What Makes an AI Agent "Agentic"?

Most "AI" tools today are reactive — they answer a prompt and stop. Agentic systems are different: they can break a goal into steps, call tools, and adapt when a step fails.

In practice, this means an agent handling customer refunds can check order history, apply policy rules, issue the refund, and only escalate to a human when the case falls outside its confidence threshold.