The evolution of AI in enterprises has transitioned from simple chatbots to sophisticated, event-driven agents capable of autonomous decision-making and real-time processing of streaming data. These modern AI agents respond to system-generated events and tackle complex tasks at a massive scale, necessitating advanced stream processing capabilities like those offered by Apache Flink. Flink's continuous event processing, contextual data access, and fault tolerance make it an ideal platform for deploying AI agents. In response to specific needs in AI agent development, Flink Agents, a collaboration between Confluent and Alibaba, introduces enhancements such as agent semantics, dynamic topology support, enhanced observability, and MCP support. These features allow developers to create robust, scalable, and adaptive agentic systems within Flink's ecosystem. The integration of AI capabilities into Flink enables enterprises to leverage real-time, contextualized data for continuous and reliable agent operation, marking a shift towards event-driven AI applications. The open-source nature and community-driven development of Flink Agents aim to address real enterprise needs while encouraging broader participation in shaping the future of event-driven AI agents.