Training your team in AI: what to teach, in what order and what is not needed
Most AI training courses teach people to write pretty prompts and change nothing the next day. What a small team really has to learn so that Monday's work is different.
· Applied AI training
Almost every artificial intelligence course sold to small businesses has the same problem: it teaches the tool and not the work. People come out with a list of tricks, apply them for two weeks and then go back to doing things the way they always did, because nobody showed them how to fit those tricks into their actual tasks.
This is what does work, in the order in which it works.
First: where it is NOT used
It seems counter-intuitive to start there, but it is what prevents the accident. Before teaching anybody to use the tool you have to make clear what information never leaves the company: client documentation with names and tax numbers, health data, payslips, credentials.
And the rule that solves nearly everything: strip the identifiers before pasting anything. A contract without the name and the document number summarises just as well, and it is no longer personal data travelling to a third party. The full legal side is in using AI with customer data within the law.
Second: what it can do and what it pretends to do
The idea that prevents the most mistakes, and the one hardly ever taught: these systems generate plausible text, not verified truth. They write beautifully and invent a legal reference with exactly the same confidence they use to cite a real one.
Two rules follow that a team has to take on board:
- It is useful for whatever you can verify. If you know enough to spot the mistake, it is a great help. If you do not, it is a well-written risk.
- Anything with legal, tax or accounting effect is checked at source. No exceptions.
Third: their own tasks, not generic examples
This is where a training session earns its money or loses it. Instead of «how to write prompts», each person brings three real tasks they do every week and solves them during the session.
What usually comes up in a small company:
- Summarising a long document and pulling out the points that require action.
- Drafting the first version of a reply that comes up again and again.
- Translating or adapting a text into a different register.
- Turning a messy list into a usable table.
- Explaining an error message from a program.
None of them is spectacular. All of them save between twenty minutes and two hours a week, every week, and that is what keeps people using it in March.
Fourth: when to stop using the chat
The signal is simple: if you do a task the same way, with the same tool and more than once a week, it is no longer a chat case, it is an automation case. Copying and pasting fifty times a month what an automatic flow would do on its own is work paid for in hours.
Being able to tell those two situations apart is probably the most profitable thing in the whole course. What lies on the other side we cover in AI and automation.
What does not need teaching
We are not popular for saying this: a small-business team does not need to know what a transformer is, nor the difference between models, nor advanced prompt engineering. That is filler content to stretch a course to eight hours.
With an understanding of the limits, respect for the data rule and their own tasks solved live, a team is operational. The rest is learned by using it.
About the legal duty to train
It is being used as a sales argument and it is worth being precise. Art. 4 of the European AI regulation does ask for literacy measures to be adopted — but it is not among the provisions that carry penalties under art. 99.4. Anybody selling you training on the grounds that you will be fined if you do not buy it is selling with an argument that does not hold up. The detail, with the new dates from the July 2026 Omnibus, is in what the AI regulation requires today.
Training the team pays off for what it saves, not for the fines it avoids.
If this sounds like you
We run the training with your team's real tasks on the table, not with textbook examples: AI training for your team. And if you also want the data and contracts side tidied up, it goes together with AI and data protection compliance.
We are Mindset & Code: automation, data and development for small businesses. We use these tools every day on real work with legal obligations attached, which is different from describing them on a slide. You can see what we do and what it costs.
Sources: Regulation (EU) 2024/1689 on artificial intelligence, as amended by Regulation (EU) 2026/1744; Regulation (EU) 2016/679 (GDPR) and Chile's Ley 21.719 for the data side. General guidance.