Lesson 1 of 8 · 7 min read
Most explanations of AI fail in the same way. They either drown you in maths you didn’t ask for, or they wave their hands and say it’s “like a digital brain,” which sounds reassuring and tells you nothing.
Neither helps you use the thing.
So here’s the version that actually matters for your job: AI is software that finds patterns in data and uses them to make predictions. That’s it. Everything else — the chatbots, the image generators, the tools your company just bought a licence for — is built on that one idea.
By the end of this lesson you’ll understand what AI genuinely does, what it can’t do, and why the difference matters every time you use it.
What AI actually means
The term “artificial intelligence” has been around since 1956, and it has meant something different in almost every decade since. That’s part of why it’s confusing — people use the same two words for a chess program from 1997 and for ChatGPT.
A more useful definition: AI is software that learns from examples rather than following instructions written by a person.
Think about the difference. Traditional software does exactly what a programmer told it to do. A payroll system calculates tax because someone wrote the rule: if salary is between X and Y, deduct Z. Every outcome traces back to a human decision.
AI doesn’t work that way. Nobody wrote a rule telling ChatGPT how to answer your email. Instead, it was shown an enormous amount of text and worked out the patterns for itself. When you ask it something, it’s not looking up an answer — it’s predicting what a good answer would look like, based on everything it has seen.
That distinction explains almost every strange thing AI does. Why it’s brilliant at some tasks and useless at others. Why it sometimes invents facts with total confidence. Why two people asking the same question get different answers.
It isn’t retrieving. It’s predicting.
AI, machine learning, and generative AI
Three terms get used interchangeably. They’re not the same thing, and knowing which is which will save you a lot of confusion.
Artificial intelligence is the umbrella term. Any software that performs tasks we’d normally consider to need human intelligence.
Machine learning is the method underneath most modern AI. It’s the process of feeding a system examples until it works out the pattern. Show it ten thousand photos labelled “cat” and ten thousand labelled “dog”, and it learns to tell them apart — without anyone ever defining what a cat looks like.
Generative AI is the newest branch, and it’s what caused the noise from 2022 onwards. Rather than sorting things into categories, it produces new things: text, images, code, audio. ChatGPT, Copilot, Gemini, Claude, Midjourney — all generative.
The nesting matters: all generative AI is machine learning, and all machine learning is AI. Not the reverse.
When someone at work says “we should use AI for this,” they almost always mean generative AI. And in this course, so do we.
What today’s AI tools actually do
Strip away the marketing and modern AI tools do a small number of things extremely well:
- Turn one form of text into another. Long into short. Bullet points into prose. English into Spanish. Notes into an email.
- Answer questions from patterns in what they’ve read. Not from a database — from what they’ve absorbed.
- Draft first versions. Of anything. Emails, plans, code, job descriptions.
- Spot patterns in information you give them. Themes across fifty customer reviews, inconsistencies in a document.
- Explain things at whatever level you ask for. The same concept for a CEO or a ten-year-old.
Notice what these have in common. Every one is a language task. Modern AI is, at its core, extraordinarily good at understanding and producing language. Its usefulness at your job is mostly a function of how much of your job is made of words.
For most office work, that’s a lot of it.
What AI is not
This is the part that gets skipped, and it’s the part that causes the problems.
AI doesn’t understand anything. It has no model of the world, no beliefs, no sense of whether what it just said is true. It produces text that looks like the kind of text that would be correct. Often that’s the same thing. Sometimes it very much isn’t.
AI doesn’t know what it doesn’t know. A person who isn’t sure will hedge. AI will state an invented fact in exactly the same confident tone it uses for a real one. This is called hallucination, and it’s not a bug that will be patched — it’s a direct consequence of how prediction works. We’ll cover this properly in Lesson 8.
AI isn’t connected to the truth. Unless a tool is explicitly searching the web, it’s working from what it absorbed during training, which has a cut-off date. Ask about last week’s news and you may get a confident answer about a world that no longer exists.
AI isn’t intelligent in the way the word implies. It can draft a decent legal summary and then fail at counting the letters in a word. Its abilities are jagged — brilliant and useless in places you wouldn’t predict. Don’t assume that because it did something impressive, it can do something that seems easier.
None of this makes AI less useful. It makes it useful in a specific way — as a fast, tireless, occasionally wrong assistant that needs checking. That’s a genuinely valuable thing to have. It’s just not what the marketing says it is.
Where you’re already using AI
AI stopped being a future technology a while ago. You’ve been using it for years:
- Your email filters spam
- Your phone recognises faces in photos
- Netflix and Spotify recommend things
- Your bank flags an unusual transaction
- Maps reroutes you around traffic
- Predictive text finishes your sentences
None of these felt like a revolution because they arrived quietly, inside things you already used. Generative AI felt different only because it showed up as a blank box asking you to type.
The point: you’re not starting from zero. You’ve been living with AI for a decade. What’s new isn’t the technology — it’s that you can now talk to it directly, which means the skill of knowing what to ask suddenly matters.
That’s what most of this course is about.
What this means for your work
If AI is pattern-matching software that’s exceptional at language and unreliable about facts, the practical implications follow:
It’s excellent for first drafts, poor for final answers. Use it to get from nothing to something. Then apply your own judgement.
It’s only as good as what you give it. Vague question, vague answer. This is why prompting is a skill and not a gimmick — Lessons 4 and 5 cover it.
Verify anything that matters. Names, numbers, dates, quotes, legal or medical claims. If being wrong has a cost, check it.
It doesn’t replace expertise — it rewards it. You need to know enough to spot when the output is wrong. People with deep knowledge get more out of AI than people without, not less.
That last one is worth sitting with. The people who benefit most from AI aren’t the ones who know the most about AI. They’re the ones who know the most about their own work.
Key takeaways
- AI is software that learns patterns from examples, rather than following rules a person wrote
- Generative AI — the kind you’ll use daily — sits inside machine learning, which sits inside AI
- It predicts what a good answer looks like; it doesn’t retrieve facts or understand meaning
- It’s confident whether it’s right or wrong, so anything that matters needs checking
- Its strength is language, which is why it applies to so much office work
- Knowing your own field is what makes AI useful to you
Next: Lesson 2 – How AI Models Learn
Now you know what AI is. Next we’ll look at how it actually learns — still no maths — because understanding that is what makes its odd behaviour predictable rather than mysterious.