> ## Documentation Index
> Fetch the complete documentation index at: https://docs.marketer.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Guardrails and autonomy

> Set workspace performance, spend, scaling, audience, and launch-review policy before enabling automation.

**Guardrails** define hard workspace policy. **Autonomy** defines how new work starts and whether an accepted launch pauses for review. Provider permissions and ad-account state still apply underneath both.

<Warning>
  Guardrails reduce risk; they do not validate the campaign, creative, destination, audience, tracking, or provider account. Keep a human review step until real outputs and recovery paths are proven.
</Warning>

## Set performance targets

Open **Settings** → **Guardrails**.

| Field               | How Marketer uses it                                            |
| ------------------- | --------------------------------------------------------------- |
| **Target ROAS**     | Results above this target can qualify for scale-up under policy |
| **Break-even ROAS** | Results below this threshold can trigger pull-back behavior     |
| **Target CPA**      | Optional acquisition-cost target used where relevant            |

Targets are decision thresholds, not promises. Align their date range, conversion definition, and attribution source with how the team evaluates performance.

Blank numeric fields use Marketer's system defaults. If a number is important to your operating policy, enter it explicitly rather than relying on an unseen fallback.

## Set currency-bound spend caps

Choose the guardrail currency first, then set:

* **Monthly cap**;
* **Daily cap**; and
* **Minimum spend floor**.

Caps are enforced only on ad accounts billed in the selected currency. Marketer does not convert a different provider currency to make it fit the cap.

<Warning>
  A USD cap does not govern an account billed in EUR, GBP, or another currency. Verify each connected ad account's native billing currency before relying on a cap.
</Warning>

The minimum spend floor expresses an operating lower bound; it does not override a provider rejection, a hard cap, a paused campaign, or unavailable funds.

## Control scaling

| Control                     | Effect                                                                           |
| --------------------------- | -------------------------------------------------------------------------------- |
| **Allow budget increases**  | Master switch; when off, Marketer cannot make scale-up increases                 |
| **Max budget increase (%)** | Hard ceiling for any one permitted scale-up                                      |
| **Approval threshold (%)**  | Changes above this size require approval instead of applying under normal policy |

Use all three together. For example, a maximum can bound the size of an allowed increase while the approval threshold sends a smaller high-risk band to a reviewer.

## Protect campaigns

Use **Untouchable campaigns** to select synced Meta campaigns the agent must never modify.

* Select by the exact Meta campaign and account.
* Revisit the list when campaigns are renamed, replaced, or archived.
* Keep legal, evergreen, or externally managed campaigns protected.
* Confirm a protected campaign remains selected after a provider reconnection or roster change.

Untouchable status protects modification. It does not remove the campaign from read-only analysis.

## Configure new-customer acquisition

**New customer acquisition** is configured per Meta ad account. Choose the Custom Audiences that represent known customers under **Customer exclusions**.

This setting uses provider audiences to prevent acquisition work from targeting existing customers. It is separate from:

* **Existing customers** reporting roles;
* **Engaged audience** reporting roles; and
* ad-set-specific targeting choices.

Build and verify audiences in [Custom audiences](/workspace-admin/custom-audiences), then confirm the exclusion on the exact Meta ad account.

## Configure autonomy

Open **Settings** → **Autonomy**.

### Default creation mode

| Mode               | How a new brief opens                       |
| ------------------ | ------------------------------------------- |
| **Ask every time** | Presents the automatic-versus-manual choice |
| **Auto order**     | Opens the product-URL order flow            |
| **Manual brief**   | Opens the write-it-yourself flow            |

This is a workspace default, not a lock. A teammate can choose another mode on an individual brief.

### Autonomy level

| Level                 | Launch behavior                                                      | Recommended use                                           |
| --------------------- | -------------------------------------------------------------------- | --------------------------------------------------------- |
| **Review everything** | Every accepted launch remains paused for review                      | Default while testing or when a launch owner must approve |
| **Hands-off**         | Accepted orders launch **ACTIVE** and can begin spending immediately | Narrow, proven managed-plan workflows only                |

The advanced control can appear as **Launch accepted orders live (ACTIVE)**. Turning it on opens **Go hands-off?** or **Launch accepted orders live?** and states the spend consequence.

Hands-off is available only on managed plans. Guardrails and spend caps remain active, but the paused review stop is removed.

## Recommended rollout

<Steps>
  <Step title="Validate read-only outputs">
    Compare real analysis with connected provider data and completed coverage.
  </Step>

  <Step title="Enter explicit limits">
    Set currency, spend caps, scaling controls, customer exclusions, and untouchable campaigns.
  </Step>

  <Step title="Keep Review everything">
    Test proposals and accepted launches while they remain paused.
  </Step>

  <Step title="Inspect a real paused launch">
    Verify provider account, destination, tracking, creative, audience, budget, naming, and approval ownership.
  </Step>

  <Step title="Expand one workflow">
    Consider **Hands-off** only for a proven scope with monitoring and a named rollback owner.
  </Step>
</Steps>

## Change review checklist

* Active workspace and effective role
* Ad account and native currency
* Target and break-even definitions
* Daily and monthly caps
* Increase master switch, maximum, and approval threshold
* Untouchable campaign roster
* Customer-exclusion audiences
* Default brief mode
* **Review everything** or **Hands-off** consequence
* Notification owner for performance, budget, and approval events

## Recovery

* If a change is blocked, identify whether the blocker is role, provider permission, guardrail, approval, account state, or sync coverage.
* If hands-off was enabled unintentionally, return to **Review everything** before starting more work.
* If the wrong currency was configured, correct the guardrail and audit affected account coverage; do not assume conversion applied.
* If an audience exclusion is missing, pause acquisition work until the correct Meta account and audience are selected.
* If a protected campaign changed, contain provider-side activity and review whether it was protected at the time and in the correct workspace.

## Next step

Configure [Notifications and billing](/workspace-admin/notifications-and-billing) so warnings, approvals, and failed-payment events have a responsible recipient.


## Related topics

- [Approvals, guardrails, and autonomy](/essential-guidelines/approvals-and-autonomy.md)
- [Branding and Autofills](/workspace-admin/brand-and-autofills.md)
- [MCP connect](/workspace-admin/mcp-connect.md)
- [Members and roles](/workspace-admin/members-and-roles.md)
- [Integrations](/integrations/overview.md)