First Conversations with AI

L1 Lesson 3 of 5 — First Steps

1

2

3

4

5

The first step: Just start

The most important thing up front: you can’t break anything. AI chatbots are made for experimenting. Every conversation is an experiment — and the first answer is always just a draft.

Three conversations compared

Conversation 1: Good — Clear context and task

You:

I’m preparing a team meeting and need an agenda. The meeting is 60 minutes, topic is Q3 quarterly planning. Attendees are 5 people from marketing and sales. Please create a structured agenda with time slots.

AI: (creates a detailed agenda with welcome, Q2 review, Q3 goals, open discussion, and next steps — each with time allocations)

Why this works: You gave the AI everything it needs — duration, topic, context, format. The more relevant information you provide, the better the result.

---

Conversation 2: Medium — Too vague

You:

Write something about quarterly planning.

AI: (writes a generic article about quarterly planning — technically correct, but not what you need)

The problem: “Write something” could mean anything — a blog post, an email, a presentation. Without context, the AI guesses, and it usually guesses wrong.

Simple fix: Add what it’s for and what format you need.

---

Conversation 3: Weak — Fact query without verification

You:

What was Company X’s revenue in 2025?

AI: (states a specific number with a source reference)

The risk: The AI might make the number up. This is exactly what Hallucination An AI response that is false or fabricated but sounds confident and convincing. s are — confidently stated misinformation. Specific numbers, statistics, and source citations should always be independently verified.

What makes a good conversation?

Do

• Provide context: Who are you, what do you need this for?
• Name the desired format: List, email, table?
• Treat the first answer as a draft and iterate
• Follow up when something doesn't fit: 'Make it shorter' or 'Focus on point 3'

Don't

• Ask one-word questions and expect perfect answers
• Accept numbers and facts without verification
• Give up when the first answer isn't perfect
• Enter sensitive data like passwords or health information

The art of follow-up

A good AI conversation isn’t a single Prompt The input you send to an AI model — your question, instruction, or task. — it’s a dialogue. You can always:

• Refine: “That’s good, but make it more formal.”
• Focus: “Just concentrate on the first point.”
• Redirect: “Not as a list, but as flowing text.”
• Challenge: “Are you sure about that number? Where did you get it?”

The AI forgets nothing within the conversation (as long as the context window holds). So you can work toward the desired result step by step.

Try it yourself

Exercise 1: The meeting experiment

Ask the AI to write an email to your team canceling a meeting. On the first try, just say “Write a cancellation email.” On the second try, provide the reason, tone, and recipients. Compare the results.

Exercise 2: The refinement dialogue

Ask the AI for a text (e.g., a summary of your last project). Regardless of the first answer — ask three follow-up questions that improve the text. Watch how the output evolves.

Exercise 3: The fact check

Ask the AI something verifiable — e.g., the CEO of a well-known company or the founding year of an organization. Check the answer with a quick web search.

Think further

The first answer is never the final product. AI is a sparring partner, not a vending machine. The better you learn to work with it, the better the results become.

In the next lesson, we’ll look at the limits: When does AI get things wrong — and why?