Recently, I was talking with an AI implementation specialist who works directly with groups of AI champions inside organizations.
She told me that a typical client selects five people from across the business. They are usually smart, interested in AI and full of ideas about where it could be used. Then the implementation work begins and, in her experience, two of the five often never really get started.
They attend the early sessions. They appear engaged. Their use cases never move much beyond an idea.
It would be easy to conclude that they lacked time, needed more training or simply were not as interested as they first appeared. Any of those things could be true. I think there is another question firms need to ask:
Are we choosing AI champions for the behavior that gets someone noticed at the beginning, rather than the range of behaviors required to make adoption stick?
Enthusiasm is visible. The rest of the role is not.
When firms ask for AI volunteers, the people most likely to be selected are often the people whose enthusiasm can be seen.
They are curious. They have ideas. They speak confidently about what might be possible. They are willing to experiment in public and they put up their hands before everyone else has decided how they feel.
Those are useful qualities. They are also only a small part of what an AI champion group will eventually be expected to do.
AI champions are not simply early adopters or enthusiastic users. They are being asked to help the firm move from individual experimentation to trusted, repeatable ways of working. That can involve identifying valuable use cases, testing outputs, documenting new workflows, navigating risk, teaching colleagues, gathering feedback and maintaining momentum after the novelty has gone.
OpenAI describes AI champions as the people who help an organization turn individual experimentation into repeatable, useful ways of working. It distinguishes between those who provide direction and structure and those closer to the work who help colleagues put AI into practice.
That distinction matters. Someone can be an exceptional personal user of AI and still be poorly placed, insufficiently supported or behaviorally unsuited to perform every part of the champion role.
In a professional services firm, the challenge is even sharper
AI adoption in a professional services firm rarely sits neatly inside one department.
The most valuable opportunities are often buried inside the work of lawyers, accountants, consultants, architects, recruiters, advisers and client-service teams. The people who understand those opportunities are also under pressure to remain billable, protect client confidentiality, exercise professional judgment and deliver work to an existing standard.
This means the champion group has to do more than create excitement about a new tool. It must help the firm distinguish a clever demonstration from a dependable way of working.
It has to ask:
- Does this solve a problem that matters?
- Can the output be trusted?
- What human judgment must remain?
- Can the process be repeated by someone other than its creator?
- Will colleagues actually use it?
- Does it improve the way the firm delivers value to clients?
Selecting five enthusiastic people does not guarantee that the group will naturally ask all six questions.
AI adoption requires four different forms of work
As I have started applying behavioral profiling to AI champion groups, I have been thinking about the work successful adoption requires. Four distinct contributions keep appearing.
1. Explore
Someone has to experiment, see possibilities and tolerate imperfect early results. Without this contribution, the firm waits for certainty while the opportunity moves past it.
The corresponding risk is experimentation without convergence. The group creates prompts, agents and demonstrations, yet never decides which ideas deserve to become part of the firm's operating model.
2. Validate
Someone has to test quality, accuracy, risk and relevance. In a services firm, this is essential. An output that is fast and plausible is not necessarily an output the firm should trust with a client.
The corresponding risk is allowing scrutiny to become a permanent holding pattern. If the standard is certainty before movement, nothing makes it out of the lab.
3. Operationalize
Someone has to turn a promising experiment into a repeatable workflow. That means clarifying ownership, documenting the process, building appropriate controls and following through when the first version does not work cleanly.
The corresponding risk is creating a dependable process that becomes too rigid to improve as the technology and the firm's needs continue to change.
4. Mobilize
Someone has to make the work meaningful to everyone else. They need to communicate the value, build confidence, listen to resistance and help colleagues use AI in the context of their actual work.
The corresponding risk is generating activity and excitement without the substance required for lasting implementation.
These are not four job titles, and they do not necessarily require four different people. They are four forms of work that the champion group must collectively be able to perform.
That is a very different selection question from, "Who is most excited about AI?"
Where behavioral profiling becomes useful
I have used DISC behavioral profiling with leaders, executive teams and boards for 20 years. The greatest value rarely comes from reading five individual reports. It comes from mapping the people together and interpreting what the group's collective behavioral composition makes more or less likely to occur naturally.
The same principle applies to an AI champion group.
A person with a stronger drive for pace and results may help the group move from discussion to action. That same drive may cause them to dismiss important detail or lose interest once the immediate result has been achieved.
A highly influential and optimistic contributor may see possibilities quickly and build energy across the firm. They may need structure and support to convert those possibilities into disciplined follow-through.
A steadier, more patient contributor may be slower to trust the technology, then become highly valuable in making adoption practical and helping colleagues feel supported through the change. If harmony matters strongly to them, they may find it harder to challenge an approach that others are excited about.
A more analytical and quality-focused contributor may test assumptions, expose risks and create the most reliable documentation in the group. They may also continue seeking information after the group has enough evidence to take a responsible next step.
None of these profiles tells us whether someone will be a successful AI champion.
DISC does not measure AI skill, capacity, commitment, credibility with colleagues or proximity to a valuable workflow. It should not be used as a pass-or-fail selection tool. It gives us another form of evidence: how a person is likely to approach the work, which contribution may come more naturally, where friction may arise and what support may help them remain effective.
Five capable people can still create an incomplete group
The greater opportunity is to examine the champions collectively.
A group full of fast experimenters may create impressive momentum while underweighting validation and implementation. A highly analytical group may build safe, sophisticated use cases that never spread beyond its own meetings. A collaborative and supportive group may bring people with it while avoiding the decisions required to stop weak experiments or challenge senior sponsors.
This does not mean every champion group needs one person from each DISC style. People are more complex than four boxes, and most organizations naturally lean toward certain behavioral patterns.
It means leaders need to understand what their particular group is predisposed to see, what it may overlook and what must be introduced deliberately into the way it operates.
That may result in practical guardrails:
- A decision deadline for a group inclined to keep investigating
- A validation checkpoint for a group that moves quickly
- Clear ownership for turning experiments into documented workflows
- A communication plan that begins before the solution is finished
- Protected time for champions whose client or operational work will otherwise win every week
- A designated challenger when the group tends to agree too easily
The aim is not to construct a theoretically perfect team. It is to help the team you have perform the whole job.
Selection is only the beginning
Behavioral composition matters, although it is not the only reason champion programs stall.
Champions also need:
- A clearly defined role and mandate
- Enough proximity to the work to identify meaningful use cases
- Credibility with the colleagues they are expected to influence
- Protected capacity to do the work
- Access to technical guidance and decision-makers
- Measures that extend beyond attendance, tool usage or the number of ideas generated
- An operating rhythm that continues after the initial launch
Recent research into AI adoption reinforces the importance of these conditions. BCG research into AI adoption found that employees approach AI with different adoption patterns and require different combinations of incentives, empowerment and leadership attention. It also identified insufficient protected learning time as a material barrier. Research discussed by Wharton similarly emphasizes peer mentoring, permission to experiment and champion networks as ways to move adoption beyond the already enthusiastic minority.
Behavioral insight cannot compensate for a champion who has no time, no authority and no meaningful connection to the work. It can help a firm stop managing five different people as though they will all engage, contribute and influence others in the same way.
Questions I would ask before choosing the group
Before naming your AI champions, ask:
- What is this group actually responsible for delivering?
- Does it collectively have the capacity to explore, validate, operationalize and mobilize?
- Are we selecting visible enthusiasm, or do we also have evidence of credibility, judgment and follow-through?
- Which contribution is likely to come naturally to this group, and which may need to be built deliberately into its operating rhythm?
- What does each person need from the role, the sponsor and the group to remain effective after the novelty has worn off?
- How will we know that AI has moved from interesting individual use to a trusted and repeatable part of the way the firm works?
AI adoption needs behavioral architecture
Firms are investing considerable attention in their AI technology architecture: which tools to use, where data sits, what connects to what and how risk will be controlled.
The human system responsible for adoption deserves the same level of intention.
Who explores? Who tests? Who turns the work into something repeatable? Who brings everyone else with them? What will this particular group do naturally, and what could it repeatedly miss?
These are not questions about who is good or bad at AI. They are questions about behavioral architecture: designing the conditions in which different people can collectively take a significant change from possibility to practice.
Your most enthusiastic AI user may become an exceptional champion. They should not be selected on enthusiasm alone.
The group that gets your firm excited about AI and the group that makes AI part of how the firm works may require different strengths. The real task is building a champion group capable of doing both.
