On November 3, 2026, Supermetrics announced that its Claude integration can create and edit ad campaigns, not only report on them. New campaigns arrive paused, waiting for a person to approve them. That makes the approval step the real control, so it deserves a deliberate review process.
- Supermetrics says campaigns created through its Claude integration are paused by default until reviewed and approved, across Google Ads, Meta, Microsoft Advertising, TikTok, LinkedIn and Snapchat.
- The release does not describe pricing, availability dates or how per-account approval settings work.
- Its adoption and time-saving figures are vendor-reported, with the time saving covering one task at one agency.
- A paused draft is only safe if the reviewer knows what to check, so agree a checklist before the first campaign.
What was announced
Supermetrics, a Helsinki marketing data company, said on November 3, 2026 that marketers can now create and edit campaigns through Claude using their marketing data, business context and human approval. The supported platforms are Google Ads, Meta, Microsoft Advertising, TikTok, LinkedIn and Snapchat.
The release says new campaigns are paused by default and stay paused until reviewed and approved, and that proposed edits can be reviewed before they are applied. Customers can set approval requirements for different accounts. A “Context Layer” lets teams define targets, naming conventions and reporting rules in natural language.
Co-CEO Mikael Thuneberg is quoted saying, “The goal isn’t to remove marketers from the process.” For buyers, the design choice matters more than the quote: the human sits at the point of activation, not at the point of drafting.
What the release does and does not establish
The release gives adoption figures. It says more than 6,000 customers have used the Claude connector since its February 2026 launch, that active users grew an average of 250% month over month, and that paid-media reporting is the dominant use case. Those are Supermetrics’ own numbers, with no method given.
It also cites agency Layer cutting a bespoke client-reporting task from up to 10 hours to about 20 minutes. That is one task at one customer, and it concerns reporting, not campaign creation. Nothing in the release measures whether AI-built campaigns perform better or worse than human-built ones.
| Question a buyer will ask | What the release says | Status |
|---|---|---|
| Can the agent launch spend on its own? | New campaigns are paused until approved | Stated |
| Can approval rules differ by account? | Customers can set approval requirements for different accounts; no detail on how | Stated, unspecified |
| Do AI-built campaigns perform better? | No performance comparison given | Not shown |
| What does it cost, and when is it available? | No pricing or availability dates mentioned | Not stated |
Where approval can fail
A pause is a control only if the reviewer can see what matters. Our earlier piece on least-privilege access for AI marketing agents covers what to restrict upstream, and this one covers the moment a human reads the draft.
The failure mode is routine rather than dramatic. A reviewer sees a clean campaign name, a plausible budget and tidy ad copy, clicks approve, and never opens the audience, location or conversion settings. Speed is the pressure: the whole point of the tool is saving time, so reviews drift toward skimming.
Our post on why Gartner expects many agentic AI projects to be cancelled covers the wider budget risk. Whatever the pilot, define who approves, what they check and how long they have before the first campaign is created.
A review checklist for paused drafts
The steps below are our recommended process, not a Supermetrics requirement, and the release does not specify what reviewers see before approving. Adapt them to each platform’s own settings screens.
- Confirm the ad account and campaign objective match the brief, since an agent can draft in the wrong account.
- Check budget, bid strategy and start and end dates against the approved spend for the month.
- Open the targeting settings: locations, languages, audience exclusions and any placements that were added.
- Verify the conversion event is the one your revenue reporting uses, not a proxy such as page views.
- Read every headline and claim against what you can substantiate, using our guide to FTC and ASA advertising rules for AI-generated content.
- Click every landing URL and confirm tracking parameters match your naming conventions.
- Approve with a capped budget first, then review the first days of data before scaling.
If a draft fails any step, send it back with a written reason. Those notes become the instructions you add to the Context Layer, so the next draft improves.
Frequently Asked Questions
Does Supermetrics let Claude spend budget without approval?
According to the November 3, 2026 release, new campaigns are paused by default until a person reviews and approves them. The release does not explain how per-account approval settings can be configured.
Is there evidence that AI-built campaigns outperform manual ones?
Not in this release. Its figures cover connector adoption and a reporting task at one agency, so they say nothing about campaign performance. Test results in your own account instead.
Who should approve AI-built campaigns inside a small business?
Someone who owns both the budget and the conversion definition. If those sit with different people, require both to sign off for the first few campaigns until the process is proven.
Approval gates only work when the person behind them knows what to look for. At Grow with Hafsa we help teams set up agent-assisted ad workflows, review checklists and testing plans as part of our AI marketing and performance marketing services, so automation speeds up the work without taking control of the budget.
The same approval logic applies to email: AI agent cold outreach and liability under CAN-SPAM and UK PECR covers what a reviewer should lock down before an agent sends.
Before approving a draft, also consider whether the agent read untrusted content; our article on prompt injection risk for AI marketing agents explains why.