Defining data governance on paper is simple; implementing it in a large organization with dozens of departments, hundreds of use cases, and fluid organizational structures is often a nightmare. Mapping business roles (data owners, stewards, and technical leads) to technical permission schemes, synchronizations with Entra ID, tag policies, and nested groups in Azure Databricks requires much more than simply applying generic templates.

In this session, we’ll share the real-world experience of a large-scale deployment at a major client where the initial challenge wasn’t the technology, but understanding the business. We’ll trace the journey: from an initial phase of manual control managed via Terraform and IaC, to the design and deployment of an agent-based architecture capable of interpreting unstructured requests from ticketing systems and determining the required permissions in Unity Catalog.

We will analyze in detail:
The reality check: How to map complex organizational charts and organizational exceptions to the native capabilities of Unity Catalog and Entra ID.
IaC as a foundation: Granular and reproducible management of groups and privileges using Terraform.
The agent-driven leap: How to train agents to infer correct assignments from ambiguous ticket requests.
Security Without Blind Faith: Implementing technical guardrails and human-in-the-loop oversight mechanisms to ensure that automation does not compromise data governance.

A session based on real-world experiences to discover how agentic AI can transform the most tedious and critical operations of a modern data platform.