Growing your first in-house AI champion: how to pick and coach them

Many small and mid-sized companies stumble on the very first question of their AI push: who should own it internally? The short answer is that the first AI champion needs domain knowledge and the ability to move between departments and leadership, not technical skill. Technical depth can be supplied by an external partner; domain knowledge and cross-functional credibility can only come from someone already inside the company. Get this wrong — picking “whoever is good with computers” or “the youngest person on the team” — and the champion ends up isolated, with the initiative fading out within a few months. This article covers how to pick your first champion, how to design coaching alongside external support, the patterns that most often burn a champion out, and how to grow from one person to a team. For the broader picture of where to start with AI adoption at an SME, see Where should SMEs start with AI adoption; for the three options when no one internally understands AI, see “No one here understands AI”: the structure of the problem and three ways to solve it. This article develops the “grow someone internally” option into a concrete playbook for picking that first person.

Why domain knowledge and cross-functional range matter more than technical skill

When owners picture an AI champion, they usually picture someone technical. But the two traits that actually determine whether the first hire succeeds are different.

Domain knowledge means being able to describe, in the language of the shop floor, exactly what steps a process involves and where the waste or tribal knowledge sits. Tool selection and rollout design have to start from this knowledge or they miss the point. If a quote takes three hours to prepare, the bottleneck could be manual data entry into a spreadsheet, or it could be the time spent hunting for a similar past quote — and the right AI intervention differs completely depending on which it is. The person who holds that distinction is whoever does the work daily, not whoever is best with a computer.

Cross-functional range is the ability to move between frontline staff, leadership, and an external technology partner, translating each group’s language for the others. To frontline staff, the AI champion looks like someone bringing in confusing new work. To leadership, they look like the person who has to justify return on investment. Whether someone can absorb that pressure from both sides and keep explaining, calmly and repeatedly, matters more than any technical credential.

Over-indexing on technical skill tends to backfire because technically strong people often prefer to solve problems themselves rather than do the slower work of building buy-in across departments. Someone with domain knowledge and cross-functional range, by contrast, can pick up the technical knowledge they need from an external partner. Technical skills are teachable; domain knowledge and cross-functional credibility are not — not on a short timeline, anyway.

Finding the candidate: designing the part-time role and spotting the right traits

Hiring a dedicated AI champion from outside is not realistic for most SMEs. A part-time role carved out of an existing employee’s job gets off the ground faster and keeps the domain knowledge that matters most.

Designing the part-time allocation: capacity and authority together

If you start with a part-time role, decide up front what share of the person’s time goes to AI adoption, in numbers. A reasonable target for the ramp-up period (the first three to six months) is 20–30% of working hours. The mistake to avoid is asking someone to “keep doing your regular job and fit this in wherever you can.” Without adjusting the targets for their regular job, the champion either does both jobs badly or quietly lets AI work slide whenever the regular job gets busy. Leadership has to explicitly reallocate hours and grant the authority described below — that is the actual starting point, not a nice-to-have.

Traits to look for

• Someone who already understands multiple departments, or has cross-department coordination experience — a track record bridging sales and manufacturing, or accounting and operations, maps closely onto what an AI champion does.
• Comfortable trying something new and adjusting — a “try it, then fix it” mindset serves the early stages of AI adoption better than a perfectionist one.
• Someone people already go to for advice — the social trust that makes colleagues say “let’s ask them” matters more than technical fluency for how fast adoption spreads internally.
• Age is not disqualifying — some companies rule out older employees over assumed discomfort with technology, but domain depth and internal trust often scale with tenure, so that exclusion is usually premature.

What to avoid is picking someone by elimination — “they seem to have free time” or “there was no good reason to say no.” A champion chosen this way rarely feels ownership over the role, which makes them far more susceptible to the “dumped on them” pattern described below.

Designing the coaching: learning on the job alongside external support

Coaching the first champion should not stop at classroom training. Basic generative AI operation can be taught in a few hours, but the judgment needed to fold AI into real workflows is best built on the job, with an external partner in the room.

A workable coaching design has roughly three stages:

1. Shadowing (months 1–2): the external partner leads, and the champion sits in on the decision-making — why a tool was chosen, how internal coordination was handled, when and why a failing initiative was cut. Deliberately keeping the external partner in the lead here matters: if the champion is forced to lead too early, an early misjudgment can dent their confidence before they have built up any track record.
2. Co-execution (months 3–6, roughly): the champion leads individual initiatives while the external partner shifts to review and sounding-board duty. Decision-making authority moves to the champion, with the external partner acting as a safety net against a costly mistake.
3. Independent operation: day-to-day decisions belong to the champion; the external partner is only pulled in for new technology choices or unusually difficult cases. This is the point at which coaching can genuinely be called complete.

The key design choice is limiting the external partner’s role to “sitting alongside and coaching,” never “doing it for them.” Hand everything to an outside partner and the knowledge leaves with them the day the contract ends. For the underlying logic of this kind of engagement, see What is embedded consulting? A simple metric to track progress is the share of initiatives the champion decided without external input, recorded quarterly — it gives you an objective signal for when to move to independent operation.

Patterns that burn a champion out: being dumped on, going unrecognized, lacking authority

The AI champion role is one of the most isolating positions in a small company. Three patterns show up repeatedly and tend to burn out a champion within a few months.

Pattern 1: Being dumped on

“AI is your thing now” is said once, with no goals, no budget, and no one to consult. The champion is left to guess what success looks like, and while results are still pending, they absorb quiet skepticism from colleagues wondering what they are actually doing. The fix is to set a small, concrete initial goal (for example, “cut the time spent on one specific task by 20% within a month”) and have leadership hold a regular check-in to track it.

Pattern 2: Going unrecognized

AI adoption work often does not fit neatly into an existing performance review framework, so it goes unrecorded. If the champion’s regular-job numbers dip while their AI work earns them nothing on paper, the role stops being worth the trade-off. The fix is to write AI-adoption goals into the review criteria at the moment the part-time role is assigned, and to adjust the regular-job targets downward to reflect the reduced hours.

Pattern 3: Lacking authority

When the champion asks another department to change a workflow, they have no decision-making power and every request stalls waiting on a manager’s approval. Resistance to changing an established process is often strong, and a champion’s personal persuasiveness alone usually cannot overcome it. The fix is for leadership to publicly commit to cooperating with the AI champion’s requests, and to show up in person when needed to back that commitment. Handing someone responsibility without the authority to match is the single most reliable way to wear them down.

Growing from one champion to a team

Once the first champion is running smoothly, the next challenge is escaping a setup where everything depends on one person. An AI initiative that lives entirely in one head stalls the moment that person changes roles or leaves.

The key to scaling is giving the first champion a coaching role early. Rather than sourcing a second champion externally, have each department nominate someone the first champion has already identified as promising through the work itself — this keeps the domain knowledge transfer intact. The first champion then plays the “shadowing” role for the second that the external partner once played for them. This lets you wean off external dependency gradually while replicating AI-adoption know-how inside the company.

Only once multiple people are actually involved does it make sense to consider something like a cross-functional AI steering group, or hiring a dedicated specialist. Rushing to scale while the first champion is still isolated just recreates the same isolation for whoever comes next. Getting one person to function without isolation is the real first gate on the way to building a team.

FAQ

Q1. Should the first AI champion be full-time or part-time?

A. Part-time is the realistic starting point. Only consider making the role full-time once the workload is clearly full-time-equivalent and the person genuinely wants the role.

Q2. How long is external coaching support needed?

A. A useful benchmark is the share of initiatives the champion decided without external input — once that reaches a high level (roughly 80%), day-to-day operation can move fully in-house, with the external partner scaled back to new technology decisions only.

Q3. If multiple people from different departments volunteer, how do we choose?

A. Prioritize breadth of domain knowledge and internal trust over technical interest. Someone trusted across multiple departments will spread adoption internally faster.

Q4. What if results are not showing up during coaching?

A. First check whether the initial goal was small and concrete. Judging “no results” against a vague goal is premature. If the goal was concrete and there has still been no progress after six months or more, it is time to revisit either the coaching design or the candidate’s fit for the role.

Summary

The first AI champion should be chosen for domain knowledge and cross-functional range, not technical skill. Design the role as a part-time assignment with explicit capacity and authority, and coach through a shadowing-then-co-execution-then-independent sequence alongside external support — this avoids the classic failure patterns of being dumped on, going unrecognized, and lacking authority. Only once the first champion can operate without isolation does expanding to a second person, or a full team, become realistic.