“Human in the loop” has become a central idea in conversations about responsible AI. But having a human somewhere in a workflow does not tell us what that human can actually see, understand, decide, or change. A person may be reviewing AI outputs while understanding very little about the wider system producing them. Someone else may recognize problems across the system but have no authority to change its architecture. And a person with the authority to redesign the system may be far removed from what is actually happening inside it. This lightning talk uses a simple organizational case to explore the difference between participating in a loop, recognizing the architecture of the loop, and having the authority to reconfigure it.

Human in the Loop Is Not Enough: Who Is Observing the Loop?