Agent Best Practices
Tips and recommendations for creating effective AI agents in Popcorn
Agent Best Practices
Follow these best practices to create AI agents that provide excellent customer experiences and efficiently handle conversations.
Writing Effective Instructions
Your agent’s instructions are crucial for its performance. Here are best practices for writing clear, effective instructions:
Be Specific and Detailed
-
Provide comprehensive information about your business, products, and services
-
Explain your brand voice and communication style in detail
-
Include specific examples of good responses for common scenarios
-
Define boundaries clearly (what the agent should and shouldn’t do)
Structure Your Instructions
-
Use clear sections with headings for different topics
-
Start with the most important information
-
Include a brief “about the company” section
-
Add specific handling instructions for different customer situations
Example: Good vs. Basic Instructions
Basic Instruction:
Better Instruction:

Capability Configuration
Optimize how your agent uses its capabilities:
Knowledge Base Best Practices
-
Organize information in logical collections
-
Keep documents focused on single topics
-
Update your knowledge base regularly
-
Test queries to ensure your agent finds relevant information
Handover Configuration
-
Create clear criteria for when handovers should occur
-
Provide detailed handover instructions to your agent
-
Ensure your agent collects necessary information before handover
-
Configure proper notifications for your human agents

E-commerce Configuration
-
Ensure product data is up-to-date and complete
-
Configure recommendation preferences based on business goals
-
Test the checkout flow for a smooth customer experience
-
Set up order tracking to provide accurate, timely information
Guiding the Conversation
-
Have your agent ask clarifying questions when needed
-
Configure the agent to confirm understanding before proceeding
-
Use structured responses for complex information
Closing Conversations
-
Ensure the agent confirms all issues are resolved
-
Configure appropriate follow-up questions
-
Set up a professional closing message
Testing and Improvement
Continuously improve your agent’s performance:
Regular Testing
-
Test your agent with different conversation scenarios
-
Simulate common customer inquiries
-
Check how capabilities work in real conversations
-
Test edge cases and complex situations
Analyze Conversations
-
Review conversation transcripts regularly
-
Look for patterns in successful and unsuccessful interactions
-
Identify opportunities for improvement
-
Update instructions and capabilities based on insights
Iterative Improvement
-
Make incremental changes to your agent’s configuration
-
Test each change before making additional adjustments
-
Gather user feedback to guide improvements
-
Monitor key metrics like resolution rate and customer satisfaction
Security and Privacy
Maintain security and privacy in agent interactions:
-
Configure your agent to avoid requesting sensitive information
-
Set up proper data handling procedures
-
Ensure compliance with relevant regulations
-
Test security measures regularly
