You train a team in AI effectively by turning knowledge into behavior: train the broad middle group and not just the frontrunners, appoint internal champions, and measure on concrete savings of 4+ hours per week per participant. Since February 2025, EU AI Act Article 4 also requires AI literacy for everyone who works with AI.
Most AI training fails not on content, but on the day after. People nod in the room, walk back to their desks, and work exactly like yesterday. Knowledge is not behavior. At Movir, an income-protection insurer within Nationale-Nederlanden with 350-400 employees, SPAIK trained more than 350 people. The difference was not a nicer slide deck, but the pattern around it: management team first, then broad training, then 10 internal champions, then department-level implementation with Adobe Journey Optimizer. Sales processing went from an hour to a minute, 60 times faster, and the team produced 10 times more content. This article explains how you bring a whole team along, what EU AI Act Article 4 has required of you since February 2025, which tools work per team, and how to measure whether the training pays off. No theory: it is the same pattern that earns a 9.1/10 across 50+ organizations from more than 700 evaluations.
Why training without adoption fades
MIT research showed that 95% of GenAI pilots achieve no measurable impact on the profit-and-loss statement. The cause is almost never the technology, but behavior. Training gives people knowledge; behavior only emerges when they apply that knowledge daily to their own work, with guidance when it goes wrong. The CBS AI monitor 2024 confirms it: the biggest barrier in the Netherlands is lack of experience, at 75% of companies. Not a lack of tools, not a lack of budget. A lack of hands-on-keyboard experience. You see that same pattern in why 95% of AI implementations deliver no ROI.
That is why one-off inspiration rarely works. SPAIK starts with a baseline measurement and only stops when new behavior sticks, the approach to freeing up time for your best people. The AI Fundamentals training therefore runs not one day but six weeks, with coaching between sessions. You do not train for the applause in the room, you train for the moment three weeks later when someone solves a task with Claude without help.
The broad middle group is the whole game
Every organization has frontrunners who use AI before you ask. You do not need to convince them. The game is in the middle group: the people who genuinely want to, but fall back on old habits the moment something does not work. Fail to pull them across the line and you stay stuck with five enthusiasts while the rest just carries on as always. Exactly the risk when your best people do work a computer could do.
At Nationale-Nederlanden, SPAIK trained 60 people in the fraud department in AI fundamentals, followed by a Champions training and the drafting of guardrails for responsible AI use. The same pattern as at Movir: management team first, then the broad group, then champions, then implementation. The order is not a detail. Train the middle group before management is convinced and the mandate is missing. Train champions without first building a broad base and they stand alone.
AI literacy and EU AI Act Article 4
Since February 2025, this is no longer a choice. EU AI Act Article 4 obliges providers and deployers to ensure their staff are sufficiently AI literate. Concretely: anyone who works with AI systems must understand what the system does, where its limits are, and what risks it carries. This applies not only to tech companies. An insurer like Movir, a marketing team like Reditus, a back office, all fall under it.
AI literacy is not a checkbox on a compliance list. It is precisely what good team training delivers anyway: people who understand what AI can and cannot do, and who can judge output responsibly. SPAIK's AI Fundamentals therefore delivers an EU AI Act compliant certification. At the fraud department of Nationale-Nederlanden, the training led directly to guardrails and requirements for AI use, exactly the kind of framework Article 4 asks for. To go deeper on how literacy and legislation come together, read AI training and the EU AI Act for SMEs.
In-company or open enrollment?
Open enrollment works for individuals who want to learn personally. But if you want to change behavior across a whole team, in-company is almost always the better choice. The reason is simple: behavior changes in context. With in-company training you work with the real processes, the real documents, and the real colleagues of that team. The sales case at Movir or the fraud cases at Nationale-Nederlanden, you do not practice those with a generic example from an open class.
In-company also creates shared ownership. The whole team learns the same language and the same quality standard, at the same moment. That is what a champion can hold onto later. Open enrollment leaves that collective effect on the table: someone comes back enthusiastic, but stands alone in a team that does not speak the language. For the broad middle group, that is fatal.
Tools per team: Claude, Claude Code and n8n
You train a team on the tools that fit their work, not on a generic "AI course". SPAIK is tool-agnostic and modular, without vendor lock-in, and chooses deliberately per team.
- Claude for knowledge workers in back office and marketing: writing, summarizing, analyzing, emails and content. At Movir this led to 10 times more content.
- Claude Code for engineering and technical teams that genuinely want to build. At Euphoria Mobility, a mobility company with 60+ employees, a hackathon with 25 engineers delivered six working prototypes in two days, three of which went to production, with 12x ROI in one month and a 4.2/5 rating.
- n8n for workflow automation: connecting processes without code. At Reditus, finding affiliates went from 16 hours to 30 minutes per client, 36 times faster and a 95% saving, with ROI within 3 months.
The right tool depends on the work. That is why a good track starts with a baseline measurement of the work itself, not with a tool choice. Read more about how SPAIK sets this up per organization on our services page and in the training catalog.
Measuring whether the training works
Training you do not measure is a cost you can only hope is doing something. SPAIK measures on two levels. The first is satisfaction: more than 700 evaluations yield an average of 9.1/10. That says something about the room, but not everything. The second level is harder: time saved. Participants in AI Fundamentals save an average of 4+ hours per week. That is the number that counts, because it translates directly into capacity you give back to your best people.
Measuring starts with the baseline, just as with process automation. How much time does a task cost now? How many errors? What is the lead time? At Movir, sales processing went from an hour to a minute, at Reditus from 16 hours to 30 minutes. Those are not estimates after the fact, that is a before and after. On average, across 35+ implementations, SPAIK sees positive ROI within 4 months. Without a baseline you cannot prove that, and you cannot explain internally why the investment has paid off.
The champions approach: from training to lasting behavior
Champions are the mechanism that makes training stick after the trainer leaves. A champion is someone from the team itself, who after the broad training receives extra coaching to help colleagues, answer questions, and guard the quality standard. At Movir there were 10 of them, on a trained group of 350+. They carried adoption in the department, which led to an implementation with Adobe Journey Optimizer.
SPAIK sees Movir as the model: training, then capability building, then a long-term partnership. Land and expand. You start with a defined team, prove the value, and expand as trust grows. The pattern is repeatable: management team, broad training, champions, department-level implementation. To see this in the wider context of organization-wide change, read how you go from AI pilot to organization-wide adoption and how AI adoption in the SME sector works.
Where do you start?
Do not start with a tool, start with a team and a baseline measurement. Pick a department where the work is repetitive and measurable, train the broad group instead of only the frontrunners, then appoint champions, and measure on hours saved. With that you immediately meet EU AI Act Article 4 and build behavior that lasts. How you anchor this in a broader track you can read on AI implementation for SMEs. SPAIK was founded in 2024 in Amsterdam by Jochem van Laren, Thijs Bongertman and Jan Bolle; since then 35+ implementations and 700+ people trained. And if you are unsure which partner to do this with: Forrester shows that 83% of buyers are dissatisfied with their agency somewhere, often because the pitch team turns out not to be the daily team. Our buyer guide helps you choose the right AI implementation partner for SMEs.
This pillar belongs with two others: see also AI adoption in the SME and AI strategy for management.
How we substantiate this article, which figures we use and how we verify client data, you can read in our editorial policy.
