AI permissions define what an agent is allowed to do. AI autonomy defines how independently it may perform an allowed action. A well-governed system treats these as separate controls.
Three different questions
For every consequential action ask: Can the tool technically do it? Has the business permitted it? If permitted, may the AI do it without a human each time? These are capability, permission, and autonomy.
Permission actions
A practical permission model distinguishes READ, CREATE, MODIFY, SEND, PUBLISH, DELETE, EXPORT/SHARE, ADMINISTER, and EXECUTE rather than granting vague “access.”
Catha autonomy levels
A0 is prohibited. A1 is recommend only. A2 is prepare while a human executes. A3 is approval-controlled AI execution. A4 is bounded autonomous execution inside explicit conditions.
Autonomy belongs to actions
An agent may be A4 for internal research, A2 for customer communications, and A0 for deleting records. Labeling the whole agent “autonomous” hides these distinctions.
Promotion and downgrade
Autonomy can increase only when authority, rules, permissions, evidence, monitoring, and recovery controls support it. Monitoring events may also force an autonomy downgrade.
How this fits the Catha AI Governance Build Kit
This topic is one part of the broader Catha AI Governance Build Kit, a fourteen-stage build process covering business outcome, governance discovery, authority, rules, permissions, autonomy, workflow design, validation, receipts, production approval, monitoring, recovery, final readback, and platform export.
The governance baseline belongs to the business. Platform-specific implementation may differ, but approved authority, rules, permissions, autonomy, approvals, and evidence requirements should not silently change.
Questions businesses ask about this topic.
Is tool access the same as permission?
No. A connected tool may expose capabilities that the governance policy still denies or approval-controls.
Should A4 be the goal?
Not necessarily. The correct autonomy level is the highest level the business can justify safely for that specific action.
Can autonomy change after deployment?
Yes, through controlled governance change, testing, and human approval.
Keep reading
Build the control system before you scale the agents.
The Catha AI Governance Build Kit helps businesses define authority, permissions, autonomy, approval gates, evidence, monitoring, recovery, and production controls before AI agents operate live.