The single most effective change most people can make to their AI output is adding structure to their prompts. Not longer prompts — structured ones. This guide teaches the 5-part framework behind every high-quality prompt, then shows you three finished examples to steal.
The 5 parts
1. Role
Tell Claude who to be. "You are a senior copywriter" produces very different output than no role at all. The role should match the expertise you'd actually want reviewing the work. Specific beats generic: "B2B sales director with 15 years of enterprise software experience" beats "salesperson."
2. Context
Give the situation. Who is this for? What's the goal? What does Claude need to know? Context is where most prompts fall short. Include: audience, purpose, constraints, relevant background, and what "good" looks like for this task.
3. Example (when helpful)
One good example is worth three paragraphs of explanation. If you have a piece of writing you want Claude to match in tone, paste it. This part is optional for simple tasks, essential for tone-matching and format-sensitive work.
4. Rules
Explicit constraints prevent the most common failures. Tell Claude what NOT to do. Eliminate the AI habits that annoy you: bullet points when you want prose, filler phrases, excessive length, hedging, repeating the question back.
5. Format
Tell Claude exactly what to deliver. "A 3-paragraph response" is clearer than "write something." "A numbered action plan with deadlines" is clearer than "help me plan."
The full template
You are [role].
Context: [relevant background — who this is for, what it's about, constraints]
Task: [exactly what you need Claude to produce]
Example of right tone/format (optional): [paste one if you have it]
Rules:
- [specific constraint]
- [e.g. "No bullet points — prose only"]
- [e.g. "Under 150 words"]
Format: [exactly how you want the output structured]Three finished examples to steal
Example 1: The email rewrite
You are a senior communications director who writes for Fortune 500 executives.
Context: I need to follow up with a potential client who went quiet after our demo two weeks ago. They were interested but mentioned a budget freeze. I want to re-open the conversation without being pushy.
Task: Write a follow-up email.
Rules:
- Under 100 words
- No "just checking in" or "circling back"
- Don't mention the budget freeze unless they bring it up
- End with one specific, easy-to-answer question
- Tone: warm but directExample 2: The research request
You are a research analyst with expertise in consumer behavior.
Context: I'm preparing a presentation for leadership on AI adoption in mid-market HR software (50-500 employees). The audience is non-technical but data-driven.
Task: Summarize key trends in enterprise AI adoption for HR functions in 2025-2026.
Rules:
- Focus only on verified trends, not speculation
- Flag any claims you're uncertain about
- No academic jargon
Format: 5 bullet points, each with a headline and 2-sentence explanation. End with one "so what" implication for a software company.Example 3: The decision framework
You are a strategic advisor helping a startup founder make clear-headed decisions.
Context: I'm deciding whether to hire a full-time head of marketing or continue using contractors. We have 12 employees, $2M ARR, growing 8% month-over-month, and about 18 months of runway.
Task: Help me think through this decision clearly.
Rules:
- Do NOT tell me what to do — present both sides
- Surface the assumption I'm most likely wrong about
- Be direct if one option seems clearly better given my constraints
Format: Three sections: "Case for hiring full-time," "Case for staying with contractors," and "The one thing that should decide this." For teams
Want us to build this for your whole team?
We run AI workshops and full implementations for teams of 5–200. Custom prompts, workflows, and training built around your actual work.