The announcement goes out and the licenses arrive. By Tuesday morning, many people still don't know which piece of work to try first. Access has solved one problem, but nobody has shown them a useful starting point or where to go when the result looks wrong.
Champions can provide that nearby help. Events give people a reason to make a first attempt, and the enablement plan brings them back to the task until review becomes routine. Each part needs evidence. After a session, you should be able to point to what someone produced, who checked it, and what happens next.
Pick champions for the work they already show
A champion is a trusted employee who understands the work and can help colleagues change how they do it. Enthusiasm and technical fluency can help, but neither proves that someone will coach a peer. Look for a person who can demonstrate a realistic scenario, sit with a colleague through the first attempt, hear resistance without dismissing it, record what happened, and pass questions beyond their role to the right owner.
Microsoft lists six champion activities. For recruitment, translate each one into something you could have observed before the program began. "Colleagues already ask this person for coaching" gives you evidence. "Seems keen" gives you an impression. Choose candidates who show at least two signals, then check that the group as a whole covers all six activities and the work groups involved.
Riley is a trusted process expert and colleagues already ask for help. Morgan is a tenant admin with no protected time for peer support. Jules is interested but hasn't yet shown coaching, process knowledge, or cross-team influence. Riley has the strongest evidence for the first place. The role still needs limits. Technical, security, licensing, and policy questions go to the people authorized to answer them.
Recognition should reward evidence. A defensible internal badge has four parts: a published criterion, the evidence that demonstrates it, a named reviewer, and where the approved result appears. If a champion submits three barriers but only one has a named owner, they don't earn the "three actionable barriers" badge just because three entries exist. The reviewer applies the published criterion and records what's missing. That single habit is what stops a champion program from becoming a participation-trophy scheme.
Recognition works because it feeds a rhythm, and the engine of that rhythm is a recurring community call: a monthly meeting where champions demonstrate a scenario, surface a barrier, and leave with decisions. A useful call produces more than attendance: a demonstrated scenario, evidence-backed feedback, named owners with due dates, an escalation route for anything outside the champion role, and a follow-up check for next time. A working session produces at least one owner and one date. Anything less is a status update. Publish a short recap in the team's channel afterward so the decisions and the open questions survive past the meeting.
Choose the event by what people should do afterward
Start with the behavior you want to see, then pick the event that gives people a chance to produce it.
For any prompting activity, teach the same four-element model you'll know from The Anatomy of a Great Prompt: Goal, Context, Source, Expectations. Then score submissions on clarity, source discipline, reuse, and verification. Prompt length isn't part of the score. A Prompt-a-thon where the longest prompt wins has measured the wrong thing.
One rule ties every format together: keep facilitation separate from implementation. An Agent Hackathon produces a concept that still needs a separate build to become a production agent. Agent Builder is a no-code route included with a Microsoft 365 Copilot license. A Copilot Studio trial can create and test agents, and publishing needs a different license path. Those are useful facts for a follow-up investigation, best decided outside the room.
Why every event needs a one-week loop
A commitment written on a sticky note is not adoption evidence. Before an event closes, record the artifact, the reuse action, an owner, and a due date. On the due date, classify each: complete (evidence submitted), needs revision (evidence exists but the artifact didn't work), or not completed (nothing came back). For "needs revision," assign a champion to help fix it. For "not completed," ask the owner to name the barrier and set one smaller action with a new date.
This one-week loop turns a good afternoon into an input for improving the program. It also protects you from the most flattering mistake in adoption: reporting the energy in the room as if it were a result.
Sequence the 30-day plan
A single event doesn't build a habit. A sequenced plan does. Use five stages: Prepare, Activate, Practice, Extend, Apply, over roughly a month, anchoring practice to the verified six-week theme progression: managing time, supercharging creativity, expanding expertise, enhancing collaboration, using agents, and applying your skills.
One wrinkle is worth naming. Microsoft's Great Copilot Journey is marketed as a "30-day" journey but is structured as six weeks, and its own materials don't map cleanly to one prompt per weekday. Don't paper over that by inventing a two-prompts-a-day rule. Treat "30-day" as 30 working-day practice slots across six five-day weeks, keep the verified themes for progression, and record whichever official activity your tenant shows, without claiming a one-to-one mapping to numbered prompts that the evidence doesn't support.
Define completion before you launch, because a link click is weak evidence of learning. Collect an artifact packet each week instead: the task attempted, the prompt used, the source or criteria used to review the response, at least one correction or confirmation, and a short reflection. A learner who produces that has demonstrated the review habit, which is the actual skill. Generated text existing on its own proves nothing.
For deeper learning, route people to Microsoft Copilot Academy in Viva Learning: individuals can self-start from it, and an authorized administrator can duplicate, customize, and publish a tailored Academy for a cohort. Don't promise item-level assignment of one exact lesson to one person. That control isn't something to assume across tenants.
