Generated image with ChatGPT
In almost every organisation we work with, the same situation eventually emerges. Some people have been using AI tools as part of their day-to-day work for months. They have experimented with them, developed their own workflows, and can no longer imagine carrying out certain tasks without them. Others have barely opened one of these tools. They may be genuinely reluctant, or they simply have not had the time or the right context to explore them. Between these two groups, a gap emerges, and it shapes the day-to-day experience of the organisation.
This coexistence creates tension. Those who are further ahead may feel they are progressing on their own, without a framework to support or guide what they are already doing. Those who are further behind may feel they are being left behind. And the organisation is left without a shared approach, because each person is navigating their relationship with AI from their own starting point, with little to connect their experience to that of the rest of the team.
The mistake of treating this as a knowledge problem
The instinctive response tends to be training: organise a course for those who do not know how to use AI. But the problem is rarely technical knowledge. More often, it is the absence of a common approach.
Someone who has been using AI for months may still lack institutional guidance if nobody has established it. They may have developed habits and workflows, but not the framework that tells them how those practices align with the organisation’s values and risk profile.
Meanwhile, someone who has stayed on the sidelines does not necessarily need a tutorial on how the tools work. They first need to understand how AI could affect their specific work and what criteria should guide its use when they do start experimenting with it.
What the AI Act actually requires
Article 4 of EU Regulation 2024/1689, as applicable following the Digital Omnibus package, refers to promoting AI literacy among staff and does not require everyone to reach the same level or guarantee a particular level of expertise for every individual.
Beyond the wording of the legislation, however, there is a sensible principle worth keeping even as the regulatory framework evolves: what each person needs to know depends on what they do.
An administrator who does not operate an AI system does not need the same level of understanding as a data manager who works with predictive models every day. Aligning a team does not mean bringing everyone to the same point. It means ensuring that everyone has the level of judgement their role requires.
A shared baseline for the whole team
There is a first level that should apply across the organisation without exception: the rules of the game.
The AI use policy, what data can be used, who has the authority to approve new tools, and how to report a problem when one arises. This level does not require everyone to know how to use AI. It requires everyone to know what to expect.
It is the foundation on which any further differentiation should be built.
Specific training for high-impact uses
The second level concerns those who make significant decisions with the support of AI: analysts, commercial teams using research tools, or HR professionals working with recruitment and selection systems.
These people need a more specific understanding of the risks associated with the systems they use, as well as how to maintain meaningful human oversight rather than oversight that exists only on paper.
This is the level where mistakes can have tangible consequences, which is why targeted training is particularly important.
The level that is almost always overlooked: advanced users
The third level is made up of people who already use AI confidently. It is often overlooked precisely because they are assumed not to need support.
Yet someone who has been using AI independently for months still needs to connect their experience to the organisation’s judgement and standards. Without that connection, their individual way of working may be creating risks that neither they nor the organisation have properly assessed.
Ignoring advanced users means assuming that competence automatically equals responsible use. It can also mean leaving unsupported the very people who are adopting the technology most rapidly.
Why there is resistance and how to address it
Resistance to AI within a team is rarely about rejecting the technology itself. It often stems from uncertainty about one’s own role, the pressure of having to learn something new on top of an already demanding workload, or the fear of falling behind colleagues who are adopting AI more quickly.
That is why resistance cannot be solved simply through more technical training. It requires conversations about each person’s role, the value they bring beyond the tools, and a framework that makes clear that AI is a tool to be used with judgement — not a substitute for the person using it.
The role of leadership
Teams align when they see that their leaders also exercise judgement around AI, not simply when they receive a policy drafted from above.
If leadership uses AI without any visible criteria, or does not use it at all while asking the rest of the organisation to adopt it, the message becomes contradictory — and teams notice.
Leadership that clearly articulates its own criteria, including its uncertainties and its limits, creates far more buy-in than leadership that simply dictates rules.
In the end, consistency between what is expected and what is practised is what sustains alignment.
Frequently asked questions
How do I manage different levels of AI adoption across my team?
The first step is to stop seeing it as a knowledge gap and understand it as a judgement gap. The goal is not for the whole team to use the same tools, but for each person to have the judgement to decide when and how to use them according to their role.
The AI Act itself points in this direction by linking AI literacy to what each role requires, rather than imposing a uniform level of knowledge.
Should I require my team to use AI?
No. Forced adoption without judgement encourages uncritical use, which is precisely where risk increases. What you should ensure is that anyone operating AI systems has the appropriate knowledge and training to do so responsibly. That is different from pushing people to use AI in roles where it adds little or no value.
What should I do with employees who are reluctant to use AI?
Start by understanding why. Resistance is often a response to uncertainty about one’s role or the fear of becoming outdated, rather than rejection of the technology itself.
Open conversations about those concerns, combined with a framework that presents AI as a tool to be used with judgement rather than as a replacement for people, can change the dynamic far more effectively than imposing adoption.
