AI / KMO / change management / AI literacy / operations
Your team doesn't trust AI yet. Here's how to change that.
The technology works. The business case is solid. The pilot proved the time savings. And then nothing happens. The tool sits there, half-adopted, while the team quietly goes back to the old way of doing things.
I see this pattern more often than I see technical failures. The AI is ready. The people are not. And no amount of training slides or all-hands announcements fixes that, because the resistance is not about understanding. It is about trust.
Why people resist (and why it is rational)
When someone has done a job a certain way for years and you introduce a tool that does part of it automatically, three things happen in their head simultaneously.
First: “Is this replacing me?” Even when management says no, the question lingers. If the tool handles 60% of what I do, what happens to me in six months? This is not paranoia. It is a reasonable inference.
Second: “I do not trust the output.” They have seen AI produce confident nonsense. They know the edge cases the tool probably does not handle. Their judgment tells them to double-check everything, which eliminates the time saving and confirms their suspicion that it was not worth it.
Third: “We have always done it this way, and it works.” Change has a cost even when the new way is better. The old way is predictable. The new way is an unknown, and unknowns carry risk. People instinctively avoid risk they did not choose.
All three reactions are rational. Treating them as ignorance or resistance to progress is the fastest way to lose your team.
What actually works (from watching it happen)
The companies where I have seen adoption succeed share a few traits, and none of them involve mandatory training or executive mandates.
Let people choose their first use case
Instead of rolling out a tool and telling everyone to use it, ask: “What part of your job do you wish took less time?” Let individuals pick the first thing they automate. When someone chooses their own pain point, they have skin in the game. They want it to work. They tolerate the early friction because the payoff is personal.
This means the rollout looks messy from above. Different people using different features at different speeds. That is fine. Uniformity can come later. Buy-in cannot be retrofitted.
Show savings in their own role, not company averages
“This tool saves the company 200 hours per month” means nothing to an individual. “You spent 3 hours on those reports last week, and here is that same work done in 20 minutes” lands immediately. Make it personal. Use their data, their tasks, their time.
Create a buddy system with early adopters
In every team there are one or two people who naturally experiment. They tried the tool on day one and found three things it does well. Pair them with the skeptics, not as trainers but as peers. “Ask Sofie, she figured out how to get it to handle the edge cases” works better than a formal training session because it comes without hierarchy.
Keep humans in the loop visibly
The fastest way to kill trust is making people feel bypassed. If the AI drafts something, let a person review and approve it. If it makes a classification, show the confidence score and let someone override it. Visible human oversight is not inefficiency. It is how you build the trust that eventually allows more autonomy.
A practical four-week plan
Week 1: Have one-on-one conversations (not announcements) with each team member. Ask what frustrates them. Listen more than you pitch. Identify two or three people who are curious.
Week 2: Let the curious ones experiment freely. No KPIs, no reporting requirements. Just: try it on one real task and tell me what happened. Collect their honest feedback.
Week 3: Share early wins informally. Not a presentation. A Teams message: “Sofie used the tool for her weekly report and it took 15 minutes instead of 2 hours. She’s happy to show anyone who’s curious.” Let people come to it.
Week 4: Offer structured time (one afternoon) for anyone who wants to try it with support available. Not mandatory. Not tracked. Just space and help. By now, the curious middle has usually started moving.
The legal reason this matters now
Since February 2025, the EU AI Act’s Article 4 requires companies deploying AI systems to ensure sufficient AI literacy among the staff who interact with those systems. This is not optional. It is a legal obligation that applies to every company using AI tools, regardless of size.
AI literacy does not mean technical expertise. It means that the people using AI tools understand what they do, what their limitations are, and when to exercise judgment. That understanding comes from the kind of gradual, supported adoption described above, not from a one-day training that ticks a compliance box.
The bridge is not training, it is trust
The technology is ready. Your team needs a bridge, and that bridge is not a better explanation of how the model works. It is the experience of using it on their own terms, seeing it help with something they actually care about, and knowing they can say no without consequences.
Build that experience deliberately, and adoption follows. Skip it, and you end up with expensive tools gathering dust while everyone pretends to use them.