By assigning specific documents from your Knowledge Base to a voice agent, you provide it with a curated source of information. When a user asks something the agent doesn't have a pre-programmed response for, it searches through the content of its assigned documents to find relevant answers.

This allows the agent to handle a wider range of inquiries accurately without needing explicit scripting for every possible scenario.

Assigning documents is done within each voice agent's configuration settings, typically under a 'Knowledge Base' tab.

You select which documents from the central repository that specific agent should have access to. Only the documents you explicitly check and save in the agent's configuration will be used by that agent.

(See the 'Assigning Knowledge Base Documents to Voice Agents' article for detailed steps on how to assign documents.)

When an agent needs to consult its knowledge during a call, it doesn't read entire documents at once. Instead, it uses sophisticated retrieval techniques to:

1. Search across the content of all currently assigned documents.
2. Identify the most relevant passages or chunks of text related to the user's query or the current conversation context.
3. Use the retrieved information to formulate an appropriate response.

The effectiveness of this retrieval depends on how well the documents are structured and whether the relevant information is present in the assigned set.

Agents always access the current version of the documents assigned to them. There is no separate 'syncing' delay or manual refresh needed for content changes.

If you update a document using the standard 'delete old / upload new / re-assign' workflow:

• The agent will stop using the old document as soon as it's deleted and unassigned.
• The agent will start using the new version of the document as soon as the new document is uploaded and assigned to the agent in its configuration.

The key is ensuring the correct, updated document is assigned to the agent in its settings.