The practical summary
- Most people stall on AI not because it is hard, but because nobody showed them the ten things that actually matter. This is those ten things.
- The single biggest move is telling the tool who you are and asking it to question you before it answers.
- The one rule that keeps you safe: never paste anything into a public AI tool you would not put on a postcard.
- You do not need to be technical. Most people reach a useful working level in a few weeks of real use.
The short version
Here is the thing nobody tells you when you first open ChatGPT and stare at the blinking cursor. The tool is not the problem. You could hand the same account to two people in the same office and one of them will save five hours a week while the other decides it is overrated and goes back to doing everything by hand.
The difference is not intelligence. It is knowing about eight small moves. That is it. Learn those and you are past the point where most people give up. Everything below is those moves, written for someone who has an actual job to do and no interest in becoming a prompt engineer.
The cheat sheet
Keep this next to your keyboard for a fortnight. After that you will not need it. Every prompt here assumes you are pasting in the real document, email or data you are working with.
| The job in front of you | What to type | The catch to watch |
|---|---|---|
| A tricky email you're putting off | Write a polite but firm email to a client who is three weeks late paying. Short, British, no waffle. | It will be too keen. Cut a sentence, make it sound like you. |
| A long document you haven't read | Summarise this in five bullets a busy managing director would actually read. Flag anything that needs a decision. | It can miss the one number that matters. Skim the original for figures. |
| Messy notes into something usable | Turn these rough notes into a clear one-page brief with headings. | Check it hasn't quietly invented a detail you didn't give it. |
| A blank page you're stuck on | I need to write X. Ask me five questions first, then draft it. | This is the whole game. Answer the questions and the draft gets ten times better. |
| A spreadsheet you don't fully follow | Explain in plain English what this data is telling me and what I should be worried about. | Nothing confidential. See the one rule below. |
| A meeting you want to walk into sharp | I'm meeting a supplier about a renewal. Give me five smart questions and three things to watch for. | Tell it your goal for the meeting or the questions stay generic. |
| Something you need to understand fast | Explain [the thing] to me like I run a small business, not like I'm a data scientist. | Ask a follow-up. The second answer is usually the good one. |
| A decision you're not sure about | Argue the opposite of this plan. Where am I most likely wrong? | Use it before you commit, not after. It's a sparring partner, not a cheerleader. |
The five habits that make the difference
Tell it who you are. Start with "I run a six-person accountancy firm" or "I'm a marketing manager at a builders' merchant". The same question gets a far better answer once the tool knows the context. Most people skip this and wonder why the output feels generic.
Make it ask before it answers. "Ask me questions first" is the single most useful phrase you will learn. It turns a machine that guesses into one that actually works to your brief.
Give it an example of good. Paste in an email or report you were happy with and say "match this tone". It copies you better than you can describe yourself.
Push back. "Too formal, try again", "shorter", "you've missed the point, the real issue is X". A first answer is a first draft, never the finished thing. The people who get value treat it like a conversation.
Check the facts, always. AI is confident even when it is wrong. For anything that carries a name, a number or a date, read it before it leaves your hands. Treat it like a bright, fast intern who occasionally makes things up.
The one rule that keeps you safe
Never paste anything into a free, public AI tool that you would not be happy to see on a postcard. Client data, staff records, anything commercially sensitive, all of it stays out unless you are on a proper business account with the privacy settings sorted. This is not a reason to avoid AI. It is a reason to get set up correctly once, then stop worrying about it.
The mistakes that make people give up
Three things sink most first attempts. Asking one vague question, getting a bland answer and concluding it does not work. Expecting it to read your mind instead of your brief. And trying it once on the hardest thing you do rather than the boring thing you do twenty times a week. Start with the boring, repeated task. That is where the hours are hiding.
See what a few hours a week back would be worth
If your team is stuck at the blinking-cursor stage, a couple of hours of proper training moves them past it for good. That is what we do.
Book a callPractical AI training for your team
The Oxford AI School delivers plain-English AI training in Oxfordshire and London, across the UK, and live online for teams in the USA and other international markets. Sessions cover basic and advanced AI skills, responsible AI policy, marketing, GEO and business planning.
Basic AI Training
Practise the core habits with guided exercises based on real work.
Explore this trainingAI Basics Deep Dive
Build stronger prompting, analysis and repeatable AI workflows.
Explore this trainingIntroduction to GEO
Learn how customers discover and assess businesses through AI search.
Explore this trainingAI for Marketing Teams
Apply better AI habits to research, content, campaigns and planning.
Explore this trainingThe Last Cheat Sheet You'll Ever Need When You Start to Use AI at Work: frequently asked questions
Which AI tool should I start with at work?
Start with one, not five. ChatGPT is the most common first choice because it is the one most people have heard of and the interface is forgiving. Claude and Google Gemini are strong alternatives. Pick one, use it for a fortnight, and don't let choice paralysis become the reason you never begin.
Is it safe to use ChatGPT at work?
It is, as long as you follow one rule: nothing confidential goes into a free public account. For real business use you want a paid business or team plan where your data is not used for training and the privacy settings are set up properly. Get that right once and you can relax.
How long does it take to get good at using AI at work?
Most people reach a genuinely useful working level in a few weeks of real, daily use. Not months, and certainly not a computer science degree. The cheat sheet above covers most of what separates a confident user from a frustrated one.
Do I need to be technical to use AI at work?
No. If you can write an email and describe what you want, you can use AI well. It responds to plain English, not code. The people who struggle tend to be the ones who overthink it, not the ones who find it too technical.
What's the best first task to try AI on?
Pick something boring that you do often, such as replying to routine emails, summarising documents or tidying up notes. High-frequency, low-stakes tasks let you build confidence and see the time saving quickly, without risking anything important while you are learning.
Can my whole team learn this, or just me?
The whole team, and that is where the real return sits. One person saving a few hours a week is useful. A ten-person team all doing it is a different business. Group training gets everyone to the same level fast and stops AI becoming one person's private hobby.
Harry Lang runs The Oxford AI School, teaching teams and businesses the basics of AI in a fun, easy and practical way. A marketing leader turned AI trainer, he can be reached at Learn@TheOxfordAISchool.com.
AI for Business Basics series
Six straight answers to the questions business owners and directors ask before they train a team.