Agentic AI Patterns - Workflows and Agents for Sophisticated AI Systems

Explore the emerging patterns in Agentic AI systems through this 52-minute conference talk from Devoxx. Learn about the two main architectural categories that define sophisticated AI agent systems: workflows that orchestrate LLMs and tools through predefined code paths, and autonomous agents where LLMs dynamically control their own processes and tool usage. Discover how multiple AI services coordinate and integrate to handle complex tasks, while understanding the fundamental differences between these approaches and their respective ranges of applicability. Examine the significant testing challenges these systems present for mission-critical scenarios due to their non-deterministic nature. Gain practical insights through concrete implementation examples using Quarkus and its LangChain4j extension, though the underlying concepts apply universally across different technology stacks. Understand the theoretical foundations of Agentic AI while seeing how these patterns can be effectively implemented and tested in real-world applications.