Home / Lesson 2  ·  10 minutes  ·  Needs: Lesson 1

How to Write Your First Prompt

Bad prompts get generic answers. The fix is a five-part recipe: role, task, context, format, constraints. Learn it once — it works in every chatbot you will ever touch.

The recipe

1Role — who is it being?

"You are a patient Excel tutor." Roles set vocabulary, depth, and tone in one line. Skip it and you get the generic default voice.

2Task — what should it do?

"Explain VLOOKUP with one worked example." One verb, one deliverable. Vague tasks ("help me with spreadsheets") get vague tours.

3Context — what does it need to know?

"I track inventory for a bakery in..." Background the model cannot guess: your level, your data, your constraints. This part separates okay answers from great ones.

4Format — what shape should the answer take?

"Keep it under 200 words with bullet steps." Length, structure, tone. Models happily ramble unless you specify the container.

5Constraints — what must it avoid?

"No jargon; if you're unsure, say so." Guardrails against the failure modes from Lesson 1 — including asking it to flag uncertainty.

Put together

You are a patient Excel tutor (role). Teach me VLOOKUP with one worked inventory example (task). I run a small bakery and know only SUM (context). Keep it under 200 words with numbered steps (format). No jargon — define every new term (constraint).

Run that, then run just "teach me VLOOKUP" in a fresh chat. The difference is the entire lesson: specificity in, quality out.

Three beginner power moves

Try it now: take one real task from your week — an email, a plan, a summary — and prompt it with all five parts. Then try it with one part. You will feel exactly which parts carry the weight.

Next on the path