Guides

Best practices

Effectively organizing and creating high-quality knowledge content is critical for your AI Agent's success. This section provides comprehensive guidance on structuring your knowledge base and crafting individual policies.

Best Practices for Knowledge Structure

Knowledge Architecture

Before creating individual policies, map out the knowledge areas your customers commonly inquire about. The optimal structure depends on your industry and business model:

Example: E-commerce Knowledge Structure

If you operate in e-commerce, you might organize your policies like this:

  • Transaction-related policies
    • How to place an order
    • How to apply gift cards or vouchers
    • Refund policy
    • Invoicing
  • Product-related policies
    • Return policy
    • Shipping and delivery methods
  • Account-related policies
    • How to issue a complaint
    • GDPR and Privacy Policy details
    • How to change account details

Example: SaaS Knowledge Structure

For software-as-a-service companies, a different structure might be more appropriate:

  • Account Management
    • Subscription plans and billing
    • User roles and permissions
    • Account security
  • Product Usage
    • Getting started guides
    • Feature explanations
    • Integration instructions
  • Technical Support
    • Troubleshooting common issues
    • System requirements
    • Maintenance and updates

Create a structure that mirrors how your customers think about and search for information related to your products or services.

Knowledge Organization

Zowie provides several tools to help you organize your knowledge base:

  • Categories: Create logical groupings for related policies (e.g., "Payments," "Returns," "Shipping")
  • Tags: Add descriptive labels to make filtering easier and more intuitive
  • Segments: Direct specific policies to relevant customer groups (e.g., "Premium Members," "New Customers")
  • Regions: Customize knowledge for different geographic locations with unique requirements

Activation Strategy

For optimal knowledge management:

  • Start small: Begin with high-quality policies for your most common customer questions
  • Implement access controls: Assign appropriate permissions to maintain quality
  • Activate strategically: Toggle policies active only after thorough review
  • Segment appropriately: Ensure customers only see information relevant to them
  • Track changes: Use the change history log for accountability and troubleshooting

Best Practices for Policy Content

Focus on Single Topics

Each policy should address one clearly defined topic:

  • Create separate policies for distinct but related issues
  • Avoid combining multiple topics within a single policy
  • Split complex topics into multiple linked policies

Good Practice: Create separate policies for "Return Policy," "Shipping Policy," and "Loyalty Program"

Poor Practice: Combining shipping details with loyalty program information in one policy

Titles

The policy title significantly impacts retrieval accuracy:

  • Mirror customer language: Use phrasing that matches how customers ask questions
  • Be specific: "How do I return an item?" is better than "Returns Process"
  • Use question format when appropriate for common inquiries

Structure of Policy

Format your policies for optimal AI understanding:

  • Use simple formatting: Stick to basic headings, paragraphs, and bullet points
  • Avoid tables, images, or complex formatting: Use plain text for seamless integration
  • Create clear hierarchies: Use headings to organize information logically
  • Break down processes: Use numbered steps for procedures

Important Limitation: Zowie AI Agent cannot extract or interpret information from tables or images in your knowledge content. All critical information must be provided as plain text. If you have tabular data or information in images, convert it to text-based alternatives like bullet points or numbered lists.

Write Clear, Explicit Content

The AI Agent cannot infer details beyond what is explicitly stated:

  • Be comprehensive: Include all relevant details, exceptions, and conditions
  • Avoid assumptions: State all conditions clearly rather than implying them
  • Include complete information: If you state "We ship within the United States," explicitly address international shipping in a separate policy

Sample Policy Structure

📘

How do I return an item?

Preparation

Return your item using the bag or box it was delivered in. If you no longer have the original packaging, you can use any other bag or box of your choosing.

Packing

Stick the prepaid return label on the parcel, and make sure that it covers the original shipping label. Seal your return parcel securely so that it doesn't get damaged on its way back.

Return Process

You can return your parcel through your local post office. Please note that we do not offer collection services. Remember to ask for a receipt as proof that you have returned the parcel and keep it until we have processed your return.

Refund Timeline

It may take up to 14 business days for the item to get back to our warehouse. Once we've received your parcel in the warehouse, we will send you a return confirmation email. Once your return has been processed, we will issue a refund to the payment method that you chose when you made your order. Please note that it can take several days before the money is back in your bank account, depending on the terms and policies of your bank.

Handling Conditional Content

For policies that vary based on customer segments, regions, or other factors:

  • Clearly define conditions: "If you purchased after January 1, 2025..."
  • Create segment-specific policies when variations are substantial
  • Address regional differences: Specify which regions policies apply to
  • Include product-specific details: Clarify which products or services the policy covers

External vs Internal Documentation

Unlike customer-facing knowledge bases that often provide simplified information, your AI Agent benefits from having access to more comprehensive information:

Leveraging Internal Resources

  • Include Agent-Only Documents: Provide your AI Agent with the same detailed documentation your human agents use
  • Add Process Details: Include internal processes, exceptions, and edge cases that might not be in customer-facing materials
  • Incorporate Product Information: Give your AI Agent access to detailed product specifications and compatibility information
  • Document Business Rules: Include internal business rules and decision-making criteria

Benefits of Detailed Knowledge

  • Improved Response Accuracy: More detailed knowledge allows the AI to provide precise, thorough answers
  • Reduced Human Escalations: AI can handle more complex queries without transferring to human agents
  • Consistent Decision-Making: AI follows the same guidelines as your human agents

What to Include vs. Exclude

Include:

  • Detailed troubleshooting procedures
  • Internal process documentation
  • Agent guidelines and scripts
  • Complete product information

Exclude:

  • Confidential business information
  • Personally identifiable information (PII)
  • Security-sensitive details
  • Information that could be misused
  • Outdated processes or policies

Best Practice: Review the knowledge materials available to your human agents and make these available to your AI Agent, excluding only confidential or sensitive information. Generally, your AI Agent should have access to the same or more detailed information than what's provided to your customers.

Content Quality Checklist

Before activating a policy, ensure it:
✓ Focuses on a single, clearly defined topic
✓ Has a descriptive title that matches customer language
✓ Provides complete information without requiring assumptions
✓ Uses simple, clear formatting
✓ Includes all relevant conditions and exceptions
✓ Anticipates logical follow-up questions
✓ Specifies timelines when relevant
✓ Uses consistent terminology
✓ Avoids unnecessary technical jargon
✓ Is assigned to appropriate segments and regions