A week into a company-wide rollout and the pattern is already visible. Two teams are using the tools daily. One team sat through the training and went quiet. A third is waiting to see how serious leadership is before committing any effort. And the delegate, responsible for AI adoption across the whole business, is fielding questions from three departments, chasing three others, and writing a board update from the same desk.
The enterprise answer to this is a centre of excellence, a dedicated team with formal governance and programme ownership lifted out of the day-to-day functions. It makes sense at 5,000 people. In an owner-managed business of 50 to 200, it arrives before the scale justifies it and creates a function whose existence can slow the adoption it is supposed to drive.
The lighter structure that works at this scale is a champions network.
What is an AI champions network?
An AI champions network is a small group of people, one from each function, who use the tools well enough to support their colleagues. They hold no formal authority over AI direction. They handle local questions, run informal coaching inside their own team, report what is working and what is failing, and give shadow AI use a route to surface safely rather than stay hidden.
Each champion stays inside their own function. The network is the connection between them, a light monthly forum, a shared channel, something that requires an hour a month rather than a calendar of meetings. The purpose is to keep adoption moving inside functions without the delegate needing to be the single point of contact for every team’s questions, stumbles, and workarounds.
The contrast with a centre of excellence is worth stating plainly. A CoE creates a dedicated team outside the functions, with formal governance and programme management. It concentrates expertise but separates it from the places where adoption actually has to happen. A champions network keeps the expertise inside the teams, while giving the delegate a structured line into what is really going on across the business.
Why does the delegate need one?
Adoption driven by a single person reaches the keen teams and stalls at the reluctant ones. Cross-functional ownership is one of the clearest differences between rollouts that build momentum and those that plateau after the first few pilots. When every function has a credible internal voice who uses the tools, the rollout reaches places the delegate cannot get to from one desk.
Korn Ferry’s research on AI readiness notes that businesses frequently assign AI leadership to capable operators who lack AI-specific experience. The gap between expectations and preparation is real, and it widens with each function the rollout is supposed to reach. A champions network narrows that gap from the inside out. The champion does not need deep technical knowledge. They need enough working familiarity to be useful, and enough credibility in their own function that colleagues will listen.
The shadow AI problem adds to the case. Research tracking enterprise AI use finds that over 90% of employees are already using personal AI tools for work, whether or not the business has sanctioned those tools. That footprint exists whether the delegate knows about it or not. A champion who is trusted inside their team gives people somewhere to take their questions rather than keeping the tools hidden.
BCG’s research on the gap between AI usage and business impact points to cross-functional ownership and genuine change management as the variables that separate businesses seeing results from those seeing activity with no measurable outcome.
Where does a champions network do its work?
Champions do much of their work inside their own function, and most of it is informal. A colleague wants to know whether a tool can handle a particular document. Someone is unsure about putting AI-generated text into a client communication. A team member has found a faster approach and wants to know if it is sound. The champion handles those questions before they reach the delegate.
This matters because the questions are real and come in volume during a live rollout. If the only route is to the delegate, every small question becomes a bottleneck and the pace of adoption slows. If there is no route at all, the questions go unanswered and workarounds accumulate out of sight.
Across the network, patterns emerge that the delegate would not see from above. One function is running into a data quality problem that another has already solved. A tool that works well in operations is being used incorrectly in finance. The network is the mechanism for sharing those observations before they become embedded problems.
The feedback loop back to the delegate closes the picture. Champions give an honest read on what is actually happening, not what is being reported up the line. That is the intelligence the rollout depends on.
When does a champions network make sense?
A champions network adds value once the rollout covers more than two or three functions and the delegate is stretched. Before that point, direct ownership is fine and a network adds coordination cost that is not yet justified. The trigger is the moment where one person cannot hold the context for every team’s questions, workarounds, and progress.
Scale matters too. In a business under 30 people, informal channels cover much of what a network would do. The structure becomes worthwhile when functions operate with enough independence that adoption genuinely differs between them, which tends to happen somewhere around 50 to 100 people.
The selection mistake to avoid is appointing champions based on enthusiasm for AI rather than credibility inside their own team. An enthusiast with low team credibility creates a role that looks right on paper but does very little in practice. The champion’s job is to influence colleagues, and that influence runs on trust built through their existing role, not on how much they like the tools.
Spencer Stuart’s work on AI leadership argues that the power-user approach, where senior people get hands-on with the tools themselves, is more effective than delegated oversight. The same logic applies to champion selection. Familiarity with the tools matters, and it has to be genuine working familiarity rather than a title.
What else needs to be in place for it to work?
The champions network is a structure, and structures need three things to function. Clear expectations of what champions are responsible for. A regular rhythm that keeps the network active. A visible connection back to the person holding the AI mandate. Without those three, champions default to their day jobs and the network exists in name only.
Clear expectations means a short description of the role, one page at most, stating what the champion handles and what is not their problem. They field local questions and surface feedback. They do not own the AI strategy or approve tool usage across the function.
The regular rhythm is usually a monthly forum of 45 to 60 minutes, cross-functional, with the delegate in the chair and champions sharing what they have seen. HR Executive’s analysis of AI adoption identifies recognition and community-of-practice structures as among the most effective levers for building the buy-in that makes adoption last. The forum serves both purposes at once.
The guardrails complete the picture. Champions need to know what they can tell their teams to do, what requires escalation, and what current policy says about tool usage and data. Without that clarity, they either give wrong answers or stop answering altogether. Neither outcome serves the rollout.
If the mandate is yours and you are carrying the adoption weight alone, a champions network is the structural move that changes that. Book a conversation to talk through how this fits your current stage.
