Agentic AI
AI applications where the model autonomously plans and executes multi-step tasks using tools.
Agentic AI is what you get when you give an LLM tools, a goal, and a loop. Instead of answering a single question, the system takes many steps retrieving information, calling APIs, writing files, evaluating progress to accomplish a higher-level objective.
The term covers a wide spectrum, from short single-turn agents (like a code-generating assistant) to long-horizon autonomous workers (e.g., software engineering agents that can take a ticket from inbox to merged PR). The reliability gap shrinks every six months but is still real.
Production agentic AI usually combines a strong reasoning model, well-designed tools, careful prompts, robust error handling, and for high-stakes work human-in-the-loop checkpoints. MCP has become the standard transport for tools.