Setup
You're going to build a four-element prompt for a real work task, submit it to an AI tool, and save the output. In the next lesson, you'll diagnose that output for failures — so this lesson deliberately separates construction from evaluation.
Instructions
- Pick a task. Choose a real work task you need to do this week — or select from these:
- Draft a follow-up email after a meeting
- Summarize a document or report
- Create an agenda or outline for an upcoming presentation
- Write a first draft of a recurring report section
- Write your prompt using all four elements:
- Role: Who are you? (Your actual role and organization context)
- Context: What specifically are you working on? For whom?
- Format: How long? What structure? What tone?
- Constraints: What MUST be included? What should be avoided? Any specific details?
- Label each element. Put [ROLE], [CONTEXT], [FORMAT], [CONSTRAINTS] tags before each section of your prompt so you can see the structure.
- Submit to any free-tier AI tool — ChatGPT, Claude, or Gemini.
- Save the output. Copy the complete, unedited AI response. Do NOT evaluate it yet — just save it as-is. You'll diagnose it in the next lesson.
Quality Check
Before submitting your prompt:
- [ ] Does it have all four elements?
- [ ] Is each element labeled?
- [ ] Is the context specific to YOUR situation (not generic)?
- [ ] Do the constraints include at least one "must include" and one format specification?
- [ ] Would a competent colleague understand what you need from reading this prompt?
How confident are you?
- "I can write a four-element prompt from scratch": 1 2 3 4 5
- "I understand why each element matters": 1 2 3 4 5
If either is below 3, review the before/after comparison in Lesson 3.2.
Bridge
You've built a real prompt and have real AI output. Now comes the skill that separates professional AI use from naive AI use: finding what's wrong with that output before it goes anywhere.