In plain words
An agent is a system that can choose actions in response to its situation. In modern AI products, the word usually means a language model connected to tools and a loop that lets it carry out a task over several steps.
A closer look
Instead of producing only one answer, an agent can inspect information, choose a tool, observe the result, and decide what to do next. The surrounding software manages access, state, and stopping conditions. Some systems follow a largely fixed workflow; others let the model choose more of the path. The boundary is a spectrum, and the term is used inconsistently.
An agent’s useful capabilities depend on the whole system: its model, tools, instructions, permissions, and feedback. A capable model with no access to your calendar cannot book a meeting. Giving it calendar access still does not guarantee that it will choose the right time.
In practice
A research agent searches for three reports, reads them, notices a missing figure, searches again, and produces a comparison with sources. Its next action depends on what it finds.
A useful distinction
“Agent” does not mean conscious, fully autonomous, or morally responsible. It describes an operational role. A technical agent is not automatically a moral agent.