Good AI training gives your team a foundation. Then everyday work takes over: rewriting a customer email, summarising a case file, preparing a report. That is exactly where a reminder of what you learned helps, for instance when an account number is still in the text you are about to share.
Take a customer service agent answering a difficult complaint on a Thursday afternoon. She pastes the customer's email into ChatGPT and asks for a short, friendly reply. She remembers Monday's training perfectly well. The customer's account number just happens to sit at the bottom of the forwarded email, below the signature, and that is not where she is looking. She is thinking about the tone of her reply.
That is not a lack of knowledge or care. A long text with one sensitive detail tucked inside is simply easy to miss. The question then is not whether the training worked, but how to bring that knowledge to bear at this moment.
What does good AI training give your team?
A shared language, to start with. After a good training session, everyone knows what the organisation means by sensitive data, which AI tools are allowed and why a free account differs from a business one. People know where to look up the rules and who to ask.
Training is also the place to show what AI does well. A letter in plain language, a summary of a long report, a first outline for a presentation. People who know what is possible use AI with more confidence and ask better questions about what is not.
And for many organisations it is simply required: since February 2025 the AI literacy obligation in the EU AI Act has asked for measures that fit the role and context of whoever uses AI. Training is a natural way to meet it.
Where does knowing turn into doing?
In the middle of the work. Training deals in principles; work deals in one particular text with a deadline attached. Between the two may lie a few days or a few months, and in that time new tools, new clients and new kinds of files arrive.
Research points the same way. Newcom's AI Monitor (July 2026) found that about a third of Dutch people who use AI at work remove sensitive information before using an AI tool. That need not mean the rest do not know better. Mostly it shows how hard it is to check a text again at exactly the right moment. Training and support while people work reinforce each other.
Which moments at work deserve attention?
The ones where someone starts from existing text with real data in it. They look different in every line of work.
In healthcare, a nurse asks an AI assistant to make a report easier to read, and a patient's name is still in the first sentence. In support teams, a whole customer email goes into an AI tool, complete with name, order number and phone number. In HR, an offer letter gets rewritten with the salary still in it. And in financial services, an IBAN sits in the transaction someone wants summarised.
These are all good reasons to use AI. In each of them, one detail is easy to overlook.
To help at those moments, you first need to know which tools the work is done in. Many organisations have rolled out one AI tool and cannot see that others are being used alongside it.
Gmail
