Autonomy Tiers
Autonomy tiers are like giving a helper different levels of permission: some tasks they can do on their own, but for anything that could cause permanent damage or a big mess, they must stop and ask you first. You only give them more freedom as you get better at checking their work.
A governance framework that maps agent permissions to action risk, where autonomy is tiered based on the reversibility and blast radius of the task. Higher autonomy is never a default state; it is an explicit, incremental grant contingent on the organization's capacity to verify the agent's output.
The classification that fixes the gate per action (read-only through irreversible/high-blast-radius); raising an agent's autonomy is an explicit governance decision, not a default, and autonomy is granted proportional to verification capacity. (C4AIL, agentic governance spec.)