AI skills
AI training for employees starts not with the tool but with where people stand.
A five-minute baseline, a short learning moment every working day across 60 working days, and the same measurement again on day 40 and 60. Vendor-neutral where possible, in your own environment where necessary.
Why one training day delivers too little
01 · The problem
The standard response to AI in an organisation is to buy a day. That's understandable, well meant, and it's the format we see the least of in the re-measurement.
The first problem is the spread in levels. Every team has someone who has been working with AI for months and someone who has never opened it. A classroom day necessarily aims at the middle, which bores the first group and loses the second. Exactly the two groups you wanted to reach.
The second is that a day doesn't change behaviour. What people learn on a Wednesday is largely gone in six weeks if nothing follows. That's not a criticism of trainers; it's how learning works. Something has to come back at the moment the novelty wears off, and that's exactly when the trainer has long since left.
The third is that you can't demonstrate anything afterwards. You have an attendance list and a feedback form. Both measure something other than whether anyone works differently. In a conversation with your board that's too little, and since the AI literacy obligation it's formally too little as well.
When a day is right
For a small team at the same level with one concrete question. Twenty people, all advanced, one topic: then you buy a single advanced training, not a programme, and you're done. We'd rather say that up front than have you discover it afterwards.
What's in the programme
02 · The structure
5 min
The baseline
Role, work context, knowledge level and self-assessment per employee, filled in on working day five in about five minutes. Out comes a matrix of five competencies against the teams, showing exactly where the gain is.
Every working day
26 learning moments
26 mandatory learning moments in three phases across 60 working days: a starting week that gets you up to speed quickly, then one short learning moment per working day with practical assignments in your own work environment. The programme is adaptive: do well on a test and you get more complex cases and deeper material instead of repetition.
60 days
Reinforcement
After the certificate comes a 60-day reinforcement phase, with a challenge on day seven and a refresher on day thirty. After that, the question whether it is still being used returns every quarter. This is the part that makes the difference between a training and a change.
Day 40 and 60
The re-measurement
The same questions as at the start, on day 40 and day 60. The difference between the measurements is the only figure that helps in a board conversation, including when it disappoints, because then at least you know which team and which competency.
Vendor-neutral where possible
03 · Which tools
The skill isn't in the buttons. Anyone who can frame a good task and judge the output does so just as well in Copilot as in ChatGPT, and keeps doing it when something else arrives next year.
In practice nearly every Dutch organisation uses Microsoft 365, so the examples usually come from there: a report in Word, an export in Excel, a meeting in Teams. That's a choice of familiarity, not principle. If ChatGPT or Gemini sits alongside it, more often than IT thinks, that appears in the modules too.
What's the same everywhere: that the model writes convincingly regardless of whether it's right, that it adopts your assumptions rather than challenging them, and that what you type in ends up somewhere. Those three things are the core of every module, and they don't age with the next product update.
Where it does become product-specific, we keep it separate. The Copilot-specific subjects (Excel analysis, agents, Cowork) live in their own modules and sessions, so an organisation that doesn't use Copilot doesn't pay for them.
Questions about AI training
04 · Short answer
How does this differ from your Copilot training?
In the examples, not in the approach. The skills that matter, framing a task, judging output, knowing what not to enter, are the same regardless of the model underneath. If your organisation mostly uses Copilot, we use that interface. If ChatGPT or Gemini sits alongside it, that comes along too.
We haven't chosen a tool yet. Can we still start?
Yes, and it's actually a good moment. The baseline shows where people stand and which tasks lend themselves to AI, which is exactly the information you need to choose a tool. We regularly see an organisation reach a different decision after the measurement than the one it was heading for.
How much time does this cost an employee?
About five minutes for the baseline on working day five, and beyond that one short learning moment per working day, a small part of the day, for 60 working days. Public holidays stay quiet and holiday leave pauses without backlog, so it fits around the work. Clearing a full day isn't necessary and usually backfires with this audience.
Does this work for people who want nothing to do with it?
Partly. What the e-learning solves is not knowing, and that's the underlying cause of a surprising amount of resistance. What it doesn't solve is a conversation that has to happen, fear about one's own job, or the sense that this is being pushed through. That needs a session in a room, and we deliberately schedule it before the modules rather than after.
Do participants get a certificate?
Yes, issued automatically after the final test, and publicly verifiable. Be aware that a certificate proves completion, not skill. The figure that helps you in a conversation with your board or a regulator is the difference between the baseline and the re-measurements on day 40 and 60.
Does this programme also cover the legal obligation?
More than amply, yes. Article 4 implies four things and this programme delivers three: level assessment per role, targeted training and re-measurements on day 40 and 60 you can record. The inventory of which AI systems are in use is yours to make, that knowledge sits with IT. If it's purely the obligation you're after, the ninety-minute e-learning is cheaper and faster.
How does this differ from the AI literacy e-learning?
Purpose and scope. The e-learning runs ninety minutes, is one-off, and covers the knowledge base the law asks of everyone: that's compliance. This programme runs for 60 working days in short daily sessions, with a re-measurement on day 40 and 60 and 60 days of reinforcement after that, and is meant to build skill. Anyone who only wants to cover the obligation needs the first.
Read on
- Going further with Excel analysis
The deeper session for anyone who wants more than having formulas explained.
- Just the legal base
Ninety minutes with a certificate, if you want to cover the obligation and no more.
- The same course, from the participant's side
What you learn in it, and why it works just as well if you use ChatGPT.
- AI literacy and the obligation
Why the same measurement doubles as your justification under Article 4 of the AI Act.
- The 26 learning moments
What each phase covers, how the 60-working-day rhythm runs and which assignments go with it.
- The Copilot version
The same approach, but entirely within Microsoft 365 and with Copilot as the starting point.
Measure first, then choose
Request a demo including the baseline measurement. You'll see where the level sits for your own group, and only then does the question of which programme fits become answerable.