What AI actually is (and isn't)
A plain-English mental model for what happens when you type into a chat box.
AI is the most overused word in business right now. It gets attached to everything from your inbox to the future of human civilisation. Most of it is marketing, but some of it is real. Here's what AI actually is, in plain language, with no hype and no doom.
It predicts, it doesn't know
Most AI tools you know are a specific type of system called a large language model, or LLM. A large language model is, at its core, a very good prediction machine.
It was trained on enormous amounts of text from the internet and books, where it learned the patterns of how words follow each other. When you give it a prompt, it predicts what words should come next, one at a time, until it produces a response. That's the whole trick. It is not thinking. It is not understanding.
There are no opinions, beliefs, or experiences behind it. It's simply a very sophisticated pattern-matching system that produces text looking like a human wrote it.
What it is
A pattern machine trained on text.
A tool with real strengths and clear limits, like every tool you've used.
What it isn't
A search engine. It generates plausible text, it doesn't retrieve truth.
A person. No judgement, no stake in your project.
The reason it feels intelligent is that human language carries human intelligence inside it. When you train it with enough writing, it can reproduce the shape of our reasoning without ever doing the reasoning itself.
Understanding what it's not is half the job
It isn't a search engine. Search engines look up real information from real sources. AI tools generate plausible-sounding text based on patterns. When they produce facts, those facts may or may not be true. This is the source of "hallucination", the polite word for when AI makes things up confidently.
It isn't a person. It doesn't care about your project. It doesn't have judgement. The tool produces output. You provide the judgement.
AI covers a huge range of technologies, from the spam filter on your email, to image generators, to self-driving cars, to LLMs. When people say "AI is going to change everything," they're usually conflating very different systems. The thing changing your work right now is mostly LLMs. The other types of AI matter less to your day-to-day.
It isn't almost-AGI. Artificial General Intelligence means a machine that actually understands the world. It could move from one abstract or physical task to another, much like a person.
What this changes in practice
None of this makes AI useless. It makes it a particular kind of tool, good at some things and structurally bad at others. Three habits follow from it, and they're the difference between AI saving you an afternoon and AI costing you a client.
Three things to do this week
01
Check anything factual before it leaves your desk
Names, dates, figures, legal points. Especially the ones stated most confidently.
02
Give it context instead of expecting it to know yours
It has never seen your client, your brand or your deadline. Say so up front.
03
Use it where being 80% right fast is genuinely useful
First drafts, rewording, summarising, unsticking yourself. Not final client facts.
Try it now
Answer this, then list which parts you're confident about and which I should verify myself: …
Conclusion
01
AI predicts the next likely words. It doesn't look anything up.
02
Confident tone is not evidence. It sounds equally sure when wrong.
03
Treat it like a tool and use it for what it's good at. That puts you ahead of people who are either dismissing it or worshipping it.
Go to next