Last updated: 05/14/2026

This guide is for AviaryAI Knowledge Base admins — the managers, team leads, or project owners responsible for uploading source documents, managing testers, and preparing the knowledge base for a voice agent deployment.


How It Works

The Knowledge Base is the source of truth your AviaryAI voice agent uses to answer member questions. Before your voice agent goes live, your team will:

  1. Upload your organization's source documents (PDFs only) into a single Topic.
  2. Sync those documents so the AI can read them.
  3. Test the knowledge base by asking questions in the voice of your members.
  4. Flag inaccurate or incomplete answers so admins can review and improve them.
  5. Review contributions — accept, reject, or update source docs based on tester feedback.
  6. Hit the accuracy benchmark (90–95% across 200–250 unique questions) before voice agent testing begins.

Goal: A knowledge base that answers member-style questions with 90–95% accuracy. That's the trigger to move from KB testing into voice agent testing.


What You'll Need


Who Should Complete This