In almost every organization experimenting with AI, the same group emerges at some point: two, maybe three people who get it, who started trying things out themselves, who enthusiastically talk about what's possible. They build a prototype, test a tool, share results in an internal chat. The rest of the organization watches, nods, and continues working the way it always has.
After a few months there is a list of interesting experiments and no change whatsoever in how the work is actually done. The two enthusiasts are still enthusiastic. Everyone around them has not moved along.
The reasoning behind this approach sounds sensible: start small, with people who are motivated, and let success spread on its own. That does work for some changes. A new way of working that visibly saves time sometimes spreads organically as colleagues adopt it.
With AI, that mechanism usually doesn't work, and the reason lies not with the people but with the organization around them. Enthusiasts often have access to systems, time and mandate that others don't have. What is for them a matter of trying something out for an afternoon requires, for a colleague without that access, a request, an approval, an IT ticket. The enthusiasm is not the problem. The absence of a path along which others can follow is the problem.
The order in which AI readiness builds up does not start with people who are willing. It starts with the structure that makes it possible: how the organization is set up, what IT infrastructure is in place, how data is managed. Two enthusiasts can work around that structure, with their own access, their own tricks, their own workarounds. The rest of the organization cannot do that, and should not want to — workarounds are no basis for something that needs to keep working.
That explains why the pattern often repeats alongside other well-known pitfalls. Anyone who reads about what needs to become clear before a pilot leads to anything sees a similar mechanism: activity without agreement on what success means remains activity. With two enthusiasts, that agreement is likewise absent — neither about what needs to become clear, nor about who follows after them.
There are a few recognizable signals. The first is that the same two or three names keep coming up whenever AI is discussed, in every meeting, in every update. The second is that questions from other departments start with "how did they do that" instead of with a route of their own. The third, and most underestimated, is that no one can explain what should happen if one of those two people leaves tomorrow.
A fourth signal is more subtle: the experiment keeps working in its own corner, disconnected from the rest of the process. That connects to what happens when a pilot outside its own team finds no connection — enthusiasm without a link to the rest of the organization remains an island, however well the island itself functions.
A fifth signal, finally, is that scaling up is proposed as the next step, while no one has checked whether the foundation underneath is already in place. That risk is described in what happens when scaling up is attempted before the foundation is in place: two enthusiasts who want to multiply their own success to ten teams run into exactly the structure they themselves had bypassed.
It is tempting to conclude that more enthusiasts are needed, or better training, or a culture program. That misses the point. The question is not whether there are enough driven people — there are often already two, and that is exactly the problem, because two is not a structure. The question is whether the organization as a whole — in how it is set up, how systems talk to each other, how available and reliable data is — can support a situation in which not two but far more people can work with AI. That is a different question from who is enthusiastic, and it is the question that needs to be answered first.
For anyone in a leadership role running into this pattern, it's useful to know what someone in a comparable position would pay attention to: what a CEO would pay attention to when assessing AI maturity describes a number of those points of attention, as does the perspective of what a COO would pay attention to regarding AI maturity from a different angle on the same organization.
If two enthusiasts demonstrate that something is possible, the question remains unanswered how much of the work can actually be taken over, and by whom. That question does not concern readiness but the work itself: which tasks lend themselves to being taken over by AI and what portion that involves. That is precisely what the work scan from FTE TO AI is meant for — it calculates, per task, what portion of the work can be taken over, regardless of who is already enthusiastically experimenting with it.
Vraag maar wat er moet staan voordat AI in uw organisatie kan landen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.