In the preceding chapters, we’ve explored fundamental agentic patterns: Chaining for sequential execution, Routing for dynamic path selection, and Parallelization for concurrent task execution. These patterns enable agents to perform complex tasks more efficiently and flexibly. However, even with sophisticated workflows, an agent’s initial output or plan might not be optimal, accurate, or complete. This is where the Reflection pattern comes into play.