// Part 1 of 2 A guide for owners and leaders of companies with 50 to 300 people. Part 2: what comes after AI training.
Where your people are now, and what closes the gap
You're not early, and your people are already experimenting. In an August 2025 Angus Reid survey for Growclass, 42% of Canadians said they use AI in their jobs, but only 12% of workers had received formal training. 68% said they'd be interested if their employer offered it (Angus Reid Group for Growclass, September 2025).
That gap is where problems start. People use personal ChatGPT accounts with no rules. One person finds a great way to do something, shows two colleagues, and it goes no further. Leadership has no idea what's working.
A course doesn't close that gap, because a course teaches the tool rather than the job. What closes it is someone on the team who already knows the job, is one step ahead on AI, and is given time to help everyone else catch up. That person is an AI champion.
The approach works in large companies. At Docusign, a six-person champions team, with no mandate, lifted engineering's AI tool adoption from 49% to 65%. A follow-on program of embedded peer champions then took active use to 93% (Docusign's own figures, 2025).
It works in smaller firms too. Part 2 shares results from two of our clients, a fractional HR firm and a home builder.
Rollout order: who in your company should use AI first?
Order matters more than most companies expect. We roll out in three steps.
- Leaders first. Owners, executives and the people others look to for how work gets done. When leaders use AI visibly, everyone else follows. In Gallup's 2026 research in organizations adopting AI, 78% of employees whose manager strongly supports AI use it frequently, compared with 44% of everyone else. Only 36% strongly agree they have that support (Gallup, July 2026). Training staff while leaders stay on the sidelines produces competing opinions and a stalled rollout (see Where AI rollouts go to die).
- Champions second. Once leaders are proficient, pick your champions. They turn leadership's direction into everyday practice inside each team.
- Everyone else third. Champions run short, practical sessions using the team's own tasks and stay available afterwards. This is where the whole company starts to move.
You can do all three steps in about 90 days. In our experience, skipping the first step is the most common reason training budgets get spent without much changing.
What is an AI champions program?
An AI champions program gives a few respected employees, about one per 15 to 20 staff, two to three protected hours a week to help their team use AI well.
An AI champion is a peer. They do the same job as their team, just one step ahead on AI, and they don't manage the people they help. Nufar Gaspar, who built Intel's internal AI champions community, calls this "peer plus one" (The AI Daily Brief, October 2026).
A good champion:
- shows how AI handles the team's everyday tasks, using the team's own examples;
- helps colleagues get unstuck without having authority over them;
- spots work that's worth improving with AI and brings it to leadership;
- tells leaders what the team needs (tool access, permissions, rules); and
- shares what works with the other champions.
A champion is not your IT department, your AI policy owner or a full-time AI lead. In a company of 50 to 300 people, it's a part-time role held by someone who keeps their day job.
Champions also differ in how much they build. Some mainly spot and share good uses. Others build simple things for their team, such as shared instructions, templates and small automations. Decide up front which kind you want. Write down what champions may build on their own and what needs approval, and agree it with your IT provider.
How to choose your AI champions
Look for three signs before anything else.
- // Sign 1
They already use AI without being asked
They're already exploring, so the role builds on a habit they have.
- // Sign 2
Colleagues already come to them with questions
The informal role exists. If their manager frees up time for it, you're making it official.
- // Sign 3
Their manager will give them real time
This is where nominations often fail.
Then check for these: they know the business well, they can persuade people who don't report to them, and they take initiative.
Avoid these common mistakes:
- // Mistake
Picking the busiest person, with no time freed up
They're often the best user and have no slack. The go-to person from sign 2 is a good pick only when their manager takes work off their plate.
- // Mistake
Making it a reward
A strong performer isn't automatically a strong AI user or a patient teacher.
- // Mistake
Picking someone new
Champions need to know which problems are worth solving, and new hires don't know yet.
- // Mistake
Senior people appointing themselves
Leaders should set direction and remove blockers, and leave the team sessions to the champions.
Treat the role as a development step. Champions build skills the company will need more of, so plan where the role leads next, such as owning shared workflows or leading a team's AI work.
How many AI champions you need, and how much time to give them
These are the starting rules we use with clients. They come from our own practice, and published research doesn't set these numbers, so adjust them as you learn.
| Rule | Our starting point |
|---|---|
| How many champions | One for every 15 to 20 employees, with at least one in every team whose work is different. |
| Protected time | Two to three hours a week per champion, and closer to half a day a week for the first month while they set up. |
| Who runs it | One owner for the whole program, usually a leader who uses AI well and can clear access and approvals. |
| Company size (employees) | AI champions to start with |
|---|---|
| 50 | 3 |
| 100 | 5 to 7 |
| 150 | 8 to 10 |
| 250 | 13 to 17 |
| 300 | 15 to 20 |
Protected time is the item most likely to quietly disappear. Put it in the champion's calendar and their manager's goals. In billable firms, two to three hours is about 5 to 8% of a champion's week. Book it to an internal project code so everyone can see it.
Published examples range from GitHub's suggested minimum of 30 to 60 minutes a week (GitHub, August 2025) to Safe Software in Surrey, BC, which gives every employee 10% of their week to explore AI (The Globe and Mail, data as of May 2025). For most growing companies, two to three hours is enough to make progress without hurting the champion's day job.
Your first 90 days
- // Week 1
Set up
- Announce the champions and explain the role to the whole company in one short message from leadership.
- Brief each champion's manager in 20 minutes on what the champion will do, how much time it takes and how the team should use them.
- Confirm tool access. Everyone should be on the same approved tool (Claude, ChatGPT, Copilot or Gemini, whichever your company has chosen) with a business account.
- Share your AI use policy. A single page is enough to start: use business accounts, name what client information may never go into an AI tool, and say who reviews AI output before it reaches a client. In law, accounting and other firms that hold confidential client files, write this page first. Our guide to AI policy for Canadian businesses covers what to put on it.
- Pick your timing. Don't launch in your busiest season.
- // Weeks 2 and 3
One win each
- Each champion picks one task their team does every week that takes too long.
- They solve it with AI and write down how, so others can repeat it.
- // Week 4
Show it
- Champions show their win to their team in a 30-minute session using the team's own examples.
- Champions meet as a group to swap what worked.
- // Days 31 to 90
Spread it
- Each team adds a second and third win, chosen with their leader.
- Champions run short team sessions and keep office hours for questions.
- The monthly champion meetup starts, with leaders attending.
- At day 90, measure the same tasks you measured on day one and decide what to scale next.
Our readiness test is simple. A champion is ready when they've taken one task their team does every week, improved it with AI, and their teammates are using it. If that hasn't happened yet, they need more support.
How to support champions after launch
Most champion programs start with energy and fade within a few months. Gaspar points to what happens after launch as the biggest predictor of success. Plan for it up front.
- A named owner. Someone keeps the program running, sets priorities and clears blockers such as access, approvals and tool requests. Programs without an owner fade.
- A regular rhythm. Hold a monthly champion meetup and keep a shared channel for questions and wins.
- Visible recognition. Credit wins in company meetings. Recognition usually motivates more than a bonus, and it tells everyone else what good looks like.
- Manager buy-in. Re-confirm the protected time with each champion's manager every quarter.
- An easy exit and refresh. Some champions will find the role isn't for them. Thank them, let them hand over, and bring in someone new.
What AI champions are responsible for
Using AI well personally is the easier part. These are the skills that turn a strong user into a good champion.
- Picking work worth improving. Who struggles with this task, how often, and what it costs. Check that someone hasn't already solved it.
- Starting simple. Often a saved set of instructions and example files is enough. Move to automations (tasks that run on their own) only when the simpler option clearly isn't enough.
- Setting guardrails. Decide what the AI must never do, what it must never include, and who checks the output before it goes out.
- Judging quality. Know when the output can be trusted and when it can't, and write down the checks. This is the skill that most separates champions from other users.
- Getting it used. Champions make sure it lives where the team works and that someone looks after it after launch week.
Much of this can be set up once and reused. For example, a champion can save the quality checks as standing instructions so the AI flags likely problems on every draft. A person still checks against the list before it goes out.
How to know it's working
Logins and number of prompts tell you very little. Look at these instead.
- // Measure
Before-and-after numbers on real tasks
Measure how long a proposal, a weekly report or a client update takes today, and measure again in 90 days. Without a "before" number, you can't prove anything. Include revenue work: proposals sent per week, turnaround time and win rate.
- // Measure
Quality as well as speed
Faster work that's worse isn't a win. Sometimes AI work takes longer and is much better.
- // Measure
Spread
How many people use each champion's shared setups, and in how many teams.
- // Measure
What people do with the time saved
BCG's 2026 research found 66% of regular frontline AI users get limited or no guidance on this (BCG, June 2026). Leaders have to decide it. Our article on who decides where the time goes walks through the options.
Why AI champion programs fail
- No owner after launch.
- Protected time is quietly taken back.
- Champions were chosen for seniority or as a reward rather than for fit.
- Leaders never became proficient, so champions have no backing.
- Champions experiment on side projects instead of tasks the team does every week.
- Wins stay inside one team and leadership never hears about them.
- The champion group is never refreshed.
One honest caveat: champions aren't the only way to get a company using AI. Some companies get there with a top-down mandate. But in growing companies without an AI department, champions are the most reliable and least expensive approach we've found to turn individual skill into company-wide habit.
Frequently asked questions
Do we need an AI policy before we start?
You need a short one, and a page is enough. Cover which tools are approved, what information must never go into them, and who to ask. Refine it as champions find gaps.
Which tool should we train staff on: Claude, ChatGPT, Copilot or Gemini?
Pick one approved business tool for everyone, based on the systems your company runs, such as Microsoft 365 (where Copilot is built in) or Google Workspace (where Gemini is built in), and what your IT provider supports. A decent tool that gets used every day beats a perfect tool nobody picks up.
What about employees who resist AI?
Don't start with them. Start with the willing majority. When a respected colleague shows how AI took two hours off a task the resisters also dislike, many come around.
Will staff worry AI is replacing them?
Some will. Say plainly what the time saved is for, such as more client work, faster turnaround or fewer late nights. Leaders need to decide that before launch. What AI means for your staff covers how to start that conversation.
How much does an AI champions program cost?
Mostly champion time, plus business licences for everyone on your chosen tool. In hours: a 120-person company with 6 to 8 champions at two to three hours a week each spends about 12 to 24 hours a week of internal time. In a billable firm, that time is usually the biggest cost, so price it at your internal rate. Outside coaching is optional.
Should champions be paid more?
Usually not at first. Protected time, visible recognition and a clear development path matter more. Revisit pay if the role grows into building tools for other teams.
How long until we see results?
Expect individual wins within the first month and measurable team-level change within about 90 days. If it's taking longer, something on the why programs fail list is usually happening.
What happens after the first 90 days?
Part 2 of this guide describes three stages of adoption: using AI, trusting AI work with owners and written checks, and redesigning how work gets done. This rollout gets you through the first stage.
Where Sidekick fits
Sidekick runs this sequence with growing companies through Fractional AI Leadership. We coach your leaders to use Claude, ChatGPT, Copilot or Gemini on their weekly reports, proposals and planning first. Then we help choose and train champions and run team sessions, measured against before-and-after numbers.
Book a free 30-minute call and we'll help you pick your first champions and the first task each one should fix. Or start with the free AI Opportunity Map, about 10 minutes.
Sources checked October 16, 2026: Angus Reid Group for Growclass, AI training in Canada; Docusign, how we drove AI adoption across engineering; Gallup, AI in the workplace; BCG, AI is reshaping jobs faster than companies are reshaping work; GitHub, Activating internal AI champions; The Globe and Mail on Safe Software; The AI Daily Brief, "How to Go from AI Power User to AI Leader", with Nufar Gaspar.
