A system is described as autonomous when it can perform its tasks and make decisions independently, without needing step-by-step human guidance or real-time intervention. This applies across a wide range: a fully self-driving car navigating traffic without a driver's input, a robotic vacuum mapping and cleaning a home on its own schedule, and an AI agent that can plan and execute a multi-step task after being given only a high-level goal are all forms of autonomy, even though they operate at very different levels of complexity and risk.
Autonomy exists on a spectrum rather than as an all-or-nothing property. Self-driving cars, for instance, are typically classified across six levels, from no automation to full autonomy in every driving condition, and most AI agents in production today sit somewhere in the middle: capable of independent action within defined boundaries, but still designed to hand control back to a human when they hit a situation they weren't built to handle.
