The Autonomy Ladder

Autonomy is earned, never assumed.

Every seller starts at the bottom and moves up only by choosing to. Each step stays inside your Guardrail Policy, is fully reconciled and logged, and is reversible — drop back a level or hit emergency stop any time.

0

Shadow Mode

No execution

The full Runbook runs and you see the complete Manifest with Data Confidence Scores — but nothing is applied. Telegram sends a summary only. Ideal for evaluating AdsPlane or testing new Guardrails with zero risk.

1

Manual Gate

Only what you approve

Execution happens only after you approve — from the web or Telegram. Approve everything, a high-confidence subset, or individual actions. This is the default, and it’s sustainable long-term for cautious sellers.

2

Guided Auto-Approval

Low-risk, in-policy actions auto-run

A single Guardrails toggle lets low-risk, in-policy, high-confidence actions skip the manual gate; everything else still waits for you. Flip it when you trust the recommendations — no rigid waiting period.

3

Trusted Automation

Wider auto-approval scope

As confidence grows you widen the auto-approved action types and thresholds in your Guardrails. More hands-off, still fully bounded, reconciled, and logged.

4

Intraday — Rally-Control

Opt-in intraday cycles

Between daily Runbooks, opt-in intraday cycles act on fast-moving signals under a separate Rally Guardrail policy with movement caps and cooldowns. For higher-volume operations that need same-day responsiveness.

AI never decides to spend — the deterministic engine does, and only within your rails. See how the loop works → or start in Shadow Mode →