A knowledge base is the set of documents an AI agent answers from, such as FAQs, policies, product sheets and help articles. Grounding an agent in a knowledge base keeps its answers specific to your company and reduces made-up answers, often called hallucinations.
The documents are split into passages and indexed. When someone asks a question, the agent retrieves the most relevant passages and gives them to the language model with the question, a method called retrieval augmented generation (RAG). The model answers from those passages and can say when they don't cover the question.
A knowledge base is only as good as its content, so review unanswered questions every week and add answers.
Beyond Presence Managed Agents support knowledge bases and RAG: upload your documents when you create the agent in Studio. Read AI Avatars for customer support.
Yes. Managed Agents answer from the documents you upload, using retrieval augmented generation (RAG).
The answers your team gives most often: FAQs, policies, product, pricing and security information.
Tell the agent to say when it doesn't know, and review unanswered questions in your conversation analytics every week.