AI agent

Placeholder note adapted from Wikipedia's AI agent article.

An AI agent is commonly described as a system that observes an environment, chooses actions, and pursues goals with some degree of autonomy. In modern usage, the phrase often refers to software that can plan, call tools, make decisions, and continue work across multiple steps.

Core idea

The agent framing makes intelligence operational: a system receives percepts, maintains state or context, and selects actions according to a goal or performance measure. This can describe simple controllers, software assistants, reinforcement learning systems, or larger agentic workflows built around language models.

Why it matters

The useful distinction is not whether a system looks conversational, but whether it can turn observations into actions. This is why agents are discussed across robotics, planning, economics, cognitive science, and current LLM product design.

Open questions

Real systems still need clear objectives, reliable tool use, evaluation, and human oversight. The word "agent" is broad, so the important product question is usually more specific: what state does the system track, what actions can it take, and how do we know those actions are correct?