Courses › AI Foundations for Everyday Work › Module 1: Understanding AI › Lesson 1.3
What you will be able to do
Decide whether a task suits AI, an ordinary tool, or your own judgement, and name the checks to make before you use anything an AI gives you.
What AI does well
A generative assistant is strongest when the job is working with words and ideas and you can judge the result yourself. In everyday work, that means:
Useful for
- Drafting a message, note or description from your facts
- Rewriting something in a different tone or at a simpler reading level
- Summarising a text that you give it
- Explaining an idea in plain words, with an example
- Listing options, questions or ideas to think about
- Turning rough notes into a list, table or template
Check carefully
- Facts, dates, names and quotations
- Sources, links and references
- Prices, measurements and any calculation
- Anything current: opening hours, rules, news
- Descriptions of people or groups
- Advice on health, law, money or safety
Some examples from different kinds of work: a baker asks for three ways to word a notice about holiday opening; a student pastes lecture notes and asks for plain-language explanations of the hard parts; a freelancer turns a call transcript into a list of what the client asked for; a teacher asks for the same explanation at two reading levels. In each case the person already knows what good looks like.
Where it goes wrong
These are the errors that matter most, with what to do about each.
| What can happen | What it looks like | What you do |
|---|---|---|
| It makes things up | A book title, a quotation, a web link or a “fact” that does not exist, written in a confident voice | Open every source it gives. If you cannot find it, do not use it. |
| It is out of date | Old rules, old prices, old opening hours; it may not know recent events | Use the official website or ask the organisation directly. |
| It lacks your context | A generic answer that ignores your customers, your location or your own rules | Add the context yourself, and cut what does not apply. |
| It slips on sums | A total that is slightly off, or a measurement or percentage that does not add up | Do the calculation with a calculator or spreadsheet. Treat any AI figure as unchecked. |
| It reflects bias | Stereotypes in how it describes people, jobs or places; assumptions about who is “typical” | Read descriptions of people critically and rewrite anything unfair. |
| It sounds sure when it is wrong | A calm, polished answer that is mistaken | Do not use tone as a guide to accuracy. Check what matters. |
| It answers differently each time | Two runs of the same request give different wording, sometimes different content | Compare the versions, and keep what you have verified. |
Privacy: what to keep out
With an online assistant, what you type is usually sent to the company that runs it and processed on its systems. Assistants that run entirely on your own device can work differently, but most beginners use online ones. Whether your text is stored, how long it is kept, whether it may be used to improve the service, and what controls you have all vary by service and plan, and they can change. So decide before you paste. Read the provider’s official privacy information and check your account settings before you share anything beyond made-up or harmless text.
Do not put these into a general AI assistant:
- Passwords, security codes or login details
- Bank, card or payment details; identity or passport numbers
- Other people’s private details: customers’ addresses, phone numbers, health or family information, pupils’ records
- Confidential documents, contracts or anything covered by an agreement or workplace rule
- Private messages that belong to someone else
A safer habit: replace real names and numbers with placeholders such as [customer] and [order number], and put the real ones back yourself afterwards. If your school or employer has a rule about AI tools, follow it. Module 6 returns to this in detail.
You stay responsible
If you send a message, publish a post or hand in work, it is yours, whoever or whatever drafted it. A wrong date in a customer message is your wrong date. Decisions with real consequences, such as health, safety, legal duties, money owed or someone’s marks and records, need a qualified person and your own judgement. AI can help you prepare. It cannot take responsibility.
When an ordinary tool is better
Using AI is not always the smartest choice. Often a simpler tool is faster, cheaper and more reliable.
| Task | Better choice | Why |
|---|---|---|
| Adding up a quote or an order | Calculator or spreadsheet | It follows exact rules, so the total is the same every time and you can check the formula. |
| Today’s opening hours, price or rule | The official website, or a phone call | It is the source. A model’s memory may be old. |
| A reply you send every week | A saved template or a snippet in your messaging app | You write it once and it never changes by accident. |
| Finding a file, contact or email | The search in your phone or mailbox | It looks at your own data and finds the exact item. |
| A measurement, a safety check, a repair decision | Measuring tools, the manufacturer’s instructions, a qualified person | The result is physical or safety-related and cannot rest on a draft. |
| A draft from rough notes, a new wording, a first outline | An AI assistant | This is where it helps: you can judge and fix the result yourself. |
A quick way to decide
Before you use AI on a task, ask three questions:
- Is it mostly words, ideas or organising? If it is arithmetic, a measurement or finding an exact record, use the ordinary tool.
- Could a mistake hurt someone or cost money? If yes, a person with the right knowledge decides, and AI at most helps you prepare.
- Can I check the result myself? If you could not tell a wrong answer from a right one, do not use AI for it yet.
If the answers are “yes, no, yes”, AI is a good fit. Otherwise use an ordinary tool, a person, or your own judgement.
Common mistakes
- Using AI as a search engine for facts. It may produce plausible text instead of a verified answer.
- Trusting a source because it has a title and a link. Open it.
- Letting it do the arithmetic. Use a spreadsheet for money.
- Pasting private details “just this once”. Use placeholders.
- Giving up after one bad answer. Add context or a limit and ask again. Then judge whether the task suits AI at all.
Try it yourself: sort your tasks, then test one limit
Part 1: sort three tasks (about 10 minutes, no assistant needed). In the worksheet, Part C, list three tasks you do in your work, studies or home life. For each, answer the three questions above and write one of: AI could draft it, an ordinary tool is better, or needs a person’s judgement. Add one sentence saying why.
Part 2: test one limit (optional, about 5 minutes, needs an assistant). Choose one:
- Source test. Ask: “Give me three sources, with links, about [a topic you know well].” Open each link. Does the page exist? Does it say what the assistant claims? Write down what you found.
- Local-knowledge test. Ask a question about something only you know, such as your own street, shop, class or team. Notice what it does when it does not know.
The results differ between assistants, so write down what yours did. That is the point.
You are done when you have three tasks sorted with reasons, and, if you did Part 2, a note of what you found.
AI is good at drafting, explaining and organising words you can judge. It can invent, be out of date, miss your context, slip on numbers and mishandle private information. Check what matters, keep private details out, and use an ordinary tool or a person when the task calls for it.
Next: the Module 1 practical activity brings the three lessons together in one real task.

