Agentic Artificial Intelligence (Agentic AI) refers to AI systems that can work through several steps to reach a goal. They can split a task, look for information, use tools, work with other agents, and try another approach when a result is not good enough. This paper reviews large language models (LLMs), multi-agent systems, reinforcement learning, retrieval-augmented generation (RAG), and memory. It also looks at possible uses in healthcare, finance, cybersecurity, disaster response, and software development. Some issues remain, such as wrong answers, computing cost, poor coordination, security, and the need for human control. The paper presents a sample design with four roles: Research, Analysis, Decision, and Recommendation. The design explains how the work could be shared among agents, but it still needs to be built and tested.