Agentic AI Patterns for Enterprise Software, Kevin Dubois

This talk focuses on the development of AI applications in enterprise settings, emphasizing the complexity introduced by the use of advanced Agentic AI systems. The speaker addresses the challenge of orchestrating multiple specialized agents that possess independent reasoning capabilities, which are essential for enhancing collaboration towards common goals. The session explores a range of agentic patterns, from rigid workflows to flexible, autonomous orchestration using large language models (LLMs), and emphasizes the practical integration of these patterns in an enterprise software environment using Java, LangChain4j, and its Quarkus extension. Additionally, the speaker discusses how to utilize existing infrastructure for effective agent collaboration while maintaining the flexibility to implement custom agentic patterns.