Your First Conversation: What to Expect
Lesson 2.3 — Your First Conversation: What to Expect
Estimated reading time: 14 minutes
The first time most people use an AI assistant, one of two things happens. Either they're underwhelmed — they ask something simple, get a bland answer, and think "what's the big deal?" Or they're genuinely surprised — they ask something real, get a response that's unexpectedly thoughtful, and immediately think of ten more things they want to try.
The difference, almost always, is in how they started the conversation.
This lesson is about how to have a first conversation that lands in the second category.
What AI Chat Actually Feels Like
Let's set realistic expectations. Talking to an AI assistant is not quite like talking to a human, and not quite like using a search engine. It's something new.
When you type a message and hit send, there's a brief pause — usually a second or two — and then the AI starts typing its response. You'll see the words appear one by one, or in small chunks, as if someone is writing in real time. This isn't for show; it's actually how the response is generated, token by token.
The responses are often longer than you'd expect. If you ask a question, you might get four paragraphs when you expected one sentence. This can feel verbose at first. You can always ask for shorter answers — just say so directly.
The AI will be polite, patient, and never annoyed. You can ask the same question five different ways. You can say "that's not what I meant, try again." You can change direction mid-conversation entirely. It doesn't get frustrated.
What it won't do is read your mind. The more clearly you explain what you want, the better the response. Vague input gets vague output. Specific input gets useful output. This is the core skill of working with AI, and we'll practice it throughout this course.
How to Start a Conversation
Open ChatGPT, Claude, or Grok. You'll see a text box at the bottom of the screen. Click into it and type your first message.
Don't overthink your first message. A lot of people freeze here, as if there's a right answer they need to get correct. There isn't. The AI has read the internet. There is almost nothing you can ask that will confuse, offend, or stump it on the first try.
That said, here are some genuinely good first messages to try:
If you want to explore what it can do:
"I'm new to AI assistants. Can you show me three or four different things you're good at, with a brief example of each?"
If you have something specific you're working on:
"I need to write an email to my landlord asking them to fix a broken heating system. I've been asking for two months and nothing has happened. Can you help me draft something firm but professional?"
If you want to learn something:
"Explain how vaccines work. I have a science background but I want the version I could explain to my ten-year-old."
If you just want to get comfortable:
"Tell me something interesting about how language works."
Any of these will get you a response worth reading. After that, you just react to what you see — and the conversation develops naturally from there.
What Good vs. Bad First Messages Look Like
This is one of the most practical things in this lesson. The difference in output quality between a weak prompt and a strong one is enormous.
Example 1: Writing help
Weak: "Write me an email."
The AI doesn't know who the email is for, what it's about, what tone is appropriate, or what you want the recipient to do. It'll write something, but it'll be generic and probably useless.
Strong: "Write a professional email to a potential client named Sarah who I met at a conference last week. I want to follow up on our conversation about her company's social media strategy and suggest a 30-minute call to discuss how I could help. Keep it brief — two short paragraphs max."
Now the AI has everything it needs: who, what, why, tone, and length. The result will be something you can actually send.
Example 2: Learning something new
Weak: "Explain machine learning."
This will get you a textbook answer — technically accurate but probably too long and pitched at an unknown level.
Strong: "I'm a nurse with no tech background. Explain machine learning to me like I understand biology but not computers. Use analogies if you can."
Same topic, but the context transforms the response. The AI knows how to calibrate what it says.
Example 3: Getting a decision
Weak: "Should I quit my job?"
The AI will respond with a long, balanced "on one hand... on the other hand" answer that doesn't actually help you decide anything.
Strong: "I'm considering leaving my job as a marketing manager after five years. The pay is good but I'm bored and I've had a job offer from a smaller company doing something more interesting for 15% less pay. What questions should I be thinking through before making this decision?"
This reframes the question usefully — instead of asking the AI to decide for you (it can't, and won't), you're asking it to help you think. That's a much more productive use of it.
Key takeaway: Before you type a message, spend 15 seconds asking: Does the AI know who this is for? What format do I want? How long should the response be? What's the context? A little more setup pays off in dramatically better answers.
The Back-and-Forth: Conversations, Not Commands
One thing that surprises new users is how much better AI assistants get when you treat them as conversations rather than one-shot commands.
Most people type one message, read the response, and then either use it or give up. But the real power is in the follow-up.
The response was too long? Say so: "That's too long. Give me the key points in three bullet points."
The tone was off? Say so: "That's a bit formal. Can you make it sound more like a friendly professional — like an email you'd get from a smart colleague, not a lawyer?"
You want a different angle? Ask: "Interesting. Now give me the counterargument."
One section was great but the rest wasn't? Tell it: "The second paragraph is exactly what I wanted. Can you rewrite the rest of the email to match that tone?"
This iterative process — give feedback, refine, give more feedback — is how professional AI users get great results. The first response is a starting point, not a final answer.
Here's what a real back-and-forth might look like:
You: Can you help me write a short bio for my LinkedIn profile? I'm a project manager with 8 years of experience, I specialise in healthcare IT, and I recently led a team that migrated a hospital system to electronic records.
AI: [Writes a 3-paragraph bio]
You: Good structure, but it sounds a bit generic in the first paragraph. Can you open with something more specific about the healthcare IT space — maybe reference the complexity of working in regulated environments?
AI: [Rewrites with more specific opening]
You: Much better. Can you shorten it by about 30% without losing the key achievements?
AI: [Produces final version]
Three exchanges, and you have something genuinely usable. That's the rhythm.
What AI Remembers Within a Conversation
Within a single conversation, the AI remembers everything. Every message you've sent, every response it's given — it's all in context. So you can refer back to things naturally:
- "Go back to the second email you wrote and change the subject line."
- "Earlier you mentioned three options. Tell me more about option two."
- "Keep the same tone as the last response but apply it to this new draft."
This works because the AI is looking at the entire conversation history each time it responds. You don't need to remind it what you said five messages ago — it knows.
The limit: Very long conversations can start to hit what's called the "context window" — the maximum amount of text the AI can hold in memory at once. For most everyday conversations, you'll never reach this. But if you're doing a three-hour research session with hundreds of messages, you might notice the AI starting to "forget" earlier parts. We cover this in detail in Lesson 2.4.
What AI Does NOT Remember Between Conversations
Here's the thing that catches almost everyone out at first: when you start a new conversation, the AI has no idea who you are or what you talked about before.
Every new conversation is a blank slate.
If you spend an hour in one chat teaching the AI about your business, your writing style, and your preferences — and then you close that tab and open a new chat tomorrow — it's gone. You'll need to re-introduce yourself.
There are ways to work around this (custom instructions, "memory" features, and copy-pasting context), and we'll cover all of them in Lesson 2.4. For now, just know it: start a new chat, start fresh.
A Few Things That Will Surprise You
It will be wrong sometimes. AI assistants can and do make factual errors. They can confidently state something incorrect. This is called "hallucination" in the field, and it's a known limitation. For anything important — medical information, legal advice, financial decisions, factual claims you'll repeat to others — always verify from a primary source.
It won't judge you. You can ask basic questions without embarrassment. You can admit you don't understand something. You can ask it to explain the same thing three different ways. It will never sigh, roll its eyes, or make you feel foolish.
It has opinions, sort of. If you ask, most AI assistants will share a perspective on a topic. They'll say "I think" or "in my view." These aren't strong convictions — they're pattern-based responses — but they can be useful prompts for your own thinking.
It can be wrong about itself. If you ask the AI what it can or can't do, it may give you an incomplete answer. The best way to find out if it can do something is just to try it.
Your Task for This Lesson
Before moving on, have at least two real conversations — one on ChatGPT and one on Claude (and try Grok if you signed up). Don't use simple test questions. Use something from your actual life:
- An email you've been putting off writing
- A concept from your work you've always found hard to explain
- A decision you're weighing up
- Something you've been curious about but never looked into
Notice the difference between a vague question and a specific one. Notice how follow-up messages change the quality of the response. You'll learn more in ten minutes of real use than in an hour of reading about it.
Up next: Lesson 2.4 explains how AI memory works — and crucially, how it doesn't — so you can stop losing context and start having more productive conversations.