Skip to main content
Build a knowledge base that helps your AI agents to deliver accurate, helpful responses. This guide covers content strategy, organization, and maintenance.

Why Knowledge Quality Matters

Your AI agents are only as good as the knowledge they have access to:

Content Strategy

Document Types to Include

Document Types to Avoid


Writing for AI Retrieval

Clear Titles

The title is crucial for retrieval: Weak:
Strong:

Answer-First Structure

Put the answer at the beginning: Weak:
Strong:

Self-Contained Sections

Each section should make sense on its own: Weak:
Strong:

Organization

Hierarchical Structure

Organize documents logically:

Naming Conventions

Use consistent naming:

Tagging Strategy

Apply tags consistently:
Tags improve retrieval for related queries.

Content Quality

Quality Checklist

For each document:
  • Title accurately describes content
  • Content answers a specific question
  • Information is current and accurate
  • Steps are clear and complete
  • No broken links or references
  • Consistent formatting throughout

Using Quality Scoring

Enable AI quality analysis:
  1. Navigate to Knowledge → Quality
  2. Select documents to analyze
  3. Review scores and suggestions
  4. Address issues identified

Quality Score Interpretation


Improvement Techniques

Keyword Optimization

Include terms users actually search for: Before:
After:

Question-Answer Format

Structure documents as Q&A:

Chunking for Retrieval

Break large documents into logical sections: Before: One 5,000-word document After:
  • Overview (500 words)
  • Getting Started (800 words)
  • Configuration (1,000 words)
  • Advanced Usage (1,200 words)
  • Troubleshooting (1,500 words)

Multi-Language Support

Translation Strategy

Translation Quality

  • Use professional translation for critical content
  • AI translation works for straightforward content
  • Always have native speakers review

Managing Translations

Keep translations in sync:
  1. Mark source language documents
  2. Track translation status
  3. Flag when source changes
  4. Re-translate updated content

Source Management

Connector Configuration

For each external source:

Manual Sources

For content not in external systems:
  1. Create dedicated manual source
  2. Use consistent formatting
  3. Include metadata (author, date)
  4. Review quarterly

Source Prioritization

When documents overlap, set priority:
  1. Primary: Official documentation
  2. Secondary: FAQs and guides
  3. Tertiary: Blog posts, announcements

Maintenance Schedule

Daily

  • Review sync logs for errors
  • Check for failed imports
  • Monitor retrieval quality

Weekly

  • Review low-quality score documents
  • Update documents based on support tickets
  • Check for duplicate content

Monthly

  • Audit document accuracy
  • Review analytics for coverage gaps
  • Update seasonal/time-sensitive content

Quarterly

  • Full knowledge base review
  • Remove outdated content
  • Assess translation needs
  • Evaluate source effectiveness

Common Issues

Low Retrieval Quality

Symptoms: Agent gives irrelevant answers or says “I don’t know” when information exists. Solutions:
  1. Improve document titles
  2. Add more keyword variations
  3. Check embedding model configuration
  4. Enable reranker
  5. Review similarity thresholds

Duplicate/Conflicting Information

Symptoms: Agent provides inconsistent answers to the same question. Solutions:
  1. Consolidate duplicate documents
  2. Set source priority
  3. Remove outdated versions
  4. Add canonical markers

Missing Coverage

Symptoms: Users ask questions with no matching documents. Solutions:
  1. Review “no answer” conversations
  2. Analyze common escalation topics
  3. Create content for gaps
  4. Check connector include patterns

Slow Sync

Symptoms: Changes in source don’t appear in AGO. Solutions:
  1. Check sync schedule
  2. Verify credentials
  3. Review sync logs for errors
  4. Consider more frequent sync

Analytics Usage

Coverage Analysis

Review which topics are covered:
Focus on reducing “Unknown” by filling gaps.

Top Retrieval Misses

Documents frequently searched but not found:
  1. Export retrieval analytics
  2. Identify patterns in failed searches
  3. Create or improve matching documents
  4. Re-test with same queries

Document Usage

Which documents are most helpful:

Content Governance

Change Management Workflow

Use Change Proposals to manage content changes systematically: Best Practices:

Types of Proposals

Knowledge Gap Tracking

Use Update Requests to track missing content:
  1. Monitor gaps: Review pending requests regularly
  2. Prioritize by volume: Multiple requests = critical gap
  3. Link to conversations: Understand user context
  4. Create content: Fill gaps with targeted documentation
  5. Close the loop: Mark resolved with notes
Triage Strategy:

Quick Wins

Immediate improvements you can make:
  1. Fix titles - Make them question-based
  2. Add FAQs - Top 10 support questions
  3. Remove duplicates - Consolidate content
  4. Update dates - Remove outdated info
  5. Enable quality analysis - Find problem areas