The nods around the meeting table, and what they actually mean

There’s a certain kind of nodding that only happens in meetings. Someone says, "the model is hallucinating because the context window is too small, we should use RAG instead," and everyone around the table nods. No one asks. Asking would mean admitting you didn’t follow, and you don’t do that in front of colleagues.

We think that nodding is expensive. If you don’t dare ask what the words mean, you can’t ask the important follow-up questions either: what does it cost, what are we sharing, who’s responsible when things go wrong? AI decisions end up with the few who happen to know the jargon, not with those who understand the business. And those are usually the wrong people.

That’s why we built a glossary, and now we think it deserves its own introduction.

25 terms, explained so you can retell them in your own words

At the time of writing, the glossary contains 25 entries: from the basics like language model, prompt, and token, through technical terms like RAG, embedding, and context window, to legal concepts like the AI Act, Article 50, and third-country transfer.

Each entry follows the same rule as everything else we publish: the explanation should be something you can retell to a colleague without borrowing a single one of our words. A token isn’t "the model’s smallest semantic unit," it’s the chunks an AI splits text into, roughly one per short word, and everything is counted in tokens: what an answer costs and how much the model can keep in its head at once. A hallucination isn’t some mysterious technical glitch, it’s when the AI answers confidently but wrong, and the interesting part isn’t whether it happens (it does, in every AI service) but how you spot it when it does.

And every entry ends with the same question: what does this mean for a workplace? That’s where most glossaries stop and ours begins. Knowing what a hallucination is doesn’t help much. Knowing that the fixes are called sources, labelling, and review, that’s something you can do something with on Monday.

The term we most want you to look up

If we had to pick just one entry for you, it would be shadow AI. That’s AI use happening in an organisation without the people in charge knowing about it: the employee pasting the customer list into a free chatbot at home at the kitchen table, because their job hasn’t offered anything better.

Your colleagues are already using AI. The question isn’t if, but where, and what they’re pasting into it. Shadow AI doesn’t disappear with bans, it just moves deeper into the shadows. It disappears when there’s a sensible alternative in the light. That’s honestly why EuroWork exists.

In your language, even when your language is small

The whole glossary is available in 35 languages. That’s not a small point. Explaining AI terms in English to a Latvian nurse or a Portuguese carpenter just swaps one incomprehensible language for another. The nodding in meetings only stops when the explanation comes in the language you think in.

The glossary is open to everyone, no login, no paywall. Look up a term, send the link to the colleague who nodded next to you, use it in training. It’s always one click away at the bottom of every page, and it grows as the debate invents new terms.

Next time someone says, "we should use RAG," you can nod with a clear conscience. Or even better: ask, "do you mean the AI should answer from our own documents instead of its memory?" and watch the whole table turn toward you like you’re the one who actually understood.

You’ll find the full glossary here: EuroWork’s AI glossary.