Evolution of Agentic Artificial Intelligence: From Classical Intelligent Agents to LLM-Based Autonomous Systems

Abstract Agentic Artificial Intelligence (Agentic AI) refers to AI systems that can plan, take actions and adapt their behavior to achieve goals with limited human supervision. This review traces the evolution of agentic AI from classical intelligent agents and multi-agent systems to today’s Large Language Model (LLM)-based agents that use tools, memory and multi-step reasoning. The chapter synthesizes key ideas, architectures, enabling technologies, evaluation methods, real-world applications and emerging safety and governance practices. It highlights how modern agentic systems combine planning, tool use, retrieval, reflection and coordination across multiple agents, while also introducing new risks such as unsafe autonomy, prompt injection and reliability failures. Finally, the review proposes research directions toward more robust, transparent and accountable agentic AI suitable for high-stakes deployment. Keywords Agentic AI, autonomous agents, large language models, tool use, planning, multi-agent systems, memory, safety, governance.