What AI can do, and what it can’t.
Where AI genuinely saves time at work, where it still lets you down, and a simple test to tell which is which before you rely on it.
- AI is strongest at language work: drafting, summarising, rewording, explaining and sorting. It gives you a fast first pass, not a finished product.
- It's weakest where accuracy really matters and is hard to check: facts, figures, legal or medical detail, and anything specific to your organisation.
- A good rule: use AI where you can easily check the result, and where a mistake would be cheap to catch.
Prefer to watch?
The video covers the same ground as the page below, in a few minutes.
Where AI genuinely helps
Most of the value at work today comes from everyday language tasks. These are jobs that take up your time but don't need your full expertise for every word.
- First drafts. Emails, letters, job adverts, meeting agendas, policy outlines. Starting from a draft is quicker than starting from a blank page, even if you rewrite half of it.
- Summaries. Turning a long report, email thread or meeting transcript into the key points and actions.
- Rewording. Making something clearer, shorter, friendlier or more formal, or adjusting it for a different reader, such as a customer instead of a colleague.
- Explaining. “What does this clause mean?” “What does this spreadsheet formula do?” “What does this acronym in the email I've just had stand for?”
- Spreadsheet help. Suggesting formulas, explaining error messages, or walking you through building a pivot table step by step.
- Ideas and structure. Brainstorming options, suggesting headings for a report, or listing questions you might have missed.
- Sorting and tagging. Grouping customer comments by theme, or pulling the dates and amounts out of a batch of messages.
Where it struggles
The same tools can let you down in predictable ways. Knowing them in advance saves a lot of embarrassment.
- Facts and figures. It can invent numbers, dates, names and references that look completely real. Never pass on a fact from AI without checking it against a source you trust.
- Maths and counting. Many tools have improved, and some now do proper calculations behind the scenes, but slips still happen. Let your spreadsheet do the sums.
- Your organisation's details. Unless it's connected to your company's own documents, it doesn't know your prices, policies, customers or systems. It fills the gaps with confident guesses.
- Specialist advice. Legal, medical, financial and HR questions often depend on fine detail and current rules. AI can help you understand a topic, but it isn't a substitute for a qualified person or your organisation's own guidance.
- Judgement calls. Whether a complaint is fair, how to support a struggling colleague, how to calm an upset customer. These need context, empathy and accountability, which AI doesn't have.
- Recent events. Its knowledge stops at a point in time. Some tools search the web, but check the date of anything that might have changed.
A simple test before you use it
Before you hand a task to AI, ask yourself two questions.
Can I easily check the result? If you can read an email draft and spot anything wrong, AI is a good fit. If checking would take as long as doing the work yourself, or you wouldn't recognise a wrong answer if you saw one, be careful.
What happens if it's wrong? A clumsy sentence in an internal note is cheap to fix. A wrong figure in a board report, an incorrect refund or bad advice to a patient is not.
Easy to check and cheap if it's wrong? Go ahead. Hard to check, or costly if it's wrong? Use AI to help you think or draft, then rely on your own checks, or on someone qualified, for the final word.
It's not just chatbots
Chatbots get the attention, but AI does plenty of other work in organisations, often quietly in the background. You may already rely on some of it without knowing.
- Forecasting. Predicting stock levels, staffing needs or which invoices are likely to be paid late, based on past patterns.
- Reading documents. Pulling details out of invoices, forms and delivery notes so nobody has to type them in by hand.
- Seeing. Checking photos for damage, reading number plates or spotting gaps on a shop shelf.
- Spotting the unusual. Flagging a payment, order or login that doesn't fit the normal pattern, so a person can take a look.
What good use looks like
The people who get the most from AI treat it like a capable assistant who has just joined the team. It's quick, tireless and keen to help, but it doesn't yet know how things work here, and its work needs checking.
Used that way, it can take the slow, repetitive parts of a task off your plate and leave you more time for the parts that need you: the judgement, the relationships and your knowledge of how things really get done.
Find your best AI task
- Jot down five tasks from this week that involved writing, reading or sorting information.
- For each one, ask the two questions: can I easily check the result, and what happens if it's wrong?
- Pick the task that scores best on both, such as tidying up a routine email or summarising a public article for your team.
- Try it with your approved AI tool, using made-up or public details only. Note how long it took compared with doing it yourself, including the time spent checking.
- Keep a short list of what worked and what didn't. It will come in handy when your team talks about AI.
Share it with your manager.
If your whole team is being asked to get up to speed on AI, a short practical session together saves weeks of guesswork.
About team training →Pass it on.
It's free, there's no sign-up, and it's written for people who aren't in IT.