Start Here · Chapter 02
Prompting that gets results
How to ask better questions, get useful outputs, and avoid common mistakes.
Why prompting matters
A prompt is not a magic spell — it is a brief for a very fast assistant who takes you literally. Vague asks produce vague answers. Specific asks produce usable drafts.
Good prompting is less about tricks and more about clarity: who you are, what you need, who the output is for, and what “good” looks like.
The four-part prompt framework
Use this structure for most tasks:
- Role — “Act as a patient teacher explaining to a non-technical manager.”
- Task — “Summarize this meeting transcript into decisions and action items.”
- Context — paste relevant background, constraints, or audience details
- Format — “Use bullet points, under 200 words, professional tone.”
Examples that work
Weak: “Write an email about the project.”
Strong: “Draft a polite email to my client explaining a one-week delay on the website launch. Tone: confident, not apologetic. Include: reason (vendor issue), new date (March 12), and next step (review call Friday). Under 150 words.”
Weak: “Help with marketing.”
Strong: “I run a local bakery. Suggest 5 Instagram post ideas for Valentine’s week. Audience: families and couples within 5 miles. No jargon.”
Common mistakes to avoid
- Asking one prompt to do ten unrelated jobs — split into steps
- Forgetting to say who the audience is
- Accepting the first draft without reading it
- Sharing sensitive company or personal data on tools your employer has not approved
- Assuming the AI remembers yesterday’s chat unless you are in a saved thread with memory enabled
Iterate like a conversation
Follow-up prompts are where quality jumps: “Shorter.” “More concrete examples.” “What am I missing?” “Play devil’s advocate.” “Give me three options ranked by effort.”
Treat prompting as collaboration. You bring judgment and context; the AI brings speed and structure.
Practice exercise
Take a task you did manually last week. Write a four-part prompt. Compare the AI output to what you actually sent or produced. List three edits you had to make. Those edits teach you what to specify next time.