Creating Custom Instructions

This is a guide on how to create optimum prompts for your AI in the Custom Instructions Section

What are Custom Instructions?

The Custom Instructions section provides clear, structured guidelines to ensure responses are consistent, professional, and easy to understand. It helps both team members and AI tools respond accurately across different scenarios—whether users are asking for data, pricing, or technical details. By following these instructions, teams can save time, maintain brand voice, handle edge cases smoothly, and onboard new members more effectively. Overall, this section acts as a reliable reference for delivering high-quality, user-friendly interactions at scale.

---

General Guidelines

1. Be Specific: Provide exact parameters (word counts, formatting rules)

2. Use Examples: Include sample responses for clarity

3. Structure Logically: Group related instructions by scenario

4. Use Clear Directives: Write instructions as direct commands

5. Define Fallbacks: Include handling for edge cases

---

Quick Summary Table: Responding to Common Scenarios

Use Case	What to Include	Sample Phrase
Pricing Inquiry	3-tier overview + contact info	"Our plans range from $29 to $199 per month depending on your needs. For a custom quote, contact [e-mail]."
Data Request	Ask for missing info (product/location), summarize key stats, share source link	"[Product] data shows strong usage in India (42%), US (33%), and UK (12%). Full report: https://example.com/data/productX"
Feature Clarification	One-sentence summary, 3 key benefits, link to learn more	"This feature helps you automate outreach. Benefits: Save time, personalize messages, increase leads. Learn more: [Link]"
Multiple Meanings	40-word max overview, list 2-3 interpretations, ask for clarification	"That could mean A, B, or C. Could you clarify which you're referring to?"
Technical Question	Simple explanation, 3 key points, optional diagram or screenshot	"Pagination splits large data into pages. Benefits: faster load, better UX, lower errors."
Fallback/Not Sure	Acknowledge limitation, offer nearest alternative, suggest contact	"I don’t have specific data on that, but here's a related report. For custom insights, contact [e-mail]."
Emoji/Reaction	Acknowledge casually with short confirmation or encouragement	"Glad that helps! 👍 Need anything else?"
Scope Limitation	State what the service does not provide	"We provide chatbot integration guidance but not direct legal support."

---

🎉 Special Response Rules:

Whenever you are trying to train your bot on giving answers in a particular style or form, be sure to be specific!

✅ Good:

For pricing inquiries: "Our plans range from $X to $Y depending on requirements. For a personalized quote, contact [e-mail]."

For emoji reactions like 👍: "I'm glad that helps! Need anything else?"

For service clarification: "We provide data analytics only, not direct [service] assistance."

❌ Bad:

Tell users to email us for pricing. Be friendly with emojis.

---

💻 Data Requests:

When users ask for data, first confirm the required details (like product or location). Then, provide a short summary, top statistics, and a direct source link for quick reference.

✅ Good:

When users request specific data:
1. Prompt for product if missing
2. Prompt for location if missing
3. Present data with: summary, top 3 statistics, direct link

Example: "[Product] data shows [key stat]. Top sources: [Country A] (X%), [Country B] (Y%).
Link: https://example.com/data/[product]"

❌ Bad:

Ask users what they want. Give them data with links.

---

🎯 Multiple-Answer Queries

For questions with multiple possible meanings, start with a brief overview, list 2–3 likely interpretations, and ask the user to clarify which one they mean.

✅ Good:

For general queries:
1. Brief overview (max 40 words)
2. List 2-3 most relevant options
3. Ask: "Could you clarify if you're interested in [Option A], [Option B]?"

❌ Bad:

Answer all possible interpretations and ask them to clarify.

---

🤖 Technical Queries

For technical questions, provide a simple one-line explanation, highlight three key benefits or points, and include a link or resource if available. Always end by relating it to the user’s use case.

✅ Good:

For feature questions:
1. One-sentence overview
2. Three key benefits as bullets
3. Direct link to details page
4. End with use case question

❌ Bad:

Tell them about features and share a link.

---

🔢 Data Presentation Rules

Always present data in a clear, consistent format—include units, round large numbers, use percentages for comparisons, and bold key figures for quick visibility.

✅ Good:

Always include units of measurement
Round large numbers appropriately
Use percentages for comparisons
Bold key statistics using **markdown**

❌ Bad:

Include numbers in answers.

---

🍂 Fallback Mechanisms

When uncertain, acknowledge the limitation, offer the closest available alternative, provide a contact option for further assistance, and suggest a booking link for custom insights.

✅ Good:

When uncertain:
1. Acknowledge limitation: "I don't have specific data on [topic]"
2. Offer closest alternative information
3. Provide booking option: "For custom insights: [booking link]"
4. Suggest contact: "For assistance: [e-mail]"

❌ Bad:

If you don't know, suggest they contact support.

---

♻️ Custom Prompt as an Iterative Process

Creating effective prompts is an ongoing process that requires refinement. Start with an initial approach, gather feedback, and gradually improve the instructions to ensure consistency and adaptability across all scenarios.

🔄 Iteration #1: Initial Attempt

When users ask about pricing, tell them our plans and give contact information.

❌ Issue: Too vague → inconsistent responses with varying details.

🔄 Iteration #2: Improved Attempt

When users ask about pricing, respond with: "We offer Basic ($X/month), Pro ($Y/month), and Enterprise plans. Contact [e-mail] for details."

⚠️ Issue: Limited flexibility → couldn't adapt to specific plan inquiries.

🔄 Iteration #3: Refined Solution

When users ask about pricing:
1. For general queries: Provide 3-tier overview with contact info
2. For specific plan questions: Show only requested plan details
3. For feature pricing: Mention which plans include the feature
4. Always end with contact suggestion

✅ Result: Consistent yet adaptable responses across pricing scenarios.

---

🗣️ Response Format & Style

Keep responses under 80 words, use a professional tone, and format multi-point answers as bullet lists with appropriate spacing. Always use "we" and "our" for company references, and bold key statistics or terms to highlight important information.

✅ Good:

- Keep responses under 80 words. 
- Use professional tone. Format multi-point responses as bullet lists with line spacing. 
- Use 'we' and 'our' for company references. 
- Bold key statistics.

❌ Bad:

Make responses short. Be professional. Use lists sometimes.

Why it's bad: Too vague, lacks specific formatting instructions and measurable parameters.

---

If everything is correct and you are still unable to move forward, write to us at [e-mail]. We will respond back within 48 business hours.