Part 2: AI for Everyone — How to talk to AI

The what, the how and the what not to

Every meaningful conversation begins with a question. | Image: AI-generated

Welcome back!

In Part 1 we established what AI is, what GenAI is (from here on, I will use the terms AI and GenAI interchangeably since the AI tools you’ll actually use day to day are all GenAI) and — just as importantly — what it isn't. If you haven't read Part 1 yet, it's worth starting there. This series builds on itself and the foundation matters.

In this article we will understand how to talk to AI. Because understanding what AI is and actually knowing how to use it are two very different things. One is knowledge. The other is a skill. And like any skill, it gets easier the moment someone shows you how.

That's where Mrs. Smith and Dr. Ray come in.

Mrs. Smith has been the head librarian at her university since forever. She knows every shelf, every catalogue, every corner of that library. Ask her where to find something and she'll have it in front of you in minutes — precise, reliable and just what you asked for.

Dr. Ray is a professor of history at the same university. He has spent three decades studying civilisations, revolutions and the forces that shape the world. Sit down with him and he won't just answer your question — he'll take you on a journey. He's brilliant, thorough, and wonderful to talk to.

They work in the same building. They respect each other enormously. And between the two of them, they have a lot to offer.

Google vs AI — what’s the difference?

Most of us learnt to use the internet the same way. Open Google, type a few words — “healthy breakfast recipes” or “fever home remedy” — and it brings back a list of pages that might have your answer. You've been doing this for years. It works. It's second nature.

So when people sit down with an AI app for the first time, that's exactly what they do. A few keywords, short and clipped, the way you'd type into a search bar. And the response they get back is fine. Useful enough. But nowhere near what the tool is actually capable of.

To understand why, let's pay a visit to Mrs. Smith and Dr. Ray.

Search points you toward information. AI helps you explore and understand it. | Image: AI-generated

Imagine you're a student at their university and you have a question about the French Revolution. You have two people you can go to.

You go to Mrs. Smith first. Within minutes she has a stack of the right books on the table in front of you — everything you need is in those pages. But finding the answers, reading through the material, making sense of it — that part is on you.

Then you knock on Dr. Ray's door. He doesn't hand you a list of books to read. He sits down with you and talks. He explains the causes, the consequences, the context, the controversies. He connects it to things happening in the world today. He answers your follow-up questions, and the ones after that. The conversation goes where you need it to go.

Google is Mrs. Smith. AI is Dr. Ray.

These days the line is blurring — even Google sometimes shows an AI-generated answer at the top of your results. But the distinction that matters here is not what technology is being used behind the search bar. It is whether you’re looking things up or having a conversation.

How to write a good prompt

The more context you provide, the more useful the response becomes. | Image: AI-generated

Think about how you would actually start a conversation with Dr. Ray.

If you walk into his office and say “Tell me about the French Revolution”, he’ll probably give you the entire history of it from start to finish. But if you sit down and say “I’m trying to understand why ordinary people in France were so angry in 1789 and what finally pushed them over the edge” — that’s when the real conversation begins.

That specific way you frame your question is called a prompt. A prompt is simply what you type into an AI app. It’s your question, your request, your instruction — whatever you want to call it.

Here is the single most valuable thing anyone can tell you about using AI — talk to it the way you would talk to a person.

Not keywords. Not a search query. A proper sentence, with context, the way you'd explain something to a friend or a colleague.

The difference this makes is immediate. Let's look at a few examples of good and bad prompts to understand how this works.

Bad prompt ❌

healthy breakfast recipes

Good prompt ✅

I am looking for some simple and healthy breakfast ideas for a family of four with vegetarian options and nothing that takes more than 20 minutes to prepare in the morning. Can you suggest a few recipes?

Bad prompt ❌

fever home remedy

Good prompt ✅

I have a mild fever of around 100°F that started last night, with no other symptoms. What are some safe home remedies I can try before deciding whether to see a doctor?

Bad prompt ❌

birthday message friend

Good prompt ✅

I want to write a short and warm birthday message for my friend who is turning 60 this year. She loves gardening and has a great sense of humour. Can you write something that feels personal and not too formal?

Notice what's happening in each of these. You're not just telling AI what you want — you're giving it the full picture. Who it's for, what the situation is, what tone you're looking for. The more it understands your context, the more helpful its response will be.

AI is not always neutral

Before we move on, there is something worth knowing about Dr. Ray.

He has spent his entire career reading, researching and forming views about the world. And whether he realises it or not, those years of reading have shaped how he sees things. He isn't biased in a deliberate way — he would never intentionally mislead you. But his perspective has been shaped by what he has read, who he has learnt from, and the world he grew up in.

AI is no different. It learnt from an enormous amount of human writing — books, articles, conversations, opinions — and humans, as we know, don't always agree. We all have our biases. And some of that found its way into AI.

This is what is known as training bias. It doesn't mean AI is untrustworthy. It means that on certain topics — particularly anything involving politics, religion, culture or society — the answer you get may lean in a particular direction without you realising it. Not because AI is trying to influence you, but because of what it was taught and by whom.

The simple habit to carry forward: on topics that matter to you, don't take one answer as the final word. Ask again from a different angle. Compare. Think.

AI is a starting point, not a verdict.

What not to share with AI

Now that you know how to talk to AI, there is an equally important question — what should you never say to it?

Remember how we said AI is like Dr. Ray? Imagine you've just met him for the first time. He is warm, helpful and clearly very knowledgeable. But would you hand him your passport? Would you tell him your bank account number, your medical history with your full name and address attached, or your passwords?

Of course not. And the same applies here.

AI apps are powerful tools, but they are still systems built and run by companies. What you type into them doesn’t simply disappear into thin air. Depending on the service, it may be stored and some of it may be looked at by the company’s team or used to help improve the system over time. This isn’t someone watching you personally — it’s a routine check to keep the tool safe and improve it. But it does mean that some things are better kept to yourself.

You don’t need to stop asking personal or sensitive questions; you just need to change how you ask them. Let’s look at a quick example. Notice that this time, the problem isn’t too little detail — it’s the wrong kind of detail.

Bad prompt ❌

My name is Meera Sharma, I am 58 years old and I have been diagnosed with Type 2 diabetes. My doctor prescribed Metformin 500mg. Is this the right medication for me?

Good prompt ✅

A 58 year old woman has been diagnosed with Type 2 diabetes. Her doctor prescribed Metformin 500mg. Is this a commonly prescribed medication for this condition and what should she know about it?

Same question. But one version hands over a full identity and the other doesn't. This is the balance worth holding on to. Context helps, identity doesn’t. Tell AI about your situation: what you’re cooking, who a gift is for, what symptoms you’re dealing with and so on. Leave out the things that identify you: your name, address, ID numbers, bank details and the like. Share the situation, not yourself.

A good rule of thumb — if you wouldn't say it out loud to a stranger sitting next to you on a bus, don't type it into an AI app.

Always verify what matters

Use AI as a starting point, not the final authority. | Image: AI-generated

Dr. Ray is a genius. But even the most confident professor gets it wrong sometimes. And that, occasionally, is precisely the problem.

Imagine you have a history examination coming up. You go to Dr. Ray and ask him the exact text of a letter written during the French Revolution. He answers without hesitation, completely assured of himself. But before you write that down in your exam, you walk down the corridor to Mrs. Smith. You ask her for the history book. You open it, find the page and confirm it for yourself.

That is cross-verification. And it is one of the most important habits you can build when using AI.

Whatever AI tells you — go and check it. Look it up on a trusted website. Read an article. Ask a professional. This is especially important for topics where getting it wrong has real consequences:

• Your health — always verify with your doctor
• Your finances — always confirm with your bank or a financial advisor
• Your legal rights — always consult a qualified professional
• News and current events — always check a reliable news source

AI is a wonderful place to start. But the final word always belongs elsewhere.

When the answer is not right, ask again

If Dr. Ray gives you an answer that doesn't feel right, you wouldn't just nod and walk away. You'd ask him to explain it differently. Or you’d try to ask the question differently. Either way, you’re aware that there is some communication gap happening here.

The same is true with AI. A bad answer is not a dead end. It's just the beginning of a better question.

If the response you get back is too long, ask it to summarise. If it's too vague, ask it to be more specific. If it got something wrong, tell it — “That doesn't seem right, can you reconsider?” If it didn't understand what you were asking, rephrase and try again. You are not locked into the first answer.

This is something many first-time users don't realise. They get one answer, feel dissatisfied and give up — assuming the tool is useless. But AI responds to persistence the same way a good professor does. The more you engage, the better the answers get.

There are no points deducted for asking again. There is no limit on follow-up questions. And Dr. Ray, unlike some professors, will never make you feel silly for not understanding the first time.

And that brings us to the end of Part 2.

We've covered a lot of ground today — how to talk to AI, what to never share with it, how to spot when it might be leading you astray, and what to do when it does. If there is one thing to take away from all of it, it's this: AI is a conversation, not a search box. Talk to it like a person, stay a little sceptical and don’t stop at the first answer.

Mrs. Smith and Dr. Ray will be back. In the next part, we are going to spend a little more time with Dr. Ray — because that habit of his, of being occasionally and confidently wrong, deserves an article of its own.

See you in Part 3!

Disclosure: This article was entirely brainstormed, structured and written by me. I used an AI assistant to generate the images, polish the grammar and refine the phrasing.