The Copilot Champion 90-day onboarding checklist

You hired a Copilot Champion. (If you didn’t, start with the job-description postfirst — this one assumes the role exists and someone’s in seat on Monday.) Now what do you actually have them do for ninety days so the license spend turns into measured behaviour change, not a Slack channel and a wiki page?

Here is the week-by-week onboarding plan we hand every client’s newly-hired Champion. It is opinionated, it is boring in the right places, and it has produced 60%+ weekly-active-user adoption at every 30–300-seat firm that has run it end-to-end.

The Champion’s first job is not to teach Copilot. It is to find out what 8 people across 4 departments actually do every Tuesday morning — and to make one of those tasks measurably faster by Friday.

The 90-day shape (one paragraph)

Days 1–14 are listening + baseline. Days 15–45 are the first measurable win in two departments. Days 46–75 are scaling the pattern that worked to four more departments and building the governance scaffolding. Days 76–90 are the steady-state operating cadence + the first quarterly review the executive sponsor will see. If you can’t name the win from days 15–45 by day 30, the role is in trouble and the sponsor needs to know.

Weeks 1–2 — listen and baseline

No training sessions yet. No demos. The Champion’s job these two weeks is to leave the office with a written document the sponsor can read in 20 minutes.

Checklist for weeks 1–2

• Meet the sponsor on day 1. Agree the single success metric for day 90 (we recommend Microsoft 365 Copilot weekly-active-user rate, defined: any meaningful Copilot interaction across Word / Excel / Outlook / Teams / Chat in the last 7 rolling days, measured from the Copilot Dashboard in the M365 admin center).
• Pull the baseline.Copilot Dashboard + Viva Insights + the seat-assignment report. Capture starting WAU, starting per-seat-per-week interaction count, and the top 5 surfaces used (almost always Word, Outlook, Teams meetings, Chat, Excel in that order; if your firm is different, that’s data).
• Run 8 listening interviews.2 senior partners / executives, 2 mid-level managers, 2 individual contributors who already use Copilot daily, 2 individual contributors who don’t. 30 minutes each. The one question that matters: “Walk me through your Tuesday morning, hour by hour, between 8am and noon.”
• Write up the ‘Tuesday morning’ doc. A 2-page summary of what people actually do, where Copilot would help, and the 3 candidate pilot tasks the Champion will tackle first. The sponsor signs off on this.
• Audit the seat assignments. The pricing/TCO postcovers the seat-reclaim discipline; week 2 is when the Champion does the first pass. Expect 8–15% of assigned seats to be near-zero users.
• Verify the governance pack exists. If there is no AI Acceptable Use Policy, no oversharing assessment, no labelling baseline, the Champion flags it now and the governance pack work runs in parallel to the pilot. The Champion does not own the pack but cannot ship without it.

Weeks 3–6 — the first measurable win

Two departments. Two tasks. Two before-and-after measurements. Nothing else matters this month.

Checklist for weeks 3–6

• Pick two departments where the candidate task is repetitive and time-boxed. Finance month-end variance commentary, sales weekly pipeline rollups, ops incident post-mortems, HR offer-letter drafting, marketing campaign-recap decks. Not“the legal team should use it more” — too vague to measure.
• Run the before-measurement.5 people per department, current state, stopwatched. “How long does this task take you today, without Copilot, with your usual tools?” Record the median.
• Ship the prompt pattern, not the training.A 1-page card per task: the prompt template, where it lives (Word / Excel / Chat), the 3 things to check before sending, the screenshot of a good output. The card goes in a SharePoint library the Champion owns.
• Pair-work the first three runs per person.The Champion sits with each pilot user for the first 3 times they do the task with Copilot. Not to teach Copilot — to debug the prompt card in real time. Cards usually need 2–3 revisions before they’re actually usable.
• Run the after-measurement at week 6.Same 5 people, same task, with the prompt card and 3 weeks of practice. Record the new median. The honest delta is usually 30–55% on the tasks that work, and the Champion will have killed at least one task that didn’t.
• Send the week-6 update to the sponsor.Two charts. Before/after time. Pilot-user WAU vs firm-wide WAU. One paragraph on what worked and what got killed. No slides. Email.

Weeks 7–10 — scale the pattern that worked

Four more departments, the same pattern. In parallel, the Champion is now standing up the operating scaffolding the role needs to be sustainable past day 90.

Scaling checklist

• Replicate the 1-page card pattern in four more departments. Same playbook: pick the candidate task, baseline, ship the card, pair-work first three runs, measure at week 4. By day 75 you should have 6 prompt cards in the library and 6 measured deltas.
• Recruit 2–3 deputy champions.One per major department. They don’t need the title or the budget; they need the time (~2 hours/week) and the willingness to be the local prompt-card author. The Champion’s job is now also their job.
• Set the weekly office hour.Same time, same Teams channel, every week, recorded. Attendance will be 4–12 people; that’s fine. The recording is the asset.
• Build the prompt library’s discoverability.A pinned Teams tab, a SharePoint link in the firm’s daily comms, and a Copilot Studio “ask the prompt library” agent for the people who won’t click through a SharePoint site. (This is also the Champion’s first agent — small, internal, no governance escalation needed.)

Governance scaffolding (parallel track)

• Run the oversharing assessmentat week 8 if it hasn’t been done. SharePoint Premium has the built-in scan; even without Premium, Purview Content Explorer + a sampled audit covers most of the risk. The Champion doesn’t fix oversharing — that’s the information-protection owner’s job — but they surface the report.
• Confirm the AI use register has every agent, prompt card, and Copilot Studio surface the Champion has shipped. One row per artefact. Owner, purpose, data accessed, review date.
• Schedule the quarterly Copilot reviewwith the executive sponsor. Calendar invite, 45 minutes, recurring. The Champion will present 4 of these per year — the cadence starts now.

Weeks 11–13 — steady state + the first quarterly review

The Champion is no longer in setup. They’re running an operation. The last three weeks are about proving the operation is repeatable without heroics.

Operating-cadence checklist

• Publish the operating doc.One page: weekly office hour time, deputy champion list, prompt library URL, intake form for “here’s a task that might be a good Copilot fit”, governance review schedule. Anyone in the firm should be able to find this in two clicks.
• Run the seat-reclaim second pass.Compare day-90 WAU per assigned seat to day-1. Reclaim any seat that has fewer than 4 meaningful interactions in the last 30 days after the Champion has attempted at least one outreach. Be honest. The reclaim discipline is what makes the licence math defensible to the CFO.
• Write the day-90 retrospective. What worked, what got killed, what the next quarter looks like. 2 pages, plain English. The executive sponsor reads this before the quarterly review.
• Run the first quarterly review.45 minutes, sponsor + 1–2 dept heads who hosted pilots + the Champion. Three slides: where WAU started vs is now; the 6 cards and their measured deltas; what the next quarter’s 4 candidate tasks are. Decide out-loud whether to expand the role’s budget / deputies / scope.

What ‘good’ looks like at day 90

Not every firm hits all of these. Most hit most of them. If you hit four out of six, the role is working.

1. WAU has moved from baseline to at least 45% firm-wide (60%+ in the pilot departments).
2. Six 1-page prompt cards shipped, each with a measured before/after delta of 25%+ on the target task.
3. Two to three deputy championsactively maintaining cards in their departments.
4. One internal Copilot Studio agentshipped (the “ask the prompt library” agent), used by at least 20 people in the last 30 days.
5. Seat-reclaim done twice with a documented audit trail; CFO can defend the per-seat spend.
6. Quarterly review on the calendar,recurring, with the executive sponsor.

What ‘bad’ looks like — and how to spot it

The single most common failure mode for the Champion role at day 90 is a Champion who has run 12 lunch-and-learns, built a beautiful wiki, and has no measured task-level delta to point to. If at day 30 you cannot name the two pilot tasks, at day 45 you have no before-numbers, and at day 60 you have no after-numbers — the role isn’t working. Have the conversation with the sponsor at day 30, not day 90. The fix is almost always “pick smaller tasks and measure them sooner”, not “replace the Champion.”

What this pairs with

The Champion JD template is the prequel. The first 90 days — 7 mistakes piece is the failure catalogue this checklist is designed to dodge. The 4-week activation playbook is the org-wide ship the Champion is operationalising afterwards.