What makes an "AI agent" fundamentally different from such a chatbot?

An AI agent goes beyond generating text by breaking a goal into steps, deciding which tools or actions are needed, and actually executing them, such as searching the web or running code, adjusting its plan based on what it learns. A regular chatbot can only produce a conversational reply based on its input, with no built-in ability to reach outside the conversation and change anything.

Answering in only one language has nothing to do with this distinction, since many chatbots are multilingual, and running without electricity is simply impossible for modern software, making both options irrelevant.

This shift toward agents is part of a broader trend in AI, where systems are increasingly judged not just on how well they talk about a task but on how reliably they complete multi-step tasks with real tools.