Let AI Ask You Better Questions

You want to try something new, but you don't know what. You could ask AI for a hundred suggestions. Or you could let it ask you a few questions first. The second approach can be far more useful.

Imagine you're planning a weekend away but haven't chosen a destination. Ask for the ten best places to visit and you'll probably receive a familiar list. Yet you don't know whether you want adventure, somewhere peaceful or an excuse to catch up with friends. More destinations won't necessarily help you make that decision.

Try saying: 'Help me work out what kind of weekend I'd enjoy. Ask one question at a time before suggesting places.' The assistant may ask who is travelling, how far you want to go or whether you'd rather explore somewhere new or relax. Each answer helps define a problem that wasn't clear at the start.

This differs from simply adding context to a prompt. When you already know your requirements, you can supply them directly. Here, the point of the questions is to help you *discover* what matters. Perhaps you realise your main priority isn't a particular place but the ability to switch off and avoid a complicated journey.

A useful question is open enough to invite thought but focused enough to move you forward. 'Describe your ideal life' is rather big for planning a weekend. 'Would you rather spend Saturday exploring or staying in one place?' gives you something manageable to consider. A good assistant can follow your answer instead of reading from a fixed questionnaire.

Suppose you say you're travelling with children, prefer a short drive and want a mixture of indoor and outdoor activities. The assistant might ask how important cost is, then suggest three kinds of weekend rather than a list of random places. You can challenge its questions too. If it assumes you're interested in theme parks, say that isn't the direction you had in mind.

This method works for choosing a hobby, shaping a project, exploring learning goals or deciding what would make a routine easier. It can even help when you feel stuck because you don't have the words for the problem yet. Tell the assistant what you're uncertain about and invite a small, thoughtful interview.

There's another reason this approach can be useful: our first request often contains an assumption we haven't noticed. 'Find me a peaceful weekend away' sounds specific, but what if peace means somewhere without crowds to you and somewhere without chores to the person travelling with you? A few questions can reveal that you're trying to satisfy two different needs. Without that discovery, a long list of destinations would only disguise the disagreement.

Consider the way a helpful friend might respond. They probably wouldn't immediately announce the ten best hotels in the country. They might ask whether you're going alone, how much time you've got and what you're hoping to get away from. An AI assistant can follow a similar pattern, though it needs you to invite that sort of conversation when the question is still open.

You can make those questions less leading by setting a simple expectation: 'Ask me one open question at a time. Don't assume the answer, and explain why each question matters if it isn't obvious.' If the assistant starts pushing a particular destination, you can redirect it. The learner should feel able to explore possibilities, not feel shepherded towards the model's preferred answer.

Sometimes the answer to a good question will surprise you. Perhaps you realise you don't want a weekend away at all; what you really want is a few hours without responsibilities. That changes the task completely. AI hasn't discovered an objective truth about your life. It has helped you put words to something you were already feeling, which gives you a better foundation for deciding what to do.

Be aware, too, of the difference between asking useful questions and conducting an endless interview. If you already know that distance, cost and accessibility are the deciding factors, you may not need another ten preferences. Ask the assistant to summarise what it's learned and offer two or three directions. You can always return to questions if the suggestions reveal another important uncertainty.

Of course, not every request needs a series of questions. If you know what you want, get on with it. But when the problem itself is fuzzy, an AI assistant can help you frame the question before trying to answer it.