Your First Conversation

It's a conversation, not a search engine

The single biggest beginner mistake is treating the chat box like Google. Twenty years of search engines have trained us all to type in caveman: "Dhaka weather tomorrow", "chicken recipe easy", "visa Malaysia requirements". Short. Keyword-y. No verbs.

A chatbot works the opposite way. It's built to understand full, ordinary sentences — the way you'd talk to a colleague. Compare:

The keyword version gets you a generic answer. The sentence version gets you an answer about your actual situation — because you gave it one. And unlike a search engine, the conversation doesn't end there: you can reply "make the picnic plan cheaper" or "I forgot, one guest is vegetarian" and it adjusts, remembering everything said so far in this chat.

Relax — you cannot break anything

Before you type your first message, a few pressure-relieving facts:

• Typos and grammar don't matter. Misspell every third word and it will still understand you. It's read more typo-filled text than any teacher alive.
• Any language works. Write in Bangla, in English, or in the mixed Banglish you actually text your friends in — all three are fine. You can even write in one language and ask for the answer in another.
• Speaking works too. Every major chatbot app has a microphone button next to the chat box. Tap it and just talk — great if typing on a phone is a chore.
• There is no wrong message. Worst case, the answer is useless and you start a new chat. Nothing is graded, nothing is saved to some permanent record of your questions, nobody is judging. (What actually happens to your chats privacy-wise is a fair question — Chapter 3 covers it.)
• Don't like the answer? Say so. "Too long." "That's not what I meant." "Try again, simpler." It never gets offended, and it usually improves.

One habit to start today: one chat, one topic

Here's the one piece of technique worth learning on day one. Each conversation is a separate room. Inside one chat, the AI remembers everything you've said in that chat — which is wonderful when you're refining a plan, and messy when you're not.

Say you spend twenty messages planning a Cox's Bazar trip, then in the same chat ask for help with a job application. The AI still has beaches and bus timings on its mind, and your cover letter may come out weirdly travel-flavoured. Old context bleeds into new answers.

The fix costs one click: new topic, new chat. Trip plan in one chat, job application in another, kid's homework help in a third. Your chat history sidebar becomes a tidy set of labelled folders instead of one junk drawer. (Why this happens — and how AI "memory" really works — is the whole story of Context & Memory, later in this chapter.)

A guided first session

Enough theory. Open your AI tool, start a fresh chat, and send any one of these five starters — pick whichever touches your real life today. Each is deliberately from a different corner of life, to show the range:

1. Everyday life: "My electricity bill seems higher than usual this month. Give me a checklist of the most common reasons and how to check each one."
2. Work: "I need to politely tell a client their payment is 3 weeks overdue, without damaging the relationship. Draft a short, courteous message."
3. Family: "My mother has just been prescribed a low-salt diet. Suggest a week of familiar Bengali-style meals that fit it, and tell me what I should still confirm with her doctor."
4. Learning: "Explain how a fixed deposit differs from a savings account, using a simple example with real-ish numbers."
5. Fun: "Plan a birthday surprise for my best friend who loves cricket and street food, with a small budget. Give me 3 ideas from simple to ambitious."

Send one. Read the answer properly. Notice that it responded to your details — the overdue payment, the low-salt diet — not just the general subject.

The "aha" exercise: watch it obey

Now the moment where most people actually get it. The magic of a chatbot isn't the first answer — it's what happens when you push back. Whatever answer you just received, send these three follow-ups, one at a time:

1. "Make it half as long." Watch it compress without losing the point.
2. "Now rewrite it in a warmer, friendlier tone." Same content, different personality.
3. "Explain it so simply that a 10-year-old would understand." Watch the jargon evaporate.

Three commands, three instant rewrites, zero complaints. No human colleague reshapes their work three times in a row this cheerfully. Once you've felt that — the answer is clay, and you're the one shaping it — you've crossed the real beginner's threshold. The full toolkit of shaping moves is in The Art of the Follow-up.

I'm having my very first proper conversation with an AI. Here's a real thing on my mind this week: [describe it in 2–3 sentences — a task, a decision, a worry, anything].

Help me with it. Then tell me one follow-up question I could ask you to make your answer even more useful.

Write a short message inviting my neighbours to a building meeting about the broken lift, this Friday at 8pm.

(After it answers, send these one at a time and watch what happens:
1. "Make it half as long."
2. "Friendlier, less formal."
3. "Now in Bangla.")

Etiquette myths, cleared up

Beginners quietly worry about the strangest things. Let's settle them:

• "Should I say please and thank you?" Not required — the AI has no feelings to hurt and works equally hard either way. But it costs nothing, and plenty of people find the conversation flows more naturally when they write like themselves. Do whatever feels normal.
• "Am I annoying it with so many questions?" Impossible. It doesn't get tired, bored, or irritated. Your fortieth question gets the same energy as your first. Ask it to re-explain something five times; it will, five slightly different ways.
• "This question is too dumb to ask." This might be the best thing about AI: dumb questions are free. No colleague raising an eyebrow, no feeling exposed in a meeting. "What's the difference between a debit and credit card, really?" — the questions we've all been too embarrassed to ask a person are exactly what this tool is perfect for.
• "It said 'I'm happy to help' — is it actually happy?" No. It writes the way helpful humans write, because that's what it learned from. Politeness in, politeness out. There's nobody home — which is precisely why you can ask it anything without social cost. (More on what's really going on inside in How AI Works.)

Key takeaways

• Talk in full sentences with your real details — it's a conversation, not a keyword search. The links-list era is over; here, the answer talks back.
• Typos, any language, mixed language, voice input: all fine. There is no wrong way to type.
• You can't break anything. Worst case: new chat, fresh start.
• One chat = one topic. It keeps answers clean and your history tidy.
• The real power is the follow-up: shorter, friendlier, simpler — the answer is clay, you're the sculptor.
• No please required, no question too dumb, no patience limit. Use that freedom.