AI Champions over AI Licenses: Why 80 % Adoption Starts with Culture
Citi reached 70 % AI adoption with 4,000 Champions. McKinsey: firms with a clear AI strategy succeed twice as often. Why culture wins.
The Picture from the Controlling Reports
We see a variation of this with almost every client that rolled out AI licenses top-down. Microsoft 365 Copilot is deployed, one license per head, a polished email from the CIO, a training session. Three months later the controller checks the usage stats: 12 % of employees have even signed in. Daily activity sits at 5 % of that group. Annualized against the contract, that's five to six-figure euro amounts flowing to Microsoft each month for a tool nobody opens.
This isn't a regional phenomenon. McKinsey documented in their January 2025 Superagency in the Workplace study that, worldwide, more than 85 % of employees don't even reach the basic level of AI tool usage available to them. The problem is rarely technical. It's cultural.
What Citi Did Differently
If there's one example that works as a benchmark, it's Citi. The US bank has built, over the past two years, a network of around 4,000 "AI Accelerators," distributed across 182,000 employees in 84 countries. Backed by a smaller group of roughly 25-30 "AI Champions" who lead the program in their respective business units.
The Champions aren't IT experts. They're caseworkers and team leads from the operating business who tried AI early and voluntarily, and now bring colleagues along step by step. They try new tools first, integrate them into their workflows, share what works.
The result: over 70 % active usage of sanctioned AI tools. Industry average sits below 20 %.
"Losing the rollout with the better model is no consolation. Winning the rollout with the better people works with any model." — paraphrased from several McKinsey publications in 2025.
The Three Fears That Sabotage Every Rollout
Without a Champions structure, you run into three recurring resistances in every organization. We see them everywhere, from the 50-person agency to the 5,000-person enterprise.
First, the fear of being replaced. According to the Beautiful.ai March 2025 survey, 64 % of managers believe their employees fear losing value because of AI. More than half see real job-loss concerns. This fear is rarely voiced openly, but it blocks adoption more than any technical hurdle.
Second, the fear of embarrassment. "What if I get a stupid result and everyone sees it?" In the first weeks, AI is a stage for many employees on which they could make fools of themselves. Without psychological safety, nobody experiments.
Third, the fear of wrong answers. Employees who've once experienced a fabricated AI response lose trust quickly. A Deloitte survey from 2025 shows trust in company-provided AI fell 31 % between May and July of that year. This is precisely why the data foundation behind the AI is so decisive. Clean sources reduce fabricated answers and build trust faster than any training session.
Champions address all three fears not through arguments but through visible practice. When the colleague at the next desk uses AI and her work gets better rather than vanishing, arguments dissolve on their own.
What "AI Culture" Actually Looks Like in Practice
AI culture is neither a PowerPoint slide nor a values poster in the intranet. In practice it consists of four concrete things.
Permission to experiment. Employees need to understand they don't have to get it right on the first try.
Clarity about limits. What is AI allowed to do, and where does human responsibility remain. Without that clarity, you end up with the compliance risks from Shadow AI we cover in detail elsewhere.
Honest talk about bullshit. AI sometimes makes things up. Saying that openly and shaping training around it builds trust. Hiding it destroys it.
Recognition for good examples. Champions who build something useful need to be visible. In the intranet, in team meetings, in quarterly reviews.
PwC's 2025 Workforce study shows the inverse in numbers: daily AI users report productivity benefits at 92 %, versus 58 % for infrequent users. Daily users also feel more secure in their jobs. The biggest fear doesn't come from too much AI. It comes from too little experience with it.
A 90-Day Champions Program
In our client work the following setup has proven robust.
First, identify champions. One per 30 to 50 employees, not from IT, but from the operating business. Volunteer, not appointed.
Second, make them visible. Champions get a written mandate from leadership, with allocated time.
Third, weekly exchange. 30 minutes, blocked in the calendar. What worked, what didn't?
Fourth, quarterly reporting. Use cases, wins, blockers reported directly to leadership. Champions get a stage.
Fifth, recognition. Symbolic, financial, or through visible career opportunities. Champions aren't a hobby. They're a leadership achievement.
A well-run Champions program also feeds directly into the AI literacy obligation of the EU AI Act, which has been in force for every European company since February 2025.
Get Started
If you bought AI licenses and the usage statistics disappoint, the fault, in our experience, is almost never the technology. We help companies set up the right Champions program, within 90 days, without the big-consulting overhead. If you're at that point, let us know.
Sources
- McKinsey, "Superagency in the Workplace: Empowering People to Unlock AI's Full Potential", January 2025
- PwC, "Global Workforce Hopes and Fears Survey 2025"
- Harvard Business Review, "Workers Don't Trust AI. Here's How Companies Can Change That", November 2025
- Beautiful.ai, "AI's Impact on the Workplace in 2025: 2nd Annual Survey of American Managers"
- AI News, "The quiet work behind Citi's 4,000-person internal AI rollout", 2025
Frequently Asked Questions
Who makes a good AI Champion?
Champions are not IT specialists. They are operational employees who already experiment with AI tools on their own initiative, are well-connected internally, and enjoy showing colleagues something new. Hierarchically, they tend to sit at mid-level, caseworkers or team leads, not in senior leadership. One Champion per 30 to 50 employees is a proven ratio.
How do I measure the success of a Champions program?
Three metrics are enough. Active weekly AI tool use per department, measured by actual outputs and not just logins. Number of use cases identified by Champions and promoted into the broader rollout. Sentiment in short employee surveys, fear and uncertainty should drop measurably within six months.
What does an AI Champions program cost?
Far less than the licenses that sit unused without one. The main investment is Champion time, typically four to six hours per week, plus a lightweight structure with monthly meet-ups, recognition, and a clear leadership sponsor. In money terms: well below the cost of one unused Microsoft 365 Copilot license per year.
