Part 2 · Set up
Autonomy stages
How much runs without you. Four stages, and none removes your approval on a message.
01What it is
A setting on your workspace that decides how much the system does on its own. It is database state and every check reads it server-side.
02Why it exists
Autonomy is earned by evidence rather than granted by preference. The reason to move up a stage is that you stopped editing drafts, and the reason to move down is that you started again.
No stage removes your approval on a message to a named person. Higher stages widen what the system decides on its own: which experiment gets budget, when a follow-up goes, what gets tried next.
03How to use it
- 01Start at stage 1. It drafts and you approve everything.
- 02Move to stage 2 once you approve most drafts without editing.
- 03Use the global pause rather than dropping a stage when something looks wrong. It takes effect at every stage immediately.
- 04Check your plan. Higher stages need a higher tier and the gate enforces it.
04What good looks like
- You give it
- Two weeks at stage 1 with a falling edit rate, then a move to stage 2.
- You get
- Routine sends go inside the limits you set, and exceptions still come to you.
- Why
- Stage 2 sends within limits and surfaces anything below the confidence bar.
Derived from the rules in the code
05What weak looks like, and what it costs
- You give it
- Stage 3 on day one.
- You get
- The gate refuses. The stage is capped by your plan, and the drafts still need approval that nobody has read.
- Why
- The pre-send gate checks the stage against the plan and refuses with a named reason.
Derived from the rules in the code