Now a Claude connector. Set it up →
Automations

Automation that prepares the work, not one that runs off with it.

Rules watch your accounts around the clock, run health checks and queue decisions in the same Scale / Iterate / Fix / Watch language. What they never do is act outside the guardrails you set.

Approval-first by design: Meta ads are created paused, Google applying is opt-in and budget-only, and every action is permission-gated and logged.

AutomationsApproval-first · guardrails on
  • Budget step-up

    7 days above break-even ROAS

    One step, max +15%, 7-day freeze

    ScalePending approval
  • Fatigue watch

    Frequency up, CTR down 20%

    Drafts a refresh, never edits live copy

    IterateRunning
  • Search-term waste

    Spend, no conversions, 14 days

    Builds a negative list for approval

    FixPending approval
  • Tracking health

    Pixel or conversion signal drops

    Blocks scaling until it is fixed

    WatchPaused
Approve 2 Pause rule Failures surface in the log

Illustrative automation board — demo figures

The story

How it actually works.

1

Monitoring is the job

Rules run the checks you would run daily — pacing, fatigue, waste, tracking health — and only raise what crosses a real threshold.

2

Output is a decision, not a change

Each trigger produces a pending decision with its evidence and a suggested move, waiting in your queue.

3

Guardrails are explicit

Step size, freeze windows, spend ceilings and which platforms may be touched at all — set once, enforced every run.

4

Nothing hides

Every run, approval and failure is logged, so you can see what fired, what changed and what did not.

Today4 moves waiting on you

Revenue

$18,420

+12% vs last week

Spend

$6,180

+4% vs last week

Blended ROAS

2.98×

−0.2× vs last week

  • Scale

    Proof angle held above break-even all week

    Meta · Prospecting · ROAS 3.4× vs 2.1× break-even

    Approve
  • Iterate

    UGC hook v2 is losing click-through

    Meta · Broad · CTR −24%, frequency 3.1

    Approve
  • Fix

    Non-buying queries taking search spend

    Google · Search · 38 terms, no conversions

    Approve
  • Watch

    New angle is still too early to call

    Meta · Testing · 2 days, 9 conversions

    Approve

Illustrative approval queue — demo figures

What you get

Built for the work, not for the demo.

Continuous account watch

Meta and Google checked on a schedule so problems surface in hours, not at the end of the month.

Pending decision queue

Everything a rule finds arrives as an approvable item with evidence attached.

Guardrails you set

Step limits, freeze windows and permissions that constrain what any rule is allowed to propose or apply.

Health checks

Tracking, pixel signal and connection status verified before anything scales.

Failures surfaced

A failed sync or rejected write is reported, not silently swallowed.

Full activity log

A timestamped record of every run, approval and applied change.

What that means for you

Outcomes you actually feel.

  • The daily account read happens whether or not you have time for it.
  • Scaling that respects learning periods because freezes are enforced, not remembered.
  • No surprise spend: nothing applies outside the limits you set.
  • An audit trail you can hand to a client or a finance lead.
  • Fewer tabs, fewer manual checks, same control.

FAQ

Quick answers.

Can automations spend my money?

Not on Meta — new ads are created paused. On Google, applying is off by default and limited to budgets within your clamped steps.

What if a rule gets it wrong?

You see the decision before it acts. Dismiss it and the dismissal sticks.

Can I turn everything off?

Yes. Rules can be paused individually or as a whole, at any time.

Automation with the handbrake designed in.

Set the guardrails once and let the checks run without giving up the wheel.