YOUR PEOPLE HAVE ALREADY STARTED
Your people are already using AI in more places than your leadership can see, and what they lack is practical ability rather than knowledge of the rules. You leave with a competence plan your leadership team can decide on: which role needs which capability, who owns it, and how it holds when the tools change again.
THE THESIS
Almost every leadership team that calls us opens the same way: our people need AI training. That reading is reasonable, and it is where we start. It is also, in our experience, a description of the instrument rather than of the problem.
What we consistently find is narrower. Confidence about what is allowed runs ahead of the ability to use the tools, so more policy moves nothing and more practice moves a great deal. And the use is already there, spread across roles and units, most of it begun on someone's own initiative and never reported upwards.
So the task is rarely an introduction. It is channelling: naming what is already happening, deciding which of it should become standard practice, and giving each role the capability that role actually needs. The energy exists. What is missing is a plan pointed at your goals.
WHAT THE WORK IS
The use that already exists comes to light: who built something on their own, and which tasks people have quietly stopped doing by hand. The picture comes from your own people describing their own work, not from our assumptions. The output is an inventory of practice, not of software licences.
Measured apart, because they move independently and only one of them is usually the bottleneck. When staff are markedly surer of what is allowed than of what works, more policy will not move anything. More practice will.
A leadership team needs to judge decisions and set boundaries. Specialists need craft. Everyone else needs the standard steps for their own tasks, in their own systems. One programme for all three serves one of them. It is also how the AI literacy duty in the EU AI Act is written: the measures must take account of the knowledge and the context the role actually has.
Capability decays when nobody keeps it current, and the tools change every quarter. The owner is rarely IT and never a consultant. We write the role into the plan, with what it is expected to do each term.
Ampliro delivers the leadership sessions. Role-specific programmes run through AIUC, our education arm. Both are measured against the plan: what people do differently in the workflow, not how many completed a module.
WHEN WE COME IN
So the specification describes the capability you need rather than the course that happens to be on offer. It often narrows the scope and lowers the cost.
Attendance was high and nothing moved. We measure where it stalled, which is usually ownership or workflow rather than content, and fix that first. Sometimes the honest answer is that training was never the lever.
Capability built at the same time as the system or the policy, so the people who have to run it are ready when it lands. That applies as much to an AI implementation as to settling your AI governance.
QUESTIONS
An AI competence plan names the capability each role needs, who owns it, how it is delivered and how it is measured. It starts from a baseline: what people already do with AI, how sure they are of the rules, and how able they are in practice. It ends in a decision a leadership team can take, with a scope and a price, rather than in a training catalogue.
No. Corporate AI training is one way to deliver AI competence, and rarely the first thing missing. In the organisations we measure, the recurring gaps are ownership, practical routine, and clarity about which data may be used with which tool. Where training is the right answer, Ampliro delivers the leadership sessions and AIUC, our education arm, delivers the role-specific programmes, measured for effect rather than attendance.
Article 4 of the EU AI Act requires you to take measures that promote AI literacy among your staff and among others who work with AI on your behalf. It has applied since 2 February 2025, and since July 2026 it is explicitly a duty of effort: you are not asked to guarantee a level of knowledge in any individual. In practice it still means the same two things: the leadership team understands what it has decided, and every role that touches an AI system knows what applies to that role. It is a floor, not a plan.
Partly, and you should. Elements of AI, and the guidance published by Digg and AI Sweden, are good, and we recommend them as a starting point. They answer what AI is and what the rules say. They cannot answer which of your own workflows should change, who owns the new step, or what your specialists need that your managers do not. That is the part we do, and it sits next to our work on AI adoption and change.
The senior who scopes the work stays accountable for it throughout, and any specialist engaged for part of it is named to you before the work starts. No anonymous delivery, and nobody swapped out midway. Role-specific programmes are delivered by AIUC, our education arm, and there too you know who is standing in the room.
Sweden and the Nordics are where most of our work sits, but location is a matter of agreement rather than a limit. Leadership sessions and role-specific programmes are delivered on site or remotely, in English or Swedish, and we travel where the engagement warrants it. What does not change with the country is the substance: the EU AI Act applies across the whole union, so the role-by-role competence logic holds wherever your people sit. Outside the EU, the regulatory part is scoped against whichever framework actually applies to you.
A competence baseline and plan is a defined engagement with a fixed price and an end. For Swedish public bodies it is normally scoped to sit under the direct award threshold, which the Public Procurement Act sets at 700,000 kronor for goods and services, unchanged from 1 January 2026 even though the EU thresholds were revised upward. Elsewhere the same fixed-price logic applies without that constraint. Delivery of role-specific programmes through AIUC is priced separately, so you can decide on the plan before you commit to the delivery.
Against a baseline taken before the work: self-rated confidence split between rules and practice, the share of relevant tasks done with AI, and time and quality in those tasks. The same measurements are repeated at follow-up points agreed in the engagement. Attendance and completion rates are recorded, and they are not the measure.
INSIGHTS
The share of firms that have adopted generative AI is now effectively the same in the EU as in the US. The share using it in more than two activities is not. The difference is not the technology but how many people can use it, and that is also what is left of Article 4 after this summer's amendment.
Read the analysis
In Sweden 88 percent say employees should be involved when new technology is introduced. A Swedish union has measured what actually happens, and the distance between the two is not a matter of opinion.
Read the analysisOne conversation is usually enough to tell whether you need a plan or a programme. Bring whatever measurements you have. If you have none, that is the first piece of work, and it is small. And if the honest answer is that competence is not your bottleneck, you will hear that.