AI-drafted change proposals for human approval
Insight is useful only when it turns into action. The AI analyst can turn its findings into change proposals — which then go through exactly the same checks as a person's proposals.
What it does
When you agree with a recommendation, ask the analyst to propose the change. It creates a proposal through the same code as the app, marked with the origin “model”, which is part of the digest and cannot be changed. The client's limits and policy apply; the model cannot override limits, even with a convincing justification.
Arguments such as “confirm” or “approve” in the model's request are ignored with an explicit message: they grant nothing. A person with the right role approves the proposal, or the client's auto policy applies it if it is reversible and within limits.
Why it helps
- From analysis to action in one conversation.
- Clear rationale attached to every AI proposal.
- Impossible for the AI to bypass approvals or limits.
How to set it up
- Discuss the issue with the analyst until you agree on a change.
- Ask: “Propose pausing this campaign” or “Propose a 20% budget decrease”.
- Open Approvals, review the model's proposal and approve or reject it.
Good to know
- Analysts (the role) cannot create proposals, and the AI inherits the same restriction for them.
- Proposals expire after 60 minutes like any other.



Try it on your own accounts
Create a workspace, connect Google or Meta in a couple of clicks and see your accounts clearly. Changes follow your approvals or the policy you configure.
Create your workspace