1.
\*\*Role\*\*
— who the AI is ("conversion-aware communication specialist", not "helpful assistant")
2.
\*\*Objective\*\*
— what good output achieves, not just what it contains
3.
\*\*Core task\*\*
— the actual instruction, with \[bracketed\] variables
4.
\*\*Working method\*\*
— 3–4 bullets on HOW to think before writing
5.
\*\*Rules\*\*
— what to remove (filler, robotic phrasing) and what never to invent (facts, dates, commitments)
6.
\*\*Output spec\*\*
— exact deliverables, including an alternate version
This method might not be for everyone and in many cases a simple few sentences of a paragraph may do the trick. I use this for answering 1000+ reviews in a month or when I feel like I need more control over the prompt rather than receiving something generic.
The "working method" section is the one most people skip and it makes the biggest
difference — it forces the model to plan instead of pattern-matching to the blandest
possible answer.
Full example using this structure (a follow-up email prompt that doesn't sound desperate):
\`\`\`
You are a conversion-aware communication specialist who knows how to
move conversations forward without sounding needy or pushy.
Core task:
Write a follow-up email based on the context below. The tone should be
professional, polite, and confident. Remind the person of the original
request without sounding aggressive. End with a clear next step.
Context: \[what you previously sent\]
Goal: \[what response or action you want\]
Working method:
- Clarify the exact action or response you want from the other person.
- Remove pressure while keeping momentum in the conversation.
- Make the message easy to answer and difficult to misunderstand.
Output requirements:
- A primary message draft
- An optional softer or firmer variant
\`\`\`
