Agentic AI: A Survey of Architectures, Memory, Planning, and Real-World Application

Artificial intelligence has rapidly evolved from narrow task-specificmodels to autonomous agents capable of perceiving, reasoning, planning,and acting in complex real-world environments. This survey provides acomprehensive overview of agentic AI systems, examining their corearchitectures, memory mechanisms, planning strategies, and practicalapplications across diverse domains. We review foundational frameworkssuch as ReAct, Chain-of-Thought reasoning, Retrieval-Augmented Generation(RAG), and tool-augmented language models. Applications in softwareengineering, healthcare, business automation, and scientific research arediscussed, followed by an analysis of open challenges includinghallucination, safety, and scalability.