AI literacy, in plain English: what the Act is really asking of your people

From 2 August, organisations across Europe are expected to make sure their staff are "AI literate." It sounds like a training problem about software. It is not.

Picture an ordinary Tuesday afternoon.

It is a quarter to five. There are eleven other things waiting. A letter needs to go out to a client, so you asked the tool to draft it, and it did, in about four seconds. You read it. It is good. Better than good, honestly. It is clear, it is polite, it hits every point, and it reads exactly like something you would have written on a better day with more time.

So you send it.

One line in it was wrong. A real figure, a real scheme, put together in a sentence that was false. Not a wild invention you would have spotted a mile off. A tidy, confident, plausible mistake, sitting in a letter that read beautifully. And it went out under your name, because a person has to put their name to it, and that person was you.

That afternoon, stripped of every bit of jargon, is what "AI literacy" is really about.

The thing being sold, and the thing that matters

Between now and 2 August, a lot of training is going to be sold with "AI literacy" on the label. Most of it will be about the tools. How to write a good prompt. Which buttons to press. Which app does what. Some of that is genuinely useful, and none of it is the thing that gets people into trouble.

Here is the thing that gets people into trouble.

Using AI is easy. Being answerable for what it produces is the same job it always was. The tool changed. The responsibility did not move an inch. A solicitor is still responsible for the letter. An accountant is still responsible for the figures. A board is still responsible for the paper it signs off. None of that changes because the first draft came out of a machine instead of a junior.

Everyone already knows this in their bones. Nobody has ever accepted "the computer wrote it" as an excuse, and nobody is about to start. Courts have already sanctioned lawyers for filing documents that cited cases the AI simply made up. The people who filed them did not invent the cases. They just did not check, and the machine's version looked real enough to sign.

A machine can write the letter. It cannot be the one who signs it.

Why it is harder than it sounds

If AI made obvious mistakes, none of this would matter. You would read the draft, see the nonsense, and fix it, the way you always have.

But that is not how these tools fail. They do not fail loudly. They fail with confidence. The invented case that reads like a real one. The true number attached to the wrong thing. The quote that sounds exactly right and was never actually said. The danger was never the rubbish you would catch in a second. It is the polish. A document that is ninety-nine per cent right and reads perfectly is far more dangerous than one that is obviously wrong, because it sails straight past a busy person at a quarter to five.

So being "AI literate" is not really about the machine at all. It is about the habit you bring to what the machine gives you.

What "AI literate" actually means

Put plainly, without a single technical word, it comes down to four things.

Knowing that whatever the tool produces, your name is the one that ends up on it.

Having a simple habit for checking the things that actually matter before anything goes out, so the confident mistake gets caught while it is still yours to catch.

Knowing the one line you never cross, which is putting client, patient or confidential information into a public tool, because once it is in, you do not get it back.

And being willing to say you used AI when it matters, and willing not to use it when the work deserves better than a machine.

That is the whole of it. Notice what is not on the list. Nothing about prompts. Nothing about features. Nothing you need a software course for.

The good news

None of this is a new skill you have to go out and buy.

It is the oldest professional habit there is, do not put your name to something you have not checked, meeting a new tool that happens to be very good at looking finished. Your best people already do this without thinking when the work is human. The whole job now is to make the same instinct kick in when the work came from a machine, precisely because the machine's version arrives looking so ready to send.

That is a smaller job than a compliance panic makes it sound. But it is a real one, and it is the one the Act is actually pointing at. When the law says make your people AI literate, this is what it means underneath the wording: make sure the people using these tools still carry the responsibility for what comes out of them, and still know how to check.

So if you are buying training for this

The test is simple. Does it teach your people to use the software, or does it build the judgement to catch the confident mistake before it goes out under their name? One is a tutorial. The other is the thing that keeps a professional, and their organisation, out of trouble.

This is exactly what we built our course around. It does not teach anyone how to use a chatbot. It puts a professional through that ordinary afternoon, the one where the wrong thing looks right and there is no time, and builds the habit of catching it.

AI for Professional Practice is a modular CPD programme for regulated professionals: solicitors, accountants, insurance and financial services professionals, company directors. Module 1 opens free on 30 July 2026. It is deliberately not a course about AI. It is about the duties professionals already carry, accountability, verification, confidentiality and judgement, and what AI changes about discharging them.

Because the machine can write the letter. It just cannot be the one who signs it.

For the enforcement timeline itself, what actually arrives on 2 August and what was quietly deferred, the companion piece is what actually happens on 2 August 2026. For the literacy obligation and its date, see the AI Act deadline most organisations are not ready for. See more insights from LearnFrame.