You do not need to become an AI expert. Start with a dull job, ask a clear question and keep your judgement firmly switched on.
There is a particular sort of AI advice that makes an ordinary Tuesday sound like a competitive sport. Learn 50 prompts. Automate your life. Become indispensable before the chap at the next desk does.
Or you could ask a chatbot to make a difficult letter easier to read. That is less impressive on LinkedIn, but considerably more useful. You do not need to master AI. You need to find out whether it can save you ten minutes without creating twenty minutes of checking.
Think of a general-purpose AI chatbot as a quick, reasonably articulate assistant for working with words and ideas. You type a request, called a prompt, and it produces an answer. It can explain, reorganise, compare and draft. It cannot know whether the answer is true simply because it sounds as though it ought to be.
Start with a job whose result you can judge yourself. Ask it to turn your rough notes into a polite email chasing a delayed delivery. Give it a list of questions you want to ask a garage before approving repairs. Have it suggest three ways to structure a speech, then pick the one that sounds like you rather than a regional sales director at an awards dinner.
Another good test is an unfamiliar bit of jargon. Ask: ‘Explain the difference between a heat pump and a gas boiler in plain English, then tell me what I would need to check before changing.’ That gives you a starting list, not an instruction to spend thousands of pounds. The useful part may be the questions it helps you ask a qualified installer.
ChatGPT has a free tier available to UK users, so there is no need to pay for a subscription to run these experiments. Features and limits can change; do not mistake a paid plan for a certificate of accuracy. If the free version cannot improve one of your real, recurring tasks, paying for more of it is unlikely to fix the problem.
Forget the elaborate ‘prompt engineering’ performances. Give the tool the job, the context, the format and the limits. For example: ‘Draft a short, courteous email to my broadband provider. My connection has dropped several times this week. I want to ask what they will do to investigate it. Do not claim I am entitled to compensation.’ You can leave out your name, address and account number. Add those yourself if you send the email.
Then talk back to it. ‘Shorter.’ ‘Less corporate.’ ‘What have you assumed?’ ‘Which parts would I need to check?’ A chatbot is often more helpful on the second attempt, when you tell it what was wrong with the first. This is not a test of your command of technology; it is editing, something you have probably been doing for decades.
You can also give it material you are entitled to share and ask it to work only from that material. A pasted, non-sensitive set of product specifications can become a comparison table. Ask it to mark anything the specifications do not say as ‘unknown’, rather than filling the gaps with guesswork. Check the result against the originals before you use it to choose what to buy.
The best AI task is one where you remain the expert on whether the answer is any good.
Keep the first experiments small. Try an email, a list of questions and a plain-English explanation. If one genuinely saves time, keep it. If it produces fluent rubbish, close the tab. You do not owe a chatbot continued employment.
Use the same caution you would with any outside service. Do not paste passwords, bank details, medical records or your entire pension statement into a public chatbot. Do not upload a colleague’s performance review or a customer list because a summary would be handy. The National Cyber Security Centre advises care over information entered into prompts; the Information Commissioner’s Office says organisations remain responsible for how they handle personal information when using AI tools.
At work, follow your employer’s policy and use only approved tools for work material. The UK GDPR is relevant when an organisation processes personal data, but switching to AI does not create an exemption from its data-protection obligations. If you are unsure whether a document is safe to upload, ask whoever handles information security at work. An anonymous summary of a problem is often enough to get a useful draft.
Privacy settings help, but understand what they do. In ChatGPT, you can turn off ‘Improve the model for everyone’ so that new conversations are not used to train OpenAI’s models. That does not remove ordinary chats from your history. Its Temporary Chat option keeps a conversation out of your history and model training while allowing retention for safety purposes. Neither setting is a good reason to feed it secrets it never needed.
There is a simpler rule: take names and identifying details out before you ask. ‘Write a tactful reply declining a request for unpaid overtime’ needs neither your manager’s name nor a copy of the staff rota. You can put the human details back into the finished draft, on your own device.
AI can invent facts, sources and quotations while sounding entirely certain. Its maker acknowledges this; the technical term is ‘hallucination’. Asking for a source is sensible, but a convincing-looking reference is not proof. Open the source and check that it exists and says what the answer claims.
This matters most where getting it wrong costs you money, health or legal trouble. A chatbot can help you list questions about taking a pension, but it does not have access to your actual State Pension record. GOV.UK has a service for checking your State Pension forecast, and MoneyHelper offers free, impartial guidance on money and pensions. Use the chatbot to prepare for those services, not to impersonate them.
Tax is another trap for a confident answer. If you ask about Income Tax, give the tax year and the part of the UK where you live. For the 2026–27 tax year, running from 6 April 2026 to 5 April 2027, Scottish rates and bands on wages, pensions and most other taxable income differ from those in England, Wales and Northern Ireland. Check any answer against HMRC’s published rules before acting on it.
The same principle applies to health. AI can help you put your symptoms into a clear list for a GP appointment; it should not decide whether you need treatment. For anything consequential, use it to understand the question and find the right professional or official service. That is not being timid. It is knowing which jobs belong to a machine and which still belong to you.
Start with a small task you can check, such as drafting an email or explaining unfamiliar terms. Give the AI clear instructions, keep sensitive information out and verify anything that affects your health, money or rights with an appropriate professional or official source. If it saves you time, use it; if not, you are allowed to remain unimpressed.
AI agents promise to book, buy and sort your admin while you put the kettle on. Here’s what they really do, what they cost and where they still fall over.
ChatGPT forgets what you need, writes like an estate agent and asks the same questions every time. Memory and custom instructions can fix most of that, if you set them up properly.
AI can sift applications, set shifts and flag performance. That does not make it a good manager. Here is where the software helps, where it fails and what you can challenge.
AI can turn a blunt complaint or awkward reply into a useful draft. Here’s how to get the wording right, protect private information and avoid sounding like a machine.
Newsletter
The best guides of the week — collected, without the noise, for anyone who likes to stay a step ahead.
