This chapter provides a foundational introduction to Agentic AI, a paradigm of autonomous, goal-driven systems that independently plan and execute multi-step tasks. It distinguishes Agentic AI from traditional predictive and generative AI by highlighting its core function of action-taking rather than content generation or classification. The chapter deconstructs the key architectural components of an AI agent—including a Large Language Model (LLM) core, reasoning and planning modules, persistent memory, and tool-use capabilities—that enable this autonomous behavior. It further explores the primary technical drivers fueling its emergence, such as the scaling of foundation models, the limitations of monolithic SaaS, physical and robotic integration, and the proliferation of specialized development frameworks. A central focus is placed on the novel security challenges introduced by these autonomous systems. The chapter argues that conventional cybersecurity measures are insufficient to address risks like unpredictability, new attack surfaces, and a lack of traceability. Finally, it establishes the need for specialized security paradigms by introducing the MAESTRO (Multi-Agent Environment, Security, Threat, Risk, and Outcome) framework as a structured approach to threat modeling for agentic environments, setting the stage for deeper analysis in subsequent chapters.