What makes an AI agent an agent, and three questions that expose agent washing

In short: an AI agent uses a model to pursue a goal by taking actions: it chooses its own next step, calls tools that change other systems, and loops until the goal is met or a limit is reached. Gartner estimated in mid-2025 that only about 130 of the thousands of vendors claiming agentic products were real. Three questions expose agent washing: what can it do without a person, which systems can it write to, and what stops it.¹

An AI agent is a system that uses a model to pursue a goal by taking actions. It decides which steps to take, calls tools such as search, databases, email, payment systems or other agents, observes the results and continues until the goal is met or a limit is reached. Three features define it: autonomy (it chooses its next step), tool use (it changes the world, not only the text on a screen) and iteration (it plans, loops and corrects).