This session demonstrates three architectural patterns for building real-time decision systems with Azure Cosmos DB: Event Sourcing with hierarchical partition keys for immutable audit trails, temporal queries, and scalable event storage ; Change Feed pattern for push-based event notifications that trigger instant materialized view updates without polling; and Event-Driven Integration for capturing domain events from existing databases (PostgreSQL, Oracle, MongoDB) through Azure Service Bus while keeping source systems authoritative. Through a live fraud detection demo that combines traditional ML for clear-cut cases and Agentic AI for uncertain scenarios, you'll see how these patterns enable fast decisioning, complete auditability, and seamless integration across heterogeneous data sources.