Fabric Data Agents promise a natural-language interface over enterprise data, but real-world adoption depends on more than just connecting a source. In this session, we'll start with an SQL database in Fabric as our data source and we will walk through what it actually takes to make a Microsoft Fabric Data Agent useful in production: grounding it with the right descriptions, instructions, and example queries; improving answer quality and SQL generation; and enforcing data access through SQL Row-Level Security rather than prompt logic.

We will also address one of today’s most interesting limitations: although English is the only officially supported language, we will show a practical configuration pattern that enables the agent to understand business questions in Italian (or any other language) and answer back in Italian (or any other language) with reliable results. The session is very hands-on and consists almost entirely of a demo showing how to tune the Data Agent based on specific needs.