Term · Accountability
The accountable human
When an AI agent makes a consequential call, there should be a person, a specific one with a name, who is answerable for it and could have changed it. A queue is not accountable. A department is not accountable. A person is.
The accountable human is the named person who is answerable for an AI agent's consequential decision and has the authority to change it. The word that matters is named. Accountability spread across a team, a process, or a policy is accountability that vanishes precisely when something goes wrong and everyone can point somewhere else. Naming a person is how you keep oversight from dissolving under pressure.
The specific, named individual answerable for a consequential AI-assisted decision, with the authority to change the outcome. Sometimes called the human of record.
Why a name and not a process
Processes diffuse responsibility by design, which is useful for throughput and corrosive for accountability. When a harm traces back to "the system" or "the team," no one decided and no one answers. A named accountable human reverses that. It restores the simple, old idea that for a decision that matters, someone in particular made it and stands behind it.
Authority has to match the name
Naming someone accountable without giving them real power to change the outcome is not oversight, it is a setup. It produces the failure I call accountability inversion: blame on the person with the least control. The accountable human only works when the name and the authority sit together, so that the person answering for the decision is the person who could actually have made it differently.
Read on
See the failure mode it prevents, Accountability Inversion, and where the role fits, escalation design and building oversight that holds.