Why MCP-Created Campaigns Pause but Budget Edits Go Live
Why do MCP-created campaigns land paused but budget edits go live instantly?
Creation is treated as the high-variance operation, so new objects arrive paused for manual activation. Edits to existing objects apply immediately. The practical effect inverts intuition: the routine-feeling action, changing a budget, is the unguarded one, and it is also the action assistants are asked to perform most often.
For the surrounding account decisions, compare Approval Gates: The Guardrail Before Any Live Ad Change and When Not to Trust an Autonomous Budget Shift.
In short
Meta’s MCP connector gates one class of action and not the other. Anything the assistant creates — campaign, ad set, ad — lands paused and waits for a human to activate it. Anything that already exists can be edited live: budget, targeting, status. Understanding why produces a usable rule for where to put your own gate.
Why the asymmetry exists
Creation is unbounded. A model asked to build a campaign chooses objective, audience, placements, budget, schedule and creative — a large space with many ways to be quietly wrong, and no prior state to compare against. Pausing the output makes that reviewable before money moves.
Editing is bounded. The object exists, the change is usually one field, and the prior value is recoverable. From the platform’s position that is a much smaller blast radius, so it ships unguarded.
The reasoning is sound and the consequence is still awkward, because blast radius per action is not the same as blast radius per week.
Action gating table
| Action | Gated by Meta? | Actual risk | Where your gate belongs |
|---|---|---|---|
| Create campaign / ad set / ad | Yes — lands paused | Moderate; reviewable before spend | Meta’s gate is sufficient for most accounts |
| Raise or lower an ad set budget | No | High — applies to live delivery, and may restart learning | Require approval above a change threshold you set |
| Pause or activate an existing ad set | No | High — an activation spends immediately; a pause can strand learning | Require approval on activation; log every pause with a reason |
| Change targeting on a live ad set | No | High — any targeting change is a significant edit that restarts learning | Treat as equivalent to creating a new ad set |
| Adjust bid or cost cap | No | High — can stop delivery outright or uncap cost | Require approval; check against your margin floor first |
| Read performance data | No | Low | No gate needed |
The rule that falls out
Gate on whether the change touches live delivery, not on whether it feels routine. Creating a paused campaign touches nothing. Moving a live ad set’s budget by forty percent touches everything, and nothing in the connector will stop it or tell you afterwards that it was the cause.
The frequency point compounds this. Budget adjustment is the single most common request anyone makes of an ads assistant, which means the unguarded path is also the well-travelled one.
Interpretation boundary
This describes the connector’s documented behaviour, not a ranking of how often each action goes wrong in practice — that depends on your account, your spend level and how the assistant is prompted. Verify the current gating behaviour against Meta’s documentation before designing around it. Treat the table as a starting map for where to place your own controls, with the thresholds supplied by you rather than inherited from this page.
Documented thresholds
Meta publishes the figures below; they are reproduced here with the article they come from. Retrieved 2026-09-22 — Meta revises this documentation without notice, so verify anything load-bearing against the current version before you act on it.
- What a budget change costs you. Meta treats budget amount as significant depending on magnitude, and publishes the calibration: $100 to $101 is unlikely to cause learning-phase re-entry, while $100 to $1000 may. Small frequent steps are safer than single large jumps.
- The floor a budget has to clear. Roughly 50 × your cost per optimisation event per week. At $12 per result that is about $600 weekly; at $30, $1,500; at $55, $2,750.
- Low bid or cost control is a documented cause of learning limited, and Meta surfaces it directly: if an ad set is limited by a low bid or cost control, a tooltip appears on hover over the status. That turns one of the six causes into a readout rather than a guess.
Source: Meta Business Help Centre — https://www.facebook.com/business/help/316478108955072, https://www.facebook.com/business/help/269269737396981.
Can software help?
Bach.ai audits your connected Meta account, estimates the revenue impact of what it finds, and proposes specific fixes. It applies a change only after you approve it. Think of it as an automated audit layer that surfaces issues and proposed fixes for your review — not a replacement for your team’s judgment, and creative production is not its core job, though the Pro and Agency plans can generate a limited number of variants.
FAQ
Do AI-created Meta campaigns start running automatically?
No. Objects created through the MCP connector arrive paused and require manual activation, which gives you a review point before any spend occurs.
Which agent actions on Meta are ungated?
Edits to objects that already exist: budget changes, targeting changes, bid and cost cap changes, and pausing or activating a live ad set. All apply immediately.
Where should approval gates go for an AI ads agent?
On any change that touches live delivery, regardless of how routine it feels. Budget moves, targeting edits and activations warrant a gate; reading performance data does not.
Method and sources
“Creation is treated as the high-variance operation, so new objects arrive paused for manual activation.”
Source: Where this guide describes platform behaviour, it follows Meta’s published advertising and Marketing API documentation, which changes without notice — verify anything load-bearing against the current version before you act on it. Every threshold the guide asks you to supply is first-party, drawn from your own account exports and commerce ledger, because no external benchmark can stand in for your own margin structure.