What AI is, in plain language
Listen instead (audio briefing)
Cut through the jargon: the dozen words you'll hear most, and what they mean for your work.
AI is a family of tools, not one thing
“Artificial intelligence” covers any computer system that does tasks we used to think needed human judgement: recognising a face, predicting demand, translating a sentence, writing a reply. Most of it has been around you for years, in spam filters, maps and bank fraud alerts.
What’s new since 2022 is generative AI: tools such as ChatGPT, Claude, Gemini and Copilot that produce new text, images, audio and code from a plain-language request. It sits inside a bigger family, as the diagram shows.

Generative AI is the newest member of the family, not the whole of it.
The words you’ll hear most
Model: the trained system that does the work. Large language model (LLM): a model trained on huge amounts of text to predict and produce language. Prompt: what you ask. Token: a small piece of text the model reads and writes; usage limits and prices are counted in tokens.
Hallucination: a confident answer that’s wrong or invented. Agent: AI that can take steps on its own, such as searching, filling in a form or sending a message, within limits you set. Training data: the examples a model learned from, which decide what it knows and what it gets wrong.
Tip: If someone uses an AI word you don’t know, ask the AI itself: “Explain [word] in two sentences for someone in [your job]. Give one example from my work.”
What it’s good at, and what it isn’t
Generative AI is very good with language: drafting, summarising, explaining, rewording, organising messy notes. It’s fast, patient and available at 2am.
It doesn’t “know” things the way a person does. It predicts likely words, so it can be wrong while sounding certain, it may not know recent events, and it doesn’t know your business unless you tell it. Used well, with your judgement on top, it saves time every day.
Worked example: asking it to explain
Lindiwe works in accounts payable. Her manager mentioned “an AI agent for invoices” and she wasn’t sure what that meant.
Worked example: Making sense of a new term
Before: Prompt
What is an AI agent?
What came back
An AI agent is an autonomous software entity that perceives its environment, reasons about goals and takes actions using tools and APIs to accomplish complex multi-step tasks…
Accurate, but it doesn’t help her picture it.
After: Prompt
Explain what an “AI agent for invoices” would do, in three short sentences, for someone who works in accounts payable. Give one example from a normal day. Say what a person would still need to check.
What came back
It would read incoming supplier invoices, pull out the supplier, amount and due date, and match them to the purchase order. For example, it might flag that an invoice from your stationery supplier is R2,400 more than the order. You’d still approve every payment and check anything it flags.
Plain, specific to her job, and clear about who stays in charge.
Same tool, same topic. The difference was telling it who she is and what she needed.
Key points
- AI is a family of tools; generative AI creates new content.
- Model, LLM, prompt, token: the basic vocabulary.
- Hallucinations are confident mistakes.
- Strong with language; your judgement stays on top.
Common questions
Is AI going to take my job?
It changes jobs more than it replaces them. People who learn to use it well tend to become more valuable.
Do I need to understand the technology deeply?
No. Knowing the basic words and limits is enough to use it safely and well.
Exercise: Spot the AI in your week
Your version: pick the line for your area of work
- Operations: writing a delay notice to a customer
- Finance: explaining a cost increase to a manager
- Sales and marketing: drafting a social media post
- HR and people: answering a staff question about leave
- Customer service: replying to a common customer question
- Admin and coordination: summarising a long email thread
- Training, L&D or skills development: preparing a short training handout
- Business owner or founder: writing a quote for a new customer
- Project or programme management: writing a weekly project update
- Executive team: preparing a short update for your team
- Other (any role): a piece of writing you do every week
Ten to fifteen minutes. You’ll notice where AI already helps you, and try it once on something from your own work.
Step 1: List five places AI already touches your week
Think about email, your phone, maps, banking and work systems. Write down five, with what each one does for you.
Example: Spam filter (keeps junk out of my inbox), maps (traffic and arrival time), bank app (fraud alerts), phone camera (searches my photos), Outlook (suggested replies).
Step 2: Ask it to explain one word
Pick one word from this module you weren’t sure of. Ask an approved AI tool to explain it for your job.
Prompt to copy:
Explain “[word]” in two sentences for someone who works as a your area in your industry. Give one example from a normal working day.
Step 3: Imagine one task with AI
Take your task (see “Your version” above). Write one sentence on what AI could do for you, and one on what you’d still check yourself.
- ☐ One sentence on what AI could do
- ☐ One sentence on what you’d check
Step 4: Explain it to someone
Tell a colleague or friend, in your own words, what generative AI is and one thing it’s not good at. If you can explain it, you’ve got it.
Hand in: Note your five places and your two sentences here.
You’re done when: You can name where AI already helps you, explain the key words in your own way, and say what you’d still check.
