Most data teams already run a Lakehouse. The harder problem now is deciding how far AI should be embedded into the data platform itself, and what that means for governance, ownership, and operational complexity. Two tools are gaining attention in this subject: Databricks Genie and Microsoft Fabric Data Agents.

This session looks at Genie and Fabric Data Agents as platform-level design choices. Both promise natural language access and AI-driven interactions with data, but they rely on very different assumptions about metadata, semantic enforcement, and control boundaries. Those assumptions matter once teams operate multiple domains, shared Lakehouses, and strict governance requirements.

Using real architectural patterns, we examine how each approach fits into a production data platform built on Lakehouse principles. We discuss where AI logic lives, how trust in results is established, what breaks at scale, and how these tools interact with lineage, security models, and DataOps workflows. The comparison focuses on consequences, not capabilities.

The goal is to help data engineers and platform architects make informed decisions about embedding AI into the core of their data platform, understanding the trade-offs before these tools become foundational dependencies.