Generative AI for Everyone

Lesson 2

Talking to AI

By the end you will reliably get useful output, the foundation for every build lesson.

See it first

Two people use the exact same model. One gets generic mush; the other gets something they actually ship. The model did not change. The way they asked did. Good prompting is not magic words; it is giving the model what it needs to help you.

The anatomy of a usable prompt

A strong prompt usually has five parts. You do not need all five every time, but knowing them turns a vague request into a precise one.

  • Role: who the AI should act as. "You are a meticulous financial analyst."
  • Task: what you want done. "Summarize this report's key risks."
  • Context: the material to work from. "Here is the report: [paste]."
  • Constraints: the rules. "Five bullets, plain English, no jargon."
  • Format: the shape of the output. "Return a markdown table."

Put together: "You are a meticulous financial analyst. Summarize the key risks in the report below in five plain-English bullets. Report: [paste]."

The one idea that matters most: context

If you remember one thing, remember this: giving the model your context, your documents, data, and writing, changes output quality more than any clever phrasing. A generic prompt gets a generic answer. The same prompt plus your actual material gets something useful.

Without context

"Write a project update."

→ Generic, hollow, could be about anything. You will rewrite it from scratch.

With context

"Write a project update from these notes: [paste your real notes]. Audience: my VP. Tone: confident, concise."

→ Specific, usable, mostly done.

This is the lever you will pull in every build lesson, and the thing the Prompt Engineering course is built around.

Iteration: the first answer is a draft

Never treat the first response as final. Steer it:

"Shorter." → "More formal." → "Add a section on risks." → "Give me two versions."

Each instruction reshapes the output. Iterating is faster than getting the perfect prompt on the first try, and it is how real work gets done with AI.

Choosing a tool

The major assistants are more alike than different. A rough guide:

ToolGood default for
ChatGPT (OpenAI)General-purpose, huge ecosystem, lots of integrations
Claude (Anthropic)Long documents, careful writing and reasoning
Gemini (Google)Tight integration with Google products
Specialized toolsSpecific jobs (image generation, transcription, coding)

Do not agonize. Pick one, learn it well, and switch only when a specific task calls for it. The skills in this course transfer across all of them.

What you can now do

  • Structure a prompt with role, task, context, constraints, and format
  • Explain why adding your own context beats clever wording
  • Use iteration to steer output instead of accepting the first draft
  • Choose a sensible default AI tool without overthinking it

Check your understanding

1 / 3
context-is-the-lever

Which addition most reliably improves the quality of AI output?

What's next

You can get useful output. Now learn when to trust it, and how to use it safely at work: Judgment: Limits, Hallucinations & Safe Use.

Test your understanding

Prof is ready

Prof will ask you questions about talking to AI, a just-enough prompting primer — not explain it. You'll be surprised what you don't know until you have to say it.

Finished this lesson?

Read through the lesson first (0/20s).