Designing an AI Champions Program: 6 Weeks, 5 People, Repeatable Wins

If you're a CEO who's already paid for two AI workshops and noticed nothing actually changed in the team's daily work — the gap isn't more training. It's that you have no champions. A champions program is what turns a workshop into a habit, and it has a surprisingly narrow design space.

What is an AI champions program — and what is it not?

A champions program is a structured six-week cohort where 5-8 employees become the internal pattern-makers for AI usage at your company. Each champion ends the program with at least one shipped automation in production and a peer group inside their team that uses it.

It is not:

• A training certificate factory ("we trained 200 people!")
• A separate AI team siloed from the rest of the business
• A side project that runs without ownership of a real workflow

Definition: AI Champion — an employee embedded in a specific team who takes ownership of building, deploying, and teaching AI workflows for that team. They keep their day job; the champion role is roughly 25-30% of their time during the program and 10-15% after.

The reason this works where one-off training fails: champions own a real workflow they're trying to improve, not an abstract competency. The BCG 5-hour training threshold is real — programs under ~5 hours produce no behavior change — and a champions program is structured around closer to 40-60 hours of applied practice spread over six weeks.

How do you pick the right 5-8 people?

The wrong selection model: "send the team's senior engineer." Senior technical depth correlates with neither AI adoption nor peer influence inside a non-technical team.

The right model uses three criteria, in this order:

1. Credibility inside their team — when this person says "this works," people try it. Higher signal than years of experience.
2. Curiosity bias — has historically explored new tools without being asked. Look for the person who already showed colleagues a ChatGPT trick last quarter.
3. Workflow ownership — does a job that has a clear, repeatable, AI-amenable workflow. Not "thinks about AI strategy" — "writes 12 client proposals a month."

Definition: Workflow ownership — the candidate is hands-on accountable for an operational process that runs regularly, has measurable inputs and outputs, and is currently more painful than it should be. Without it, the champion has nothing to apply training to.

What you're explicitly NOT optimizing for: technical seniority, official rank, or volunteer enthusiasm with no underlying workflow.

What's the weekly format that actually works?

Six weeks, six rhythms. Each week has the same shape so the cohort can predict their load.

• Monday (90 min): Cohort lab. All champions together, plus a coach. New concept, live demo, Q&A.
• Tuesday-Wednesday: Applied work. Each champion works on their own automation. Roughly 6-8 hours per week.
• Thursday (45 min): Shoulder-to-shoulder hot seat. One champion in the spotlight; the coach and peers debug their automation live. Rotates each week.
• Friday (30 min): Cohort sync. What shipped this week, what's blocked, who needs help.

Definition: Shoulder-to-Shoulder hot seat — a coaching format where one champion presents their in-progress automation and gets debugged in real time by a coach and peers. The point is not the fix; it's that everyone watches how the fix is reasoned through.

The 6-week duration is not arbitrary. A 3-day intensive crashes hard against retention; a 12-week program loses energy by week 8. Six weeks is long enough to ship something real and short enough that the executive sponsor is still paying attention.

What does each champion deliver?

Concrete artifacts, not abstract competencies. By end of week 6, each champion has:

• One shipped automation in their actual workflow, with a human review gate.
• A written rubric or prompt that another employee can read and reuse.
• A 15-minute internal teach-back delivered to their team, recorded.
• A baseline + week-6 measurement of the workflow they touched (time saved, throughput, quality).

The teach-back is the most under-rated artifact. It forces the champion to compress what they learned into something teachable, and it surfaces gaps the cohort lab didn't catch. The recording becomes onboarding material for future hires in that role.

Copy/paste selection + program template

For the executive sponsor running the rollout:

AI CHAMPIONS PROGRAM — [COMPANY], [QUARTER]

Cohort size: ___ champions (target: 1 per 15-20 staff in scope)

Selection (one row per candidate):
| Name | Team | Credibility (1-5) | Curiosity (1-5) | Workflow owned |
|------|------|-------------------|-----------------|----------------|
|      |      |                   |                 |                |

Time commitment per champion: ~25-30% during program, ~10-15% after.
Manager sign-off REQUIRED before invite (champion needs cover for displaced work).

Cadence:
- Week 0: kickoff, baseline measurement of each workflow
- Weeks 1-6: Mon lab / Tue-Wed work / Thu hot seat / Fri sync
- Week 6: teach-back to own team, recorded
- Weeks 7-10: post-program — measurement of transfer (other people using the automation)

Sponsor responsibilities:
- Attend Week 0 + Week 6 in person
- Block champion calendars (this fails without it)
- Review week-6 measurements personally
- Fund week-10 transfer review

Champion responsibilities:
- Ship one production automation by week 6
- Record 15-min teach-back
- Mentor at least 2 peers in weeks 7-10

That sheet, plus a name in each row, is 80% of program success.

Tool tip (Course for Business): The 6-week program structure we run uses the AI Champions (1:15-20) ratio as a hard constraint — too few champions and the rest of the team can't get help; too many and the depth dilutes. Each champion goes through the Shoulder-to-Shoulder hot seat at least twice during the program, which is where the meaningful skill transfer actually happens. The Augment-don't-replace framing is what keeps champions trusted by their peers: they're not building automations to delete their teammates' jobs, they're building tools their teammates ask for. See how the program maps to your team at https://course.aiadvisoryboard.me/business.

How do you measure that it actually worked?

Four numbers, measured at week 6 and again at week 10:

1. Shipped automations — count of champion-built tools in real use (not in a sandbox).
2. Transfer rate — % of the champion's team that uses the automation weekly by week 10.
3. Saved time per workflow — measured against the week-0 baseline.
4. Champion retention — are the champions still in the role and still championing at week 12?

The transfer rate is the unforgiving one. If a champion built something nobody else uses, the program failed for that champion — and the fix is usually that the teach-back was skipped or the workflow choice was too narrow. Aim for ≥60% transfer by week 10 in the affected team.

Team scan (what AI champions report after week 1)

Patterns we see consistently in the first week of a cohort across SMB rollouts:

• ~80% of champions complete the week-1 lab with at least a working prototype
• First friction: champions underestimate how much rubric-design time their workflow needs
• First win: a champion ships a 10-line draft assistant for their highest-volume task
• Cohort coherence emerges by Thursday hot seat (champion #3 watches champion #1 and applies the same pattern)
• Manager surprise: champions deliver ~30% more peer questions to the lab than expected
• First risk: 1-2 champions try to over-engineer (build the "everything agent") — coach redirects to narrow scope
• Use case ranked #1 in week-1 retro: a workflow-specific drafting tool, not a chatbot
• Saved-time estimate: champion's own workflow drops by 20-40% by end of week 1, climbing through week 6
• Champion morale: highest in cohorts where the executive sponsor showed up live to the Monday lab
• Sustained adoption signal: champions who finished teach-back in week 6 are 3x more likely to still be championing at week 12

Micro-case (what changes after 7-14 days)

A 140-person services SMB ran the program with 7 champions, one per major team. By day 14: every champion had at least one prototype in their workflow, two had shipped production-grade automations behind a human review gate, and the executive sponsor (the COO) had already started getting unsolicited Slack messages from non-champions asking when they could join the next cohort. The biggest surprise was the teach-back artifact — a 15-minute recording from the proposal-team champion ended up being shown to new hires in onboarding by week 12. Six months later, three of the seven champions had moved into a part-time "AI lead" role for their function, the other four had stayed in their original roles but continued to ship one new automation per quarter, and the company had a pipeline of internal candidates for the next cohort.

Note on this case: This example is illustrative — based on typical patterns we observe with companies of 30-500 employees, not a single named client. Specific numbers are rounded approximations of common ranges, not guarantees.

Tool tip (Course for Business): The most common reason champions programs fail is not training quality — it's that managers didn't block calendars and champions tried to do champion work in their margin time. The 6-week program enforces this explicitly: managers sign a calendar commitment in week 0, the coach surfaces calendar conflicts in the Friday sync, and the Shoulder-to-Shoulder hot seat format makes any champion who's falling behind visible to the cohort early. Augment-don't-replace is the principle that keeps champions trusted; the AI Champions (1:15-20) ratio is the constraint that keeps the program economically rational. Book a 30-min mapping call at https://course.aiadvisoryboard.me/business.

FAQ

Can we run this with 3 champions instead of 5? You can run it, but you lose cohort effect. Three champions means each one is on the hot seat every other week, the peer debugging gets thin, and one drop-out kills momentum. Five-to-eight is the band where cohort dynamics produce more value than the coaching alone.

What if our company doesn't have 75-150 people (the 1:15-20 ratio band)? For 30-50 person companies, run a single cohort of 3-4 champions and accept the smaller cohort effect. For 200+ person companies, run two parallel cohorts with separate coaches and a shared monthly check-in — running one cohort of 12 dilutes hot-seat quality.

Do champions need to be full-time on this? No, and they shouldn't be. The program is calibrated to ~25-30% of a champion's time during the six weeks and ~10-15% after. Full-time champions lose touch with the daily workflow they're supposed to be improving — the part-time setup is a feature, not a constraint.

What's the post-week-6 cadence? Monthly champion sync (60 min), quarterly cohort retro (90 min), and one shipped automation per champion per quarter as a soft target. The post-program phase is where transfer rate and retention either compound or collapse — most programs underinvest here.

Conclusion

A champions program is the bridge between training and habit. Without it, you have certificates. With it, you have shipped automations, internal teachers, and a repeatable cohort engine that the next 5-8 champions can step into.

Pick a quarter. Pick 5-8 candidates against credibility, curiosity, and workflow ownership. Block their calendars. Run the six-week cadence. Measure transfer at week 10, not week 6.

If you want every employee to ship their first AI automation in five days — book a 30-min call and we'll map your team's first week at https://course.aiadvisoryboard.me/business.