Using AI well, day to day
Write a good prompt, pick the right tool, check the answers, and protect your data.
本页目录
1. Writing a good prompt
The five-element method
The role — Tell the AI who it should be: "You are a career counselor…".
The context — Give useful information: your situation, your audience, your goal.
The task — Clearly state what you want, with an action verb: "Summarize", "Compare", "Suggest three ideas".
The format — Specify the expected shape: a table, a 5-point list, 150 words, formal or casual tone.
Constraints and examples — State what to avoid, or give an example of the result you want.
Bad prompt: "Tell me about sleep." Good prompt: "You are a sleep doctor. I'm a student who sleeps badly before exams. Give me 5 concrete tips, one sentence each, suited to exam revision periods."
In the good prompt example ("You are a sleep doctor…"), which element matches "Give me 5 concrete tips, one sentence each"?
In "You are a sleep doctor", the AI is given a … instruction.
An AI's first answer should always be treated as final: there's no point rephrasing your request.
2. Choosing the right tool for the task
Which type of tool for which need?
| I want to… | Type of tool |
|---|---|
| Write, rephrase, summarize | General-purpose assistant |
| Look up information with sources | Answer engine |
| Understand documents I provide | Document assistant |
| Create images | Image generator |
| Create a voiceover | Voice synthesis |
| Write code | Coding assistant |
You want to look up recent information and get the sources used. What type of tool should you pick?
Sort each need into the right tool family
What should you always check before picking an AI tool based on its price?
3. Checking the answers
Why check?
An AI can hallucinate: produce a false answer stated with confidence. Building good verification habits is essential.
5 good habits
Ask for sources, then open them — The AI can invent links or references.
Cross-check numbers, dates and names — Against a reliable, independent source.
Be wary in sensitive areas — Health, law, finance: AI helps you understand, it doesn't replace a professional.
Ask the same question differently — If the answers diverge, that's a warning sign.
Specify the date for recent info — A model without web search doesn't know current events.
Sort each action: good verification habit, or not?
For a medical topic, an AI answer can help you understand, but doesn't replace a health professional's advice.
4. Protecting your data
Stay accountable
Don't share passwords, credit card numbers, medical data or confidential documents. Check in the settings whether your conversations are used to train models. Disclose that you used AI when expected. Always proofread before sending or publishing: you remain responsible for what you share.
It's fine to type a credit card number into an AI assistant to help you with your budget.
Who remains responsible for AI-generated content that you publish?
An AI tool's privacy settings sometimes let you opt out of having your conversations used to train models.
Boss challenge: good habit or bad habit?
Sort each behavior according to whether it's a good or bad habit when using AI.