Free Guide

Your First Week With AI

You are looking at a blank text box for the first time, and it is not obvious what to do with it. This guide walks you through it in order: what this actually is, what to type first, how to ask for things properly, and how to check whether you can trust what comes back. No sign-up, no jargon.

0 of 8 done

The Checklist

Work through it in order

Tick each one off as you go. It's fine to come back later, the tracker just runs in your browser for this visit.

1 What is this, actually?

An AI chat is a computer program you talk to by typing. You type a message, it types a reply. That is the whole idea. It is not a search engine handing you a list of links, and it is not a person on the other end.

It has read an enormous amount of writing, so it is very good at producing sensible-sounding answers. It does not always know whether what it wrote is actually true. It can be completely wrong while sounding completely sure of itself. Keep that one sentence in mind for every step below.

One more thing before you start: you do not need a perfect first message. You need to send one. Every step after this one will help you write better prompts, but pressing send on something rough is what actually gets you moving.

2 Find the box, say hello

Somewhere on the screen, usually near the bottom, there is a text box waiting for you. Click into it, type a sentence, and press Enter or the send button.

Hello. I am completely new to this. What are you able to help me with?

Send that. Read what comes back. That's it, you have now used AI. There is no setup and no manual you needed to read first.

3 Ask for one real thing, plainly

A message to an AI is often called a prompt. Beginners can get a long way with a simple four-part shape. You will not need all four every time, but knowing they exist means you can reach for whichever one is missing when an answer falls flat.

  • Role. Who should it act as? A blunt editor, a patient teacher, a tenant advocate.
  • Context. What does it actually need to know? Your situation, the numbers, the background.
  • Command. What do you want done, in plain words?
  • Format. How do you want it back? A list, a short paragraph, a table, under 100 words.

Act as a patient, plain-speaking tenant advocate. My landlord hasn't fixed a leaking kitchen tap I reported two weeks ago. Write a short email asking him to fix it this week. Keep it under 100 words, polite but firm.

Why this matters A vague request gets a vague, generic answer. A specific one gets a useful one. The AI is answering the question you actually asked, which is usually vaguer than you meant it to be.
Speed up your learning Whenever an answer turns out exactly right, ask it: "Write me the exact prompt that would have gotten me this answer straight away." It hands you back the shortcut for next time, so you improve without having to study anything.
4 Talk back, don't start over

If the first answer is not quite right, do not delete everything and start a new question. Reply in the same conversation and correct it. The AI can see what was said earlier and will adjust.

  • "That's too long, cut it in half."
  • "Too formal, make it sound like me talking to a friend."
  • "You got the date wrong, it's the 14th, not the 4th."

Most useful answers take two or three exchanges to get right. That is normal, not a sign you are doing it wrong.

5 Check anything that matters

AI can state something false while sounding entirely confident about it. This has a name, a hallucination, covered in the glossary below. It is not rare, and it looks like every other sentence.

  • For facts, dates, prices, or numbers, ask "where does that come from" or check it yourself.
  • For anything medical, legal, or financial, treat the answer as a starting point, not advice. Confirm with an actual professional.
  • For names, links, quotes, or citations, check them independently. These get invented more often than you would expect, and they can look completely real.
Why this matters The cost of double-checking a wrong recipe is low. The cost of double-checking a wrong figure on a form, a wrong dose, or wrong legal advice is not. Match how carefully you check to how much it would cost to be wrong.
6 Keep your private information private

Once you have typed something into an AI chat, treat it as sent, the same way you would treat an email to a company you don't fully know. Do not paste in:

  • Passwords, login codes, or security answers
  • Medical record details
  • Bank account, card, or tax file numbers
  • Someone else's personal details without their permission
  • Passport, licence, or Medicare numbers
Why this matters Some AI providers may review or use conversations to improve future versions unless you've turned that off in settings. Private information typed into a chat can end up read by someone other than you. A simple rule of thumb: if you wouldn't say it out loud in a shop full of strangers, don't type it in either.
7 Know what it's bad at
  • Expecting one message to be enough. Real use is a back-and-forth, not a single perfect question.
  • Assuming it knows today's date, current prices, or recent events. Unless it tells you it has looked something up, treat "current" information as possibly out of date.
  • Mistaking confidence for correctness. A calm, well-written sentence is not proof that it's true. See step 5.
8 Five things to try today

Pick whichever one is actually true for you today. Real tasks teach you more than made-up ones.

Explain how interest on a savings account works, like I'm eleven years old.

Help me write a text to a friend, cancelling plans for Saturday. Keep it warm, not cold.

Give me a simple week of dinners for one person who does not enjoy cooking.

I'm nervous about a phone call I have to make tomorrow. Help me plan what to say.

Read this and tell me what's confusing about it: [paste in something you wrote]

Reference, Not A Step

Words you'll keep hearing

You do not need to memorise these before you start. Look them up here the first time you hit one.

Prompt
The message you type in. Asking a good question is often called "prompting".
Chat / conversation
The whole back-and-forth in one thread. The AI can see earlier messages in the same chat, which is why correcting beats restarting.
Model
The actual AI system doing the answering. Different models exist, some faster, some more careful, roughly like different staff with different strengths.
Hallucination
When the AI states something false while sounding completely confident. Not a glitch you'll be warned about, just a wrong sentence that reads like a right one.
Context
Everything the AI currently has in view: your message, earlier messages, anything you've pasted in. More relevant context usually means a better answer.
Token
A rough chunk of text, close to a short word or word piece. Mentioned mostly around cost and length limits. You can safely ignore it as a beginner.

Once This Feels Easy

Ready for week two?

Master prompts, reusable system prompts, and a few control words that make answers noticeably more useful. Still plain English, still free.

Read Part Two

Need more than a checklist?

We build the AI tools this guide told you to be careful with.

If you want AI set up properly on your own website, chatbots, content assistants, automation, submit a brief and we'll tell you what fits and what it costs.

Submit a Brief See AI Integration