AI initiatives often stall not on model quality, but on governance throughput: stakeholders are overloaded, Jira-based intake is slow, and users avoid long complicated forms. In this session, we share how we built an agentic compliance copilot on the Databricks Agentic Mosaic AI platform that replaces rigid tickets with a conversational intake. The chatbot asks open-ended questions, lets users “just talk,” and then automatically structures, validates, and enriches the information required for AI Governance stakeholders. It prepares reviewer-ready packets covering EU AI Act (legal compliance), Security, Data Protection (GDPR), and architecture, while maintaining an auditable trail and empowering each stakeholder with exactly what they need.

We’ll walk through the problem and current process, then detail our solution design together with the challenges we have faces along the way. Agent orchestration, similarity search, RAG, human in the loop. We’ll dive into the technical architecture on Databricks (Unity Catalog, MLflow, Mosaic AI, Vector Search), including guardrails and hallucination control. We’ll close with the operational outcomes—shorter cycle times, reduced reviewer load, and higher user satisfaction—plus a practical tips you can adapt to streamline AI governance in your own organization.