You Can't Buy AI Adoption, You Build It

Geschatte leestijd: 5 minuten

Inhoudsopgave

1. Why rolled-out tools stall
2. Shadow AI solves itself
3. What "building it yourself" does and does not mean
4. Don't start with a purchase, start with an afternoon
5. Frequently asked questions

You probably know the scenario. A company buys an AI tool, rolls it out with a short explanation, and three months later almost nobody uses it anymore. The licences keep running, the enthusiastic champion has dropped off, and the rest of the team has quietly gone back to the old way of working. Not because the tool is bad, but because no one really understands what it does or why it would help. The mistake is not in the software. The mistake is in the assumption that you can buy adoption. You cannot. You build adoption, and you may take that quite literally.

Why rolled-out tools stall

A tool that is handed to you feels fundamentally different from something you made yourself. With the first you stay a spectator: you get a manual, a login and the implicit message that you now have to work differently. With the second you become an owner. You understand why something works the way it does, you know its limits, and you trust it because you have seen it from the inside. That difference decides whether an investment sticks or slowly evaporates.

People rarely resist change they shape themselves. They resist change that happens to them. Without understanding you also get an unpleasant combination: on one hand distrust ("is this thing going to do my job soon?") and on the other hand inflated expectations ("this will surely solve all our problems"). Both lead to disappointment, and disappointment is the fastest route to an unused licence.

During an in-company workshop at TopX we saw this up close. The employees who had done the least with AI beforehand, and who were the most sceptical about it, turned out to be the most enthusiastic after just one afternoon of building. Not because they had seen an impressive demo, but because they had got something working with their own hands. The abstract concept "AI" had turned into something concrete that they understood and could explain to a colleague.

Shadow AI solves itself

Here is an uncomfortable truth: your employees are probably already using AI. You just do not know exactly where, how often, and with which data. Someone pastes a customer email into ChatGPT to quickly get a tidy reply. Another has a sensitive document summarised in a tool that no one has approved. This is called shadow AI, and it is a real risk when it comes to privacy, confidentiality and compliance.

Banning it does not work, because the time saving is too big and the threshold too low. What does work is offering a better alternative that people trust. Give your team a safe, internal AI solution that they understand and helped build themselves, and the secret tabs disappear on their own. Not because they have to, but because the official route is now more pleasant and faster than the workaround.

That also fits the EU AI Act, which obliges organisations to invest in AI literacy: employees must understand what AI does, where the risks are, and how to use it responsibly. Building something yourself is the most direct form of that literacy. Nowhere do you learn the limits of a system as quickly as when you assemble it yourself and see where it goes wrong.

What "building it yourself" does and does not mean

Let us clear up a misunderstanding right away: nobody has to become a developer. "Building it yourself" does not mean your whole team is going to write production-grade AI agents that keep the organisation running. It is about understanding, not production. The goal is for people to experience first-hand how an AI solution is put together, so they learn to work with it with realistic expectations and healthy trust.

In our build sessions, someone creates a working agent in an afternoon using low-threshold tools such as n8n. Think of a chat function that answers customer questions, a knowledge base you search in plain language, or a mail function that drafts tidy replies for you. They are not perfect end products, and that is exactly the point. The tangible result is not the agent, but the insight that comes with it.

One afternoon of tinkering teaches your team more than ten presentations about "the power of AI". Afterwards someone not only knows that AI can do something, but also roughly how, and above all where the limits lie. That insight is what sticks once the workshop is over, and it is precisely the foundation you need before you seriously start building something that does have to go into production.

Don't start with a purchase, start with an afternoon

The question is long past whether your team will work with AI. They already do, whether you see it or not. The real question is whether they understand it: whether they grasp what they are using, trust what it does, and know where the limits lie. Lasting adoption stands or falls with that.

So do not start with a quote for a new platform, but with an afternoon in which your team builds something themselves. Take an honest look as well at where AI is already being used unseen in your organisation; that tells you exactly where the need is greatest. At And AI we do not start with understanding on the work floor before we build anything for nothing. Not because it sounds nice, but because it is the only way AI keeps sticking instead of evaporating into a licence that no one opens anymore.

Frequently asked questions