Best Practices & FAQ

Tips, testing strategies, and answers to common questions about configuring your persona’s prompts.

Best Practices

1. Writing Effective Introductions

Do:

Include specific years of experience and quantifiable achievements
Mention industries, company types, and deal/project sizes
Highlight unique expertise or specialization
Use natural, conversational language

Don’t:

Use generic descriptions (“experienced professional”)
Overuse buzzwords without substance (“synergistic thought leader”)
Make claims you can’t support with your knowledge sources
Write more than 250 words

Good Example:

“Marcus Lee is a startup operator turned VC investor with 8 years of experience building and scaling early-stage companies. He was employee #3 at TaskFlow (acquired by Salesforce for $150M) and later joined Elevation Partners where he led seed investments in 20+ B2B SaaS startups. Marcus specializes in helping founders navigate product-market fit, go-to-market strategy, and fundraising from pre-seed through Series A.”

Bad Example:

“Marcus is a passionate and experienced professional in the startup ecosystem with expertise in various areas including product, sales, and operations.”

2. Crafting Clear Chat Objectives

Do:

Be specific about WHO you help and WITH WHAT
Focus on OUTCOMES, not activities
Align with your actual expertise
Keep to 1-2 sentences (30-60 words)

Don’t:

Use vague goals like “provide helpful advice”
List everything you could possibly do
Make promises beyond your expertise

Good Example:

“Help early-stage B2B SaaS founders validate product-market fit, design effective go-to-market strategies, and prepare for seed/Series A fundraising with practical, operator-tested advice.”

Bad Example:

“Provide expert-level insights and guidance on startups, technology, business strategy, marketing, sales, operations, and anything else entrepreneurs might need help with.”

3. Defining Your Thinking Style

Do:

Describe HOW you approach problems (frameworks, mental models you actually use)
Specify tone and communication preferences (direct vs diplomatic, data-driven vs intuitive)
Include WHAT QUESTIONS you ask before giving advice
Mention how you balance competing priorities

Don’t:

Just list personality traits (“friendly, helpful, knowledgeable”)
Describe generic professional behavior
Forget to mention how you validate understanding before advising

Good Example:

“Elena is pragmatic and ROI-focused. She approaches marketing challenges using the Pirate Metrics (AARRR) framework and always starts by asking about current funnel metrics, customer acquisition cost, and conversion rates. She balances brand-building with performance marketing, prioritizing whatever moves the revenue needle fastest. Her tone is direct and actionable - less theory, more ‘here’s what worked for me and why it will work for you.’”

Bad Example:

“Elena is a strategic and analytical thinker who is passionate about marketing. She is friendly, professional, and always eager to help.”

4. Creating Example Conversations

Do:

Write realistic multi-turn dialogues (3-4 exchanges minimum per conversation)
Show your diagnostic questions and thinking process
Include specific details, numbers, and frameworks you’d actually use
Demonstrate how you handle vague or unclear questions
Cover 2-3 different scenarios (beginner, intermediate, specific problems)

Don’t:

Write single-turn Q&A pairs
Make responses too long or too short (2-4 sentences each is ideal)
Use perfect, unrealistic user questions
Skip the back-and-forth that reveals your methodology

Why this matters: Example conversations teach your persona HOW to interact, not just WHAT to say.

5. Writing Example Response Patterns

Do:

Label distinct patterns clearly (e.g., “Pattern 1: Data-Driven Recommendations”)
Show each pattern with 2-3 concrete quote examples
Highlight what makes each pattern unique to YOUR style
Cover patterns for different situations (diagnosis, recommendation, handling objections)

Don’t:

Just describe the pattern in abstract terms
Forget to show actual example phrases/sentences
Make patterns too similar to each other
Skip patterns you actually use frequently

6. Field Length Guidelines

Field	Minimum	Optimal	Maximum
Introduction	100 words	150-200 words	250 words
Area of Expertise	5 items	8-10 items	15 items
Chat Objective	20 words	40-50 words	75 words
Target Audience	2 types	3-4 types	6 types
Thinking Style	50 words	100-125 words	175 words
Conversation Flow	4 steps	5-6 steps	8 steps
Example Conversations	2 dialogues	3 dialogues	5 dialogues
Example Patterns	3 patterns	4-5 patterns	7 patterns
Guardrails	3 rules	5-7 rules	10 rules

7. Testing Your Configuration

Initial Test Checklist:

Identity is Clear: Can someone understand who this persona is from reading the introduction?
Objective is Specific: Is the chat objective focused on a specific audience and outcome?
Style is Differentiated: Does the thinking style clearly differentiate this persona from generic AI?
Examples are Realistic: Do the example conversations sound like real interactions?
Patterns are Observable: Are the response patterns clear and repeatable?
Flow Makes Sense: Does the conversation flow logically progress from discovery to action?
All Fields Have Content: Have you filled out at least the essential fields with quality content?

Testing Process:

Save and Test: Fill out essential fields and click Save
Start a Chat: Go to the test widget or start a conversation with your persona
Ask 5-10 Questions: Cover basic questions, intermediate problems, and edge cases
Evaluate Quality: Does it sound like you? Is it following the conversation flow?
Refine: Go back to Prompt Configuration and update specific fields based on gaps
Re-test: Try the same questions again to verify improvements
Iterate: Repeat until responses consistently match your desired personality and quality

Common Questions

Q: Do I need to fill out ALL fields?

A: No. The essential fields are:

How should your persona introduce itself?
What topics can your persona help with?
What should conversations achieve?

The optional fields (Communication Style and Examples & Guardrails sections) improve quality but aren’t required to get started.

Q: Can I edit these fields after my persona is live?

A: Yes! You can update your prompt configuration at any time. Changes apply to NEW conversations immediately. Existing conversations continue using the old configuration.

Q: How long does it take to fill this out?

Quick start (essential fields only): 15-30 minutes
Complete configuration (all fields): 1-2 hours
Professional-quality (with examples and iteration): 2-4 hours

The time investment pays off in better conversation quality and fewer support issues.

Q: What if I’m not a writer?

A: You don’t need to be! Follow these tips:

Use the templates provided for each field
Start with bullet points then convert to sentences
Record yourself answering questions, then transcribe
Use the professional examples as inspiration
Iterate: Start simple, improve over time based on real conversations

Q: Can I use AI to help write these fields?

A: Yes, but be careful:

✅ Good use: Use AI to structure your thoughts or rephrase bullet points
✅ Good use: Ask AI to format your conversation flow into numbered steps
❌ Bad use: Let AI generate generic content without your specific expertise
❌ Bad use: Copy example conversations from other personas

Remember: Your persona represents YOU. The content should reflect your actual expertise and style.

Q: How do I know if my configuration is working?

A: Look for these signs:

✅ Users say “this feels like talking to a real person”
✅ The persona asks relevant diagnostic questions
✅ Responses match your actual expertise level and style
✅ The persona admits when something is outside its expertise
❌ Responses are too generic or robotic
❌ The persona makes claims not backed by your knowledge sources
❌ Users are confused about what the persona can help with

Q: Should I update my configuration over time?

A: Yes! Your configuration should evolve as:

You learn what users actually ask about
You identify gaps in your persona’s responses
Your expertise or methodology evolves
You add new knowledge sources

Review your configuration every 3-6 months or after major changes to your knowledge base.

Q: What’s the difference between “Example Conversations” and “Response Patterns”?

A: These two fields are often confused - here’s the key difference:

Example Conversations (Goals & Audience):

Complete multi-turn dialogues (3-4 exchanges per conversation)
Shows how conversations flow from start to finish
Demonstrates diagnostic questions and back-and-forth
Example: Full conversation showing client qualification → questions → tailored advice
Length: 100-200 words per conversation
Count: 2-4 conversations

Response Patterns (Examples & Guardrails):

Labeled techniques with short example quotes
Shows specific communication patterns in isolation
Highlights distinctive elements of your style
Example: “Pattern 1: Always cite data” with 2-3 short quotes
Length: 1-2 sentences per quote
Count: 3-5 patterns

Think of it this way:

Conversations = showing the full movie
Patterns = highlighting specific scenes and techniques

Together: The LLM learns both structure (from conversations) and style (from patterns).

For a detailed comparison with examples, see: Understanding Example Conversations vs Response Patterns

Q: Can I see my persona’s actual prompts being used?

A: The system uses your configuration to build prompts dynamically, but the exact prompts aren’t shown to users. What you configure here shapes how your persona thinks and responds behind the scenes.

Q: My persona isn’t following my conversation flow. What’s wrong?

A: Common causes:

Flow is too rigid - Allow flexibility, don’t force every conversation to follow exactly
Missing example conversations - Add 2-3 examples showing the flow in action
Thinking style conflicts - Ensure your thinking style aligns with your conversation flow
Not enough testing - Try more varied questions to see where it breaks down

Fix: Update your conversation flow to be more flexible, add supporting example conversations, and test again.

Q: Users say my persona is too generic. How do I fix this?

A: Add more specificity:

Thinking Style - Describe your actual methodology with frameworks you use
Example Patterns - Show specific language and phrasing you repeatedly use
Example Conversations - Include your diagnostic questions and thought process
Expertise - Be more specific about what you can help with

Test: Ask “What makes you different from other [your profession]?” - Does the response capture your uniqueness?

Q: Can I have multiple personas with different configurations?

A: Yes! Each persona has its own independent prompt configuration. Common use cases:

Support vs Sales personas - Different objectives and tones
Beginner vs Expert personas - Different language complexity
Different service offerings - Each with specific expertise

Q: How do I handle topics outside my expertise?

A: Use the Guardrails field to set boundaries:

**Always:**
- Admit when a question is outside your expertise area
- Recommend consulting appropriate experts (lawyers, doctors, etc.)
- Suggest related topics within your expertise if relevant

Your persona will learn to gracefully decline and redirect.

Need Help?

Resources:

Video Tutorial: Configuring Your First Persona (Coming soon)
Professional Examples - Complete configurations
Community Forum: Prompt Configuration Tips
Schedule a Setup Call with Our Team

Support:

Live Chat: Available in the dashboard (bottom-right corner)

Document Version: 1.0 Last Updated: January 2025 Maintained By: MyClone Team

Ready to configure your persona? Go to Dashboard →

Previous: Professional Examples ← Back to Overview: Prompt Configuration ↑