Autonomy levels
Autonomy is a setting, not a vibe. By default the team asks before anything externally visible — publishing, sending, charging, or touching a paid connection. As trust builds, the owner widens the auto-approve list one playbook at a time. The CEO never escalates autonomy on its own.
In one breath
- Approvals Inbox shows pending decisions; Power Grid shows standing rules.
- Owners promote case-by-case approvals into standing rules.
- Default is low; the owner widens, never the system.
How autonomy is configured
Autonomy lives per-playbook and per-tool. The Approvals Inbox shows what is pending; the Power Grid shows what each specialist is currently allowed to do without approval. The owner can promote any pending request into a standing rule ("always send launch emails I have approved the copy for"), or revoke autonomy at any time.
The four steps owners actually take
Most owners walk a predictable path: read-only → draft-only → publish-after-approval → publish-on-rule. Black Box never assumes the next step; it presents the option after a clean run. This keeps autonomy a deliberate decision rather than an accumulated drift.
Why we default low
Bad autonomy on day one destroys trust on day two. We would rather you start manual and earn the auto-approve, than start auto and watch a specialist publish something that misses your voice.