Agent Authority

Authority is distinct from capability. An agent may be technically able to send mail, write to a system of record or approve a request, and separately permitted to do so. Conflating the two is how automation causes harm: nobody decided the agent could act, it simply could. Declaring authority means enumerating the actions available, the scope each is bounded by, the conditions under which a human must confirm, and what happens when the agent encounters something outside its remit. The declaration is also what makes an audit tractable, since the question becomes whether the agent acted within a stated boundary rather than whether the action seems reasonable.

Audiences and roles bound what any process can reach in Centralpoint, and Data Triggers make the conditions for automated action explicit rather than implicit in code. Actions taken are recorded alongside the AI activity that prompted them, so an agent's behaviour can be reviewed against the authority it was granted.