Governance template 02 · Academy resource
Permission Boundary Checklist
A short control checklist for making the boundary around an AI-supported action explicit, explainable and reviewable.
Use this when
A team has identified a useful AI capability and is tempted to move from draft support to action without defining the stopping point.
Permission is not a technical setting. It is an organisational decision about what may happen, under whose authority, with what evidence and with which exception path.
Use this before configuration
Complete the checklist before a workflow is presented as ready. If the owner, evidence or stop point is unclear, the configuration is not ready.
Read the boundary aloud
A good boundary can be explained in one sentence: what the AI Worker may do, what it may prepare, and where a named human must take over.
Working prompts
Use the questions before the form.
Proposed action
What action or class of action is being considered?
Allowed without escalation
What may happen within the boundary?
Not allowed
What must never happen from this workflow?
Named authority owner
Who may authorise a change to this boundary?
Evidence standard
What must be true or evidenced before a human decision?
Exception and escalation
What event stops the process and how is the human owner reached?
If an exception path is missing, the boundary is incomplete. If the owner is unnamed, the boundary is not governed.