Prompt Writing Guide
The Golden Rule: Shit IN = Shit OUT
The single most important thing to understand about AI:
If you give it vague, lazy prompts → You get generic, useless results
If you give it detailed, thoughtful prompts → You get excellent, tailored results
It's that simple. The quality of what you put in determines the quality of what you get out.
Think of AI like a brilliant assistant who:
- Knows everything but nothing about YOUR specific situation
- Can do anything but needs clear instructions
- Works fast but can't read your mind
Your job: Provide the context, details, and specifics that make the output useful.
The Golden Formula
[Role/Context] + [Task] + [Specifics] + [Format/Tone] = Effective Prompt
Breaking It Down
Role/Context: Who you are and the situation
- "I'm a sales manager preparing a quarterly review."
- "I'm planning a team offsite meeting for 15 people."
Task: What you want AI to do
- "Write an email..."
- "Analyse this data..."
- "Create a list of..."
- "Summarise this document..."
Specifics: Details that matter
- Length (words, paragraphs)
- Key points to include
- What to avoid
- Target audience
Format/Tone: How it should be delivered
- Tone: professional, warm, formal, conversational
- Format: bullet points, paragraph, table, email
- Style: concise, detailed, creative, analytical
Examples: Bad vs Good
Example 1: Email Writing
❌ Bad Prompt:
Write an email about the project
✅ Good Prompt:
I'm a project manager. Write a 150-word email to my team announcing that we're launching the new client portal next Monday.
Include:
- Go-live date and time (Monday 9am)
- What's changing for them
- Where to find training materials
- Who to contact with questions
Tone should be clear and encouraging (not overwhelming). Make them feel prepared.
Why it's better: Clear context, specific details, defined length, tone guidance.
Example 2: Data Analysis
❌ Bad Prompt:
Look at this data
✅ Good Prompt:
I'm a sales director. Analyse this Q1 sales data and:
1. Identify top 5 performing products
2. Highlight any concerning trends
3. Compare performance to same period last year
4. Suggest 3 actionable insights for Q2 planning
Format as an executive summary (max 300 words) with bullet points for key findings.
[paste data]
Why it's better: Specific analytical tasks, clear output format, actionable focus.
Example 3: Content Creation
❌ Bad Prompt:
Write a social post
✅ Good Prompt:
I'm a marketing coordinator. Write a LinkedIn post (200 words max) about our company's new remote work policy.
Key points:
- Hybrid model: 3 days in office, 2 days remote
- Employees choose which days
- Starting next month
- Focus on work-life balance and productivity
Tone: Professional but approachable. Emphasis on employee wellbeing and trust. Include 3 relevant hashtags.
Why it's better: Platform specified, word limit, key messages, tone, format requirements.
The Six Principles of Great Prompts
1. Be Specific
Vague: "Help me with marketing" Specific: "Create 5 email subject lines for our Black Friday sale targeting customers who haven't purchased in 6 months"
2. Provide Context
No context: "Write a report" With context: "I'm presenting to the executive team about customer satisfaction. Write a 1-page summary of these survey results highlighting key trends and recommendations"
3. Define the Output
Undefined: "Analyse this" Defined: "Create a table with 3 columns: Finding, Impact, Recommendation"
4. Set Constraints
Unconstrained: "Write an email" Constrained: "Write a 100-word email. Keep it under 3 paragraphs. Don't mention pricing."
5. Specify Tone
No tone: "Respond to this complaint" With tone: "Respond with empathy and professionalism, acknowledging the frustration without being defensive"
6. Use Examples (When Helpful)
Write 3 product descriptions in this style:
EXAMPLE:
"The Pro Desk Chair - Engineered for all-day comfort with ergonomic lumbar support and breathable mesh. Adjustable height and arms. £299"
Now write for: [your products]
Advanced Techniques
Technique 1: Chain of Thought
Ask AI to think step-by-step.
Before giving your final answer, think through:
1. What are the key customer needs?
2. What makes this solution unique?
3. How does it compare to alternatives?
Then write the proposal.
Technique 2: Role Playing
Make AI adopt a specific perspective.
You are an expert business consultant with 20 years experience. Review this strategy and identify potential blind spots.
Technique 3: Multiple Options
Ask for variations to choose from.
Give me 5 different versions of this headline, ranging from conservative to bold.
Technique 4: Iterative Refinement
Build on previous responses.
Prompt 1: "Draft an email about..."
Prompt 2: "Make it 30% shorter"
Prompt 3: "Add more emphasis on urgency"
Prompt 4: "Perfect. Now create a subject line."
Technique 5: Templates
Create reusable prompt structures.
I need a [type of content] for [audience] about [topic].
Key messages:
- [point 1]
- [point 2]
- [point 3]
Tone: [tone]
Length: [length]
Format: [format]
The Voice Method: The Easiest Way to Prompt
Instead of typing, press the Voice Mode button in Claude (the microphone icon) and simply speak to it like a colleague.
Speaking is faster, more natural, and helps you convey nuance and context that's harder to type. You can even use the "full-circle method" to design better prompts by speaking.
Example:
"I am trying to create a proposal for a new client — can you help me write a prompt for that?"
The AI will then generate a clear, structured prompt for you. You can refine it together through conversation before running the final version.
Why it works: Speaking lets you think aloud, express tone and emotion, and course-correct mid-sentence. It's like brainstorming with a creative partner instead of typing instructions.
Try it — your prompts will improve instantly.
Common Prompt Patterns
Pattern 1: Summarisation
Summarise this [document type] in [length]:
- Focus on [aspect]
- Highlight any [specific elements]
- Format as [structure]
[paste content]
Pattern 2: Brainstorming
I need ideas for [goal/project].
Context:
- Audience: [who]
- Budget: [constraint]
- Timeline: [when]
- Must include: [requirements]
Give me 10 creative ideas, ordered from safest to most innovative.
Pattern 3: Analysis
Analyse this [data/text] and:
1. [specific question]
2. [specific question]
3. [specific question]
Present findings as [format] with [structure].
Pattern 4: Transformation
Transform this [input type] into [output type]:
Requirements:
- [constraint 1]
- [constraint 2]
- [constraint 3]
[paste input]
Pattern 5: Comparison
Compare [A] and [B] across these dimensions:
- [dimension 1]
- [dimension 2]
- [dimension 3]
Format as a table with columns for each option and a final "Recommendation" column.
Troubleshooting: When Results Are Bad
Problem: Output is too generic
Solution: Add more specific details and context
Problem: Wrong tone
Solution: Be explicit: "Use a warm, conversational tone" or "Keep it formal and professional"
Problem: Too long/short
Solution: Specify exact length: "100 words" or "3 bullet points" or "1 paragraph"
Problem: Missing key information
Solution: List must-include points explicitly
Problem: Factually incorrect
Solution: AI isn't reliable for facts. Always verify or provide the facts yourself
Problem: Doesn't sound like "me"
Solution: Provide examples of your style or specific phrases to use/avoid
The Iteration Mindset
Remember: First outputs are drafts, not finals.
Iteration examples:
- "Make this more concise"
- "Use simpler language"
- "Add more detail about [topic]"
- "Remove the section about [topic]"
- "Make it sound more [adjective]"
- "Rewrite the opening to be more engaging"
- "Give me 3 alternative versions"
Quick Reference Card
Before every prompt, ask yourself:
- ✅ Have I explained who I am and why this matters?
- ✅ Is my task crystal clear?
- ✅ Have I specified length/format?
- ✅ Have I defined the tone?
- ✅ Have I mentioned any constraints?
- ✅ Will AI have all the information it needs?
Remember: Great prompting is a skill that improves with practice. Shit in = Shit out: More detail and context = Better results.