Skip to main content

Autonomy the customer grants

Autonomous Actions

Autonomy in A2Z Reach is granted per action type, never per agent, and only after that type has built a review record the customer can see. Each type sits on a four-state dial — autopilot, needs approval, always human-gated, or locked — and graduation is opt-in and reversible at any time.

Let proven action types run unattended once they have earned it, with per-type autonomy dials, a visible track record, and outreach permanently human-gated.

What this helps your team do

Grant autonomy per action type

Move a single action type to autopilot while every other type keeps waiting for approval, instead of trusting or distrusting the whole system at once.

See the record behind the dial

Review how many runs a type has completed and how many were approved unedited, edited, or rejected before deciding whether it should graduate.

Keep the irreversible work gated

Action types that leave the building or cannot be undone stay human-gated or locked regardless of track record.

How it works

  1. STEP 01

    Start every type on approval

    New action types draft their work and wait. Nothing runs unattended before it has a reviewed history.

  2. STEP 02

    Build a reviewable record

    Each run records whether the customer approved it unedited, edited it, or rejected it, and a type becomes graduation-eligible only after clearing a minimum run count and unedited rate.

  3. STEP 03

    Graduate, monitor, or revert

    The customer flips an eligible type to autopilot. Autopilot runs stay logged and verified, and the dial can be flipped back at any time.

Evidence available

  • Autonomy state per action type
  • Runs completed, approved unedited, edited, rejected, and pending
  • Unedited percentage against the graduation threshold
  • Runs remaining before a type becomes eligible
  • Full log of every autopilot run

Interpretation limits

  • Unattended execution is an Enterprise pricing row; Growth and Agency receive review-ready output that a person still approves.
  • Outreach and other externally visible or irreversible action types remain human-gated or locked and never graduate to autopilot.
  • Earned autonomy reflects a past review record and is not a guarantee about any future run.

Product sources

Related insights

Frequently asked questions

Can an agent send outreach without a person seeing it?

No. Action types that leave the building, such as outreach, are always human-gated. They do not become eligible for autopilot regardless of how strong the review record is.

Is autonomy granted to an agent or to a task?

To an action type. Trust is earned and revoked per type, so a strong record on one kind of work never silently extends to a different kind.

Can autonomy be taken back?

Yes. Graduation is opt-in and reversible — a type on autopilot can be returned to approval at any time, and autopilot runs remain logged and verified.

Find the evidence behind your AI visibility gaps.

Start with your brand website, review the generated measurement setup, and see the prompts, engines, competitors, and sources behind the first report.

Analyze My Brand