---
title: PlainRouter Actions — Govern and verify every ad account change
description: Turn AI recommendations into policy-checked, approved, executed, and verified Meta account changes.
canonical: https://plainrouter.com/product/actions
last_updated: 2026-08-26
---

# Move faster without handing AI a blank check

Actions turns a recommendation into a reviewable proposal with evidence, policy checks, approval, controlled execution, and provider verification. Your team spends less time relaying changes and keeps control of what reaches Meta.

[Join design partners](/register)

## Sitemap

See the full [sitemap](/sitemap.md) for all pages.

## The expensive part of automation is uncertainty

A recommendation has limited value if an operator must reconstruct its reasoning, manually repeat the change, and later check whether it landed. Actions keeps the decision context attached from proposal through outcome, reducing handoff cost without hiding consequential changes.

- **Decision latency — Shorter:** Put the recommendation, rationale, evidence, and intended change in one queue.
- **Account changes — Governed:** Evaluate scope, limits, evidence, and execution mode before a provider write.
- **Execution status — Verified:** Separate “request accepted” from evidence that Meta reached the intended state.

## How Actions works

1. **Gather current evidence.** PlainRouter records the account, signal, and performance context used to support the recommendation.
2. **Create a canonical proposal.** The agent names the exact target, requested change, rationale, and stable retry key. No provider write happens yet.
3. **Evaluate workspace policy.** The gate checks allowed action types, execution mode, dynamic evidence, account state, and configured limits.
4. **Approve or reject.** An operator reviews the whole batch in one queue, or policy can permit bounded automation within the selected mode.
5. **Execute, verify, and measure.** PlainRouter records the provider receipt, verifies the resulting state, and follows the outcome or recovery path when needed.

## What changes for your team

- **Review decisions, not scavenger hunts:** See what will change, why it was proposed, which evidence was used, and whether policy passed without rebuilding context.
- **Encode how your team operates:** Choose suggest-only, bounded automation, or fuller automation while keeping shared policy at the execution boundary.
- **Know the difference between sent and landed:** Provider receipts, follow-up reads, and verification states make uncertainty visible instead of treating an API response as success.
- **Keep a durable decision history:** Proposals, policy decisions, approvals, execution, verification, outcomes, and recovery events remain tied to the responsible actor.

## The approval gate carries context, not just a yes or no

Actions is designed to make automation operable. It preserves the evidence and execution state needed to understand a change before, during, and after it reaches the advertising platform.

- Supported proposals include budget adjustments, delivery-status changes, creative uploads, and paused creative duplicates.
- The authenticated human or agent identity is recorded as the proposer; callers cannot supply a different identity.
- Unknown or unhealthy dynamic evidence can block a proposal instead of being silently ignored.
- Retries use persisted execution receipts to avoid repeating provider writes when the outcome is uncertain.
- The audit trail covers proposals, policy, approval, execution, verification, outcomes, and recovery.

## Actions FAQ

### Does every action need manual approval?

That depends on the workspace execution mode and policy. Teams can remain suggest-only or permit bounded automation, while consequential changes still pass through the same policy boundary.

### What can Actions change today?

Actions supports Meta budget adjustments, active or paused delivery status, creative asset uploads, and paused ad duplication with a selected creative.

### What happens if Meta accepts a request but the result is unclear?

PlainRouter preserves the receipt, avoids blindly repeating the write, and continues through verification or recovery states until the result can be classified.

### Can I see who proposed and approved a change?

Yes. The audit history ties each proposal and decision to the responsible human or agent, alongside the policy and execution events.

## Bring us the change your team is afraid to automate

We are working with design partners to encode real operating policies, shorten approval cycles, and make provider-side results easier to trust.

[Join design partners](/register)