Autonomous Agent

An agent that operates without step-by-step human input, deciding its own actions to reach a stated goal.

Autonomous in this context doesn’t mean “fully self-directing forever” — it means “doesn’t need a human to approve each step.” An autonomous agent receives a goal, makes decisions internally about what to do next, and only escalates to a human on stop conditions (uncertainty, errors, completion).

The autonomy spectrum runs from human-in-the-loop (every action approved) through human-on-the-loop (human monitors, can interrupt) to autonomous within scope (free to act inside defined boundaries). Production deployments rarely use full autonomy outside narrow scopes — the cost of errors at scale outweighs the gain in throughput.