Measure the quality, accuracy, and relevance of your agent's responses with three new Retrieval-Augmented Generation (RAG) evaluation metrics. The new Answer Faithfulness, Answer Relevance, and Context Relevance metrics added to the Analytics Foundation Data Model give you a trust score of your AI agent's interactions. Metadata labels for Agent topics and Agent types were also added to the semantic model builder.
Where: This change applies to Agentforce in Lightning Experience in Enterprise and Unlimited editions with the Einstein for Sales, Einstein for Platform, or Einstein for Service add-on.
When: These changes are available starting the week of February 9, 2026.
Why: To monitor and quantify your AI agent's customer support quality, you need to know how accurately AI agents are answering customers questions and requests:
- Average Answer Faithfulness Score: Measures how well the agent adheres to your grounded source data (like Knowledge Base articles).
- Average Answer Relevance Score: Measures how well the response addresses the customer's question or request.
- Average Context Relevance Score: Measures if the agent retrieved the correct data source to answer the question.
How: To start using these metrics, in Setup, toggle on Audit and Feedback and Knowledge/RAG Quality Data and Metrics.
SEE ALSO | Salesforce Help: Set Up Agent Analytics, Data Model and Calculated Fields for Agent Analytics, Set Up RAG Quality Metrics