Session

The reason AI agents struggle to move from PoC to production is that the design and implementation of "which AI may do what, and how far" remains ambiguous. In this session, we organize the AI governance design needed to enable secure access to enterprise data along three axes: safety, accountability, and consistency of permissions. We present a reference architecture that combines Microsoft Foundry with the MCP of CData Connect AI—which connects in real time to over 350 data sources—and deliver concrete design patterns aimed at the transition from PoC to production operation.