AI agents are defined by the work they are responsible for.
An AI agent is a software component given a bounded role, context, and a way to produce or act on information. The useful distinction is not whether it can chat; it is whether it can move a well-defined part of a workflow forward in a way a person can review.
Specialists beat vague handoffs
A single general-purpose assistant can be useful, but it often asks one prompt to carry research, drafting, critique, and formatting at once. Specialized AI agents keep those responsibilities separate. The researcher looks for evidence, the writer builds a draft, the editor tests clarity, and the publisher prepares the approved work.
How agents differ from chatbots
A chatbot is mainly an interface for a conversation. An AI agent can use a persistent brief, return structured work to another stage, and wait for a review decision. It may use a conversational interface, but the value comes from the workflow around the conversation.
The role of orchestration
Multi-agent systems need an agreed context and clear handoffs. In AURA, PIPELINE.md represents that shared operational memory: what is next, what has been learned, what changed, and what still requires approval.