AI training for your team: why it is now mandatory

On 2 August the European Union's AI Act entered general application. If you run a company in Panama, your first reaction is the right one: it does not apply to you.

But one part of that law is worth reading even from nine thousand kilometres away, because it is not a bureaucratic requirement. It is the description of a problem you already have.

What changed, concretely

Since that day, in Europe:

• If someone is talking to a bot, they have to be told. The user must know they are interacting with a machine, unless it is obvious.
• AI-generated content gets marked. Synthetic images, video, audio and text carry a machine-readable mark.
• Deepfakes must be identified when they could make someone believe they are authentic.
• Companies deploying AI must build competence in their people. Not just whoever builds it: whoever uses it.

Fines reach 15 million euros or 3% of worldwide turnover, whichever is higher. And the rules for high-risk systems were pushed to December 2027 and August 2028, which tells you something on its own about how hard this is proving to apply.

The part that does affect you today

Of those four, the interesting one is not the fines. It is the last one: the obligation that your people actually know how to use the tool.

Look at what that rule is saying once you strip the legal language. It is acknowledging that most companies already bought AI and their people do not know how to use it. So much so that it had to be written into law.

That problem has no nationality. I see it here every week.

The pattern I see in Panama

The conversation repeats itself. A company pays for licences of some AI tool for the whole team. Three months pass. When I check who actually uses it, it is two people, and they use it to draft emails.

The cost is there. The result is not. And the conclusion they draw is almost always the wrong one: "AI is not useful for what we do".

It is not that it is useless. It is that nobody showed them what to ask it for. It is exactly what happens with an ERP that cost a fortune and gets used as a spreadsheet.

A study from this year puts it in numbers: 54% of organisations already report real benefits from their AI investment. That figure has two readings and the second one matters — almost half report none. And the brake they identify is not the technology: it is the talent gap.

What I would do before buying more licences

Four things, in this order:

1. Look at who is actually using it. Almost every tool gives you that data. If you have twenty licences and six active users, you already know where the problem is, and it is not the vendor.
2. Find the three most repetitive tasks of the month. Not the most important ones: the most repetitive. That is where automation pays quickly and visibly.
3. Teach with their own work, not generic examples. A workshop where each person brings their own boring task and leaves with it solved beats ten recorded courses.
4. Write down how to ask. The difference between a useless result and a good one is almost always how specific the request was. That is taught in an afternoon and documented once.

And about the law, to close

Panama does not have a regulation like this yet, but it is worth not getting comfortable, for two reasons.

The first is commercial: if you sell to clients in Europe, their obligations reach you through contracts long before a local law exists. It is already happening.

The second is common sense. Telling people there is a bot on the other side, marking what you generated with AI, and making sure your team knows what it is doing are all reasonable things. If you wait until they are mandatory, you will be improvising against a deadline.

If you want us to review how your team is using the tools you already pay for, write me. It is a short conversation and the problem usually shows up in the first half hour.