Agentic AI
Also known as: AI agent
Agentic AI describes AI systems that are goal-directed rather than reactive — given an objective, they plan, take multiple steps, use tools, check their own work, and continue until the task is done, instead of just answering a single prompt.
A traditional chatbot is reactive: you ask, it answers, the loop ends. An agentic system is autonomous within bounds: you hand it an outcome, and it decides the steps, calls tools, reacts to what it finds, and keeps going until it reaches the goal — with a human approving key actions.
In practice, agentic AI is what lets a tool open your files, pull data from several sources, write a report, and flag what matters — all from one instruction. Claude Cowork and the OpenAI Codex app are both examples of agentic AI applied to the desktop.