An AI agent is a system that uses a language model not just to answer a question but to pursue a goal: it plans, takes actions through tools (such as search, code execution, or APIs), observes the results, and decides what to do next, looping until the task is done.
A foundational technique is the 2022 paper “ReAct: Synergizing Reasoning and Acting in Language Models” (Yao et al.), which combined step-by-step reasoning with tool actions so a model could both think and act within one loop. Anthropic’s engineering article “Building Effective AI Agents” offers a practitioner’s view, distinguishing simple “workflows” from true “agents” and arguing for the simplest pattern that solves the problem.
Agents are the basis of tools that can write and run code, browse, and operate software on a user’s behalf.
Why business readers should care: Agents move AI from “advice” to “action,” automating multi-step work. That raises both the upside (real task completion) and the stakes (an agent that takes wrong actions), making controls and oversight essential.