There’s a particular kind of frustration that comes from sitting in a meeting, or reading an article, and nodding along at words you don’t actually understand. AI. LLM. Prompting. Hallucination. Everyone else seems to know what they mean. You make a note to look them up later, and then you don’t, because when you do, the explanations are either too technical or too breathless — and neither one actually helps.
This page exists for that moment. It’s not a dictionary. It’s not a course. It’s the explanation a knowledgeable colleague would give you over coffee — unhurried, plain, and honest about what actually matters.
What “AI” Actually Means (In This Context)
Artificial intelligence is one of those terms that has been stretched so far it means almost everything and almost nothing. For our purposes — and for most of the conversations you’re likely to have — AI refers to software that can do things we used to assume only humans could do: understand language, generate text, answer questions, summarise documents, write emails.
The AI tools most people are using right now don’t think the way humans think. They don’t have opinions, ambitions, or feelings. What they do have is an extraordinary ability to recognise patterns in language — and to produce responses that feel, often remarkably, like something a thoughtful person might write.
That’s useful to know. It won’t always behave like a person, even when it sounds like one.
Generative AI
Generative AI is the specific branch of AI that creates things — text, images, audio, code. When someone says “I used AI to write a first draft” or “AI summarised that report for me,” they’re talking about generative AI.
The tools most professionals are using — ChatGPT, Claude, Gemini, Copilot — are all generative AI. The word “generative” just means it generates output rather than, say, detecting spam or recognising faces in photos (which are also AI, but different kinds).
LLM — Large Language Model
This is the technical name for the engine behind most of the AI tools you’ve heard of. An LLM is a type of AI trained on vast amounts of text — books, websites, articles, code — until it becomes very good at predicting what words should follow other words.
The “large” refers to the scale of the training, not to anything you’d see or feel. You don’t need to understand how it works any more than you need to understand combustion engines to drive a car. What’s worth knowing is this: LLMs are very good at language tasks, and surprisingly good at reasoning tasks, but they have real limitations — including one called hallucination, which we’ll get to.
ChatGPT is built on an LLM. So is Claude. So is almost everything else in this space.
Prompting
A prompt is the instruction you give to an AI tool. It’s what you type into the box.
“Summarise this email thread.” That’s a prompt. “Write a first draft of a performance review for someone who exceeded their targets but struggles with communication.” That’s also a prompt — a better one.
Prompting is sometimes talked about as though it’s a technical skill requiring special training. It isn’t. It’s closer to learning how to give clear instructions to a capable colleague who is new to your organisation. The clearer and more specific you are, the better the result. You already know how to do this — you’ve been doing it with people for years.
Hallucination
This is the word used when an AI tool produces something that sounds completely confident and turns out to be wrong. It might cite a source that doesn’t exist. It might state a fact that isn’t true. It might give you a name, a date, or a figure that it has, essentially, made up.
It’s called hallucination because the AI isn’t lying — it doesn’t know it’s wrong. It’s producing text that fits the pattern of a plausible answer, even when the underlying information isn’t there.
The practical implication: always verify anything specific. Dates, statistics, names, quotes, legal or medical information — check them. AI is a remarkable first-draft tool. It’s not a reliable final source.
ChatGPT and Claude — What’s the Difference?
Both are AI assistants built on large language models. ChatGPT is made by OpenAI. Claude is made by Anthropic. They do broadly similar things — answer questions, write, summarise, reason through problems — and both are worth knowing.
The differences between them are real but subtle, and they change with every new version. The more useful question isn’t which one is better, but which one you find easier to work with. Start with one. You can always try the other later.
Context Window
When you’re having a conversation with an AI tool, it can only “remember” a certain amount of the conversation at once. That limit is called the context window.
In practical terms: if you’re working on a long document, or a conversation that’s been going for a while, the AI may lose track of things you said earlier. It’s not being difficult — it simply can’t see back that far. Starting a fresh conversation, or re-pasting the relevant information, usually solves it.
Agentic AI
This is a newer term, and you’ll hear it more. Agentic AI refers to AI that doesn’t just respond to a single question — it takes a sequence of actions to complete a more complex task. Instead of answering “what’s in this document?”, an agentic AI might read the document, search for related information, draft a summary, and send it to the right person.
We’re still in the early stages of this. Most professionals don’t need to think about it yet. But it’s worth knowing the word, because it’s increasingly what people mean when they say AI is going to change how we work.
None of these terms require a computer science degree. They require about fifteen minutes and someone willing to explain them plainly — which is what this page is for.
If something here raised a question rather than answering one, that’s a good sign. It means you’re thinking about it properly.
All AI terms are also collected in the Glossary →