How to use AI well.
Three steps: say exactly what you want, give the context it can't guess, and refine the first answer instead of accepting it. Most people skip steps two and three, which is why they think AI isn't very good. No coding needed — the skill is clear communication.
Step one: say what you actually want
The single biggest beginner mistake is typing a vague request and accepting whatever comes back. An AI assistant can't read your mind — it only has the words in front of it.
| Weak prompt | Strong prompt |
|---|---|
| "Write an email." | "Write a short, friendly email to my landlord asking to fix a broken heater. Polite but firm. Under 120 words." |
| "Explain quantum physics." | "Explain quantum physics to a 15-year-old using everyday analogies. Three short paragraphs." |
| "Help with my CV." | "Here's my CV and the job description. Which three bullet points should I rewrite to match the role better, and why?" |
Step two: give it context
Context is the information the assistant needs but doesn't have: who you are, who it's for, what "good" looks like, and any constraints. This matters more than clever wording — it's why the same question gets a generic answer for one person and a perfect one for another.
- Who it's for — "for a nervous first-time patient" changes everything.
- Your situation — your role, your constraints, your deadline.
- An example — paste something in the style you want. Examples beat descriptions.
- Format and length — "a bulleted list", "under 200 words", "a table".
Step three: iterate — don't accept the first answer
Treat the first response as a draft, not a verdict. The fastest improvements come from simply saying what's wrong: "Too formal — make it warmer." "Cut it in half." "You missed the deadline detail; add it." Most people never do this, which is why they conclude AI "isn't that good".
What to never do
- Never paste secrets — passwords, ID numbers, bank or card details, or confidential client data.
- Never trust facts blindly. Assistants can state wrong things confidently — that's called a hallucination. Verify anything that matters.
- Never use it for advice that needs a professional — medical, legal, or financial decisions need a qualified human.
- Don't paste and publish. Edit it into your own voice; unedited AI text reads like unedited AI text.
A first session to try right now
Open any assistant and run these three, in order. It takes about five minutes and teaches more than any article:
- 1. "Explain what you're good at and what you're bad at, in plain English."
- 2. "I want to [your real task]. What information do you need from me to do it well?" — then give it what it asks for.
- 3. "Now do it." Then critique the result and ask for one specific change.
That loop — ask, supply context, refine — is essentially all of practical AI use.