Training your AI Knowledge Agent

Your AI Knowledge Agent is only as good as the material you train it with. This guide will help you understand what types of files to upload to maximize your AI agent's effectiveness in understanding your business and products.

Recommended File Types

The AI Knowledge Agent can process various file formats, but some contain richer information than others:

  • PDFs - Ideal for structured documentation, user guides, and technical specifications
  • Word Documents (.docx, .doc) - Great for internal documentation and policies
  • HTML Files - Perfect for existing web documentation
  • Markdown Files (.md) - Excellent for technical documentation
  • Plain Text Files (.txt) - Good for simple FAQs and notes
  • PowerPoint Files (.pptx, .ppt) - Useful for product presentations and feature overviews

Note: The system does not currently support image-only files or scanned documents without text. For best results, ensure your PDFs contain actual text content rather than just scanned images.

Content That Trains Your AI Best

To maximize the effectiveness of your AI Knowledge Agent, prioritize uploading these types of content:

  • Product Documentation - User manuals, technical specifications, and API documentation
  • FAQs - Frequently asked questions and their answers from your customers
  • Troubleshooting Guides - Step-by-step instructions for resolving common issues
  • Feature Guides - Detailed explanations of your product's features and capabilities
  • Training Materials - Onboarding materials that explain your product to new users
  • Knowledge Base Articles - Existing support articles that answer common questions
  • Release Notes - Information about product updates, new features, and bug fixes

Organizing Your Documentation

For the best results, prepare your documentation with these principles in mind:

  1. Be Comprehensive - Include detailed information about all aspects of your product
  2. Use Clear Language - Avoid jargon or highly technical terms without explanation
  3. Structure Logically - Organize content with clear headings, subheadings, and sections
  4. Include Examples - Real-world examples help the AI understand practical applications
  5. Update Regularly - Keep documentation current with your latest product changes

Best Practices for File Preparation

Before uploading your files, consider these best practices:

  • Split Large Documents - Rather than one massive file, consider breaking content into logical sections
  • Use Descriptive Filenames - Name files in a way that indicates their content, e.g., "product-installation-guide.pdf"
  • Include Metadata - Ensure documents have proper titles and metadata when possible
  • Remove Obsolete Information - Don't upload outdated documentation that might confuse the AI
  • Prioritize Quality Over Quantity - Well-structured, clear documents are better than numerous low-quality ones

Content to Avoid

Some types of content may not be helpful or appropriate for training your AI:

  • Highly confidential or sensitive internal information
  • Customer-specific data or personally identifiable information (PII)
  • Outdated documentation that no longer reflects your current product
  • Marketing materials with minimal technical details
  • Documents with primarily visual content and minimal text

Testing Your AI Knowledge Agent

After uploading your documentation, it's important to test how well the AI has understood your content:

  1. Ask common questions your customers typically have
  2. Try technical queries that require detailed knowledge of your product
  3. Test edge cases and specific scenarios your customers might encounter
  4. Review the AI's responses for accuracy and completeness

If you find gaps in the AI's knowledge, consider what additional documentation might help fill those gaps.

Pro Tip: For the best results, include both beginner-friendly content and detailed technical documentation. This gives your AI Knowledge Agent the ability to respond appropriately to users with varying levels of familiarity with your product.