Agentic Reasoning and Acting (RAG) is emerging as a significant advancement in the field of generative AI applications. Traditional RAG systems are typically hard-coded pipelines focused on retrieval and generation. However, Agentic RAG introduces a more dynamic approach by utilizing large language models (LLMs) for decision-making. The ReAct framework, developed by Shunyu Yao and colleagues, exemplifies this shift by incorporating multi-turn interactions, query/task planning layers, and tool interfaces for external environments. Additionally, Agentic RAG systems feature reflection and memory capabilities for personalization, intelligent routing for optimal query handling, multi-step reasoning that mimics human thought, and adaptive learning to refine strategies over time. LlamaIndex further enhances Agentic RAG by taking generative AI applications to the next level. These advancements are poised to revolutionize how we interact with LLMs, offering more sophisticated and personalized AI experiences.