AI Voice Agent Knowledge
The Knowledge tab controls what information your AI Voice Agent has access to when answering caller questions. This includes static knowledge bases and dynamic context fetched before each call.
Available on: All AI Voice Agent plans.
Knowledge Sources
Your agent draws on two types of knowledge:
Static Knowledge (Knowledge Bases)
Pre-loaded content that stays constant across calls:
| Source | What it provides |
|---|---|
| Web page URLs | Product pages, FAQs, pricing, policies — auto-refreshes every 24 hours |
| Files | PDFs, documents, spreadsheets — uploaded once, available permanently |
| Q&A pairs | Specific question-answer mappings — highest accuracy for known questions |
| Plain text | Free-form content — company information, procedures, scripts |
See Create a Knowledge Base for setup details and content limits.
Dynamic Knowledge (Fetch Context)
Real-time data retrieved before each call starts:
- Caller account details from your CRM or database
- Recent order status, open tickets, or pending actions
- Custom data from any API endpoint
See Fetch Context Before Every Call for setup.
How the Agent Uses Knowledge
When a caller asks a question, the agent:
- Searches the knowledge base for relevant content
- Checks fetched context for caller-specific information
- Synthesizes an answer combining both sources
- Cites sources — you can verify what content was used in the transcript (see KB Sources in Transcripts)
If no relevant knowledge is found, the agent follows its configured behavior — either acknowledging it does not have the information, offering to transfer to a human, or taking a message.
Knowledge Tab Configuration
On the Knowledge tab of your agent:
- Select a knowledge base — choose from existing knowledge bases or create a new one
- Enable Fetch Context — connect your API endpoint for dynamic data
- Set priority — when both static and dynamic knowledge exist, the agent prioritizes the most relevant source
Knowledge Base Limits
| Plan | Knowledge bases per agent |
|---|---|
| Trial | 1 |
| Pay-As-You-Go | 0 |
| Agent Lite | 1 |
| Agent Max | Up to 5 |
Best Practices
- Cover your top 20 questions first. Review recent call transcripts or support tickets to identify the most common questions. Build Q&A pairs for these.
- Keep content specific. Broad, vague content leads to broad, vague answers. Break large topics into focused entries.
- Update regularly. Product changes, pricing updates, and policy changes should be reflected in the knowledge base immediately.
- Use Q&A pairs for critical answers. When accuracy matters most (pricing, compliance, medical/legal), Q&A pairs give you the most control over the agent's response.
- Review source attribution. Regularly check transcripts to verify the agent is using the right sources.
Related Articles
- Create a Knowledge Base for AI Voice Agents — content types and limits
- Manage Knowledge Bases — editing and maintaining
- KB Sources in Transcripts — verifying source accuracy
- Fetch Context Before Every Call — dynamic pre-call data