The Big Idea
At almost every workshop, someone catches me during the tea break and asks, "Can ChatGPT do this?" or "Does Copilot know about that?" My answer is always the same.
Just ask it.
AI chat tools are really good at talking about themselves. They can tell you what they can do, what they can't, what they need from you, and when they're about to make something up. But you have to ask. It doesn't volunteer.
This is a small habit that changes everything about how useful AI becomes to you. Once you start asking the AI what it can do, you stop waiting for a teacher or a techy friend to tell you.
The Mindset Shift
Most of us learned to use the internet through Google. You type a few keywords, you get a page of links, you click the best one. That's a search-engine mindset. It's about finding.
An AI chat tool is different. You're not searching, you're instructing a helper. It works more like explaining a job to a new assistant than typing into a search bar. The more context you give, the better it does. You can describe your situation in whole sentences. You can paste in a document. You can ask a follow-up. You can say, "no, do it differently."
Stop searching. Start guiding.
The shift sounds small. In practice it changes what you can do on your own, without waiting for help.
Why This Works
AI chat tools have read an enormous amount of text, including their own documentation, help pages, and user guides. They know, in a fairly reliable way:
What they can do. Writing, summarising, translating, explaining, tutoring, coding, image understanding, and more. If you ask, they'll list it.
What they can't do. They can't access your real-time location. They can't read your email unless you paste it in. They usually can't open a link unless you have a special tool turned on. They'll tell you.
What they need from you. If you ask "can you help me with my rates notice," a good AI will ask to see the notice, or ask what exactly you want to know.
Where they're likely to get things wrong. Ask "how confident are you in that answer" and most AIs will tell you honestly.
AI tools are mostly honest about themselves but not always correct. They can be out of date about their own new features. If an AI says "I can't do X," and you read somewhere else that it now can, try it anyway. Don't take the AI's word as final, but do take it seriously.
Worked Examples
Here are six things people have asked me whether an AI can do. In each case, the answer is the same: ask the AI. These are real examples of what the AI is likely to say back.
Yes, I can help with that. To do a good job, it would help to know a few things: how many people are you feeding, roughly what's your budget for the week, are there any foods you or your family don't eat, and do you have a freezer and an oven or just a cooktop?
If you don't want to answer all that, I can also just start with a sensible budget plan for two adults cooking with basic equipment, and you can tell me what to change.
Yes. I've read the letter. It's a notice from your energy retailer telling you that your concession has not been applied to your account for the last two bills. It says you need to contact them within 30 days to have the concession backdated, otherwise you'll only get it from the date you call.
In plain terms: you're probably owed money back. The quickest action is to phone the number at the top of the letter and ask them to backdate the concession. I can draft what you could say on the call if you want.
I can usually help you troubleshoot it, yes. I can't actually see your network, so I'll need to ask you some questions as we go. To start: what exactly is happening? Is the Wi-Fi showing up as connected but nothing loads, or can you not see your network at all? And is this on one device or all of them?
Once you tell me that, I can walk you through the most likely causes in order, from simplest to most involved.
Absolutely, this is something I'm genuinely good at. Tell me what job you're going for, and paste the job ad if you have it. I'll ask you questions the way a real interviewer would, one at a time, and give you feedback on each answer before we go to the next.
We can also practise specific things you're worried about, like what to say about a gap in your work history, or how to answer "tell me about yourself" without rambling.
I can listen to audio you upload to me, yes, and I can transcribe it and tell you what's in it. What I can't do is listen live while you speak, or call someone on your behalf.
If you want to try: record the voice message, save it as a file on your phone, and attach it to a message to me. I'll write out what was said and we can go from there.
Good question, and the honest answer is: usually no. By default I don't carry memory between separate conversations. Each new chat starts fresh.
Some AI tools have a memory feature you can turn on, but it's off unless you switch it on. If you want me to pick up where we left off, paste in what we covered last time and I'll work from that.
Notice the pattern. In every case the AI does three useful things: it answers the "can you" question honestly, it tells you what it needs from you to actually help, and it offers a next step. You get those things for free just by asking.
Four Starter Questions to Ask Any AI
Any time you pick up a new AI tool for the first time, or you're not sure what it'll do with a particular job, open with one of these. They work for Claude, ChatGPT, Copilot, Gemini, or any other chat AI.
The four openers
"What can you help me with?" A plain overview, in the AI's own words.
"What can't you do?" The limits, honestly stated, before you waste time.
"What do you need from me to do a good job on this?" Turns the AI into a collaborator instead of a guesser.
"How confident are you in that answer?" Catches hallucinations before they become mistakes.
These are not clever prompts. They're just conversation. That's the point.
The Take-Home
You don't need a course, a manual, or a tech-savvy friend to find out whether an AI can help you with something. You need to open the app and type the question in plain English.
The AI will tell you yes, no, or "here's what I'd need to do it." That's the whole skill.
The AI is not an oracle you query. It's a helper you direct.
Once that clicks, digital literacy stops being a thing other people have and becomes something you can build, one conversation at a time. That's what this training is really about.