Fabric Data Agents allow you to ask natural language questions directly to data wherever it resides, whether in OneLake (lakehouse), the Data Warehouse, Power BI semantic models, or even KQL databases. In this session, we'll explore how the agent leverages user credentials (via Microsoft Entra) to query only the data scopes to which it has access, interpret the question, and translate it into queries.
We'll also see how to guide the agent with custom prompts and contextual rules that drive response quality.
An end-to-end demo will move from the authorized dataset to the generated response, demonstrating the complete step-by-step flow.