineering
Organizations often discover that the same model works for AI adoption in other functions: finance, operations, customer service. Engineering champions can share how they built their network, while business champions focus on governed AI agents in workflows. For that wider scope, kalAIDoCreate offers an orchestration platform with audit trails, SSO and integrations, plus an expert team that supports enablement. Our article on AI agents for enterprise workflows describes where such agents work today.
How we support champion programs
neooutput runs the champion training track, helps design the mandate and rituals, and offers ongoing office hours where champions can bring hard cases. Many organizations start with an enterprise agentic coding workshop and select champions from its participants.
FAQ
How many champions do we need?
A common starting point is one champion per team of 6 to 12 engineers, plus at least one from platform or security. Adjust after the first review based on demand.
Should champions be volunteers?
Volunteering is a good start, but combine it with selection criteria and manager agreement. The role needs respected engineers with protected time, not just enthusiasm.
How much time should champions spend on the role?
Ten to twenty percent of working time is typical in the first months. Without protected time, the program usually loses momentum quickly.
What happens after the first 90 days?
Review the impact, rotate or add champions where needed and adjust the mandate. Some organizations make the role permanent, others rotate it yearly to spread skills.