Why it matters
Knowledge bases provide AI agents with a reliable source of truth. Instead of guessing or relying solely on prompt instructions, agents can reference your help articles, FAQs, guides, and documentation to give accurate and up-to-date answers.
A well-structured knowledge base ensures consistency across messaging and call interactions, reduces hallucinations, and keeps information aligned with how your product actually works.
Key Concepts
Knowledge base: A collection of pages containing approved information your AI agent is allowed to reference.
Pages: Individual, focused documents within a knowledge base that help the agent retrieve specific content quickly.
Knowledge selection: Choosing which knowledge bases an agent can access.
Document ingestion: The process where the system breaks uploaded content into smaller chunks for efficient retrieval.
Source of truth: The information the AI is permitted to rely on. The agent cannot use content outside the knowledge you attach.
Updates: Any changes to documents require updating or syncing the knowledge base to ensure accuracy.
Step-by-Step: Create and Manage Knowledge Bases
Open Agents from the sidebar.
Click the Knowledge tab to open the knowledge base manager.
Click Add knowledge base to create a new one, or select an existing knowledge base to edit.
Review the table to see the title, status, creator, and last updated date.
Click Edit knowledge to manage its pages.
Click Create page to add new content.
Enter a page name and page content in the text fields.
Use Edit page to update titles or content for existing pages.
Click the gear icon to edit the knowledge base’s name or description, or to delete it.
Save changes and attach the knowledge base to your AI agent from the agent editor.
Tips and Best Practices
Structuring High-Quality Pages
Keep each page focused on one clear topic.
Use short, direct sentences.
Start with a brief summary, followed by steps or explanations.
Add keywords people commonly search for such as “reset password” or “billing settings”.
Include examples to help ground the agent’s responses.
Avoid internal jargon unless clearly defined.
Keeping Knowledge Accurate and Useful
Update pages whenever workflows or product features change.
Break large, complex topics into multiple smaller pages for better retrieval accuracy.
Ensure titles are specific and unambiguous.
Review knowledge base usage in agent responses and refine content where misunderstandings appear.
Understanding How Knowledge Works for AI Agents
The agent retrieves chunks of relevant text at conversation time.
If content is not in the knowledge base, the agent cannot rely on it.
Knowledge bases act as a retrieval layer, not training material.
Attach only the knowledge bases that are relevant so responses stay focused.
Use fewer, more accurate sources rather than one large document full of mixed topics.
Organizing Large Knowledge Collections
Use separate knowledge bases for different teams, products, or workflows to avoid overlap.
Keep pages short, cleanly separated, and consistently formatted.
Add clear descriptions inside each knowledge base to help other team members understand its purpose.
Troubleshooting
Issue | Possible Cause | Fix |
Agent gives outdated information | Knowledge base not updated | Edit or replace outdated pages and save changes |
Agent references irrelevant content | Too many knowledge bases attached | Remove unnecessary knowledge bases from the agent |
Agent ignores knowledge | No knowledge bases selected or content unclear | Attach a knowledge base or rewrite ambiguous pages |
Conflicting answers | Overlapping or duplicated content | Consolidate pages and remove duplicated information |