AI Can Now Build Your Ad Campaigns. Who Approves Them?

For most of my career, making an ad was the expensive part. Someone wrote the copy, someone designed it, someone set up the campaign, and every one of those steps cost a day and a fee. In the last two weeks, two announcements quietly moved that cost close to zero.

On 28 October, Microsoft Advertising said it will now build an advertiser’s first Performance Max campaign for them, using their website and brand materials. A week later, Supermetrics said its Claude connector, used by more than 6,000 customers since February, can now create and edit campaigns on six ad platforms: Google Ads, Meta, Microsoft Advertising, TikTok, LinkedIn and Snapchat.

Put plainly, you can now describe a campaign in a sentence and have AI build it. The question that matters has moved. It is no longer who builds the ad. It is who approves it.

6ad platforms where a marketer can now ask Claude, through Supermetrics, to build or edit campaigns — every new campaign created paused until a person approves it.

Why I am writing thisOwners are already being pitched “AI that runs your ads while you sleep.” Some of it is real. The part nobody mentions is who is responsible when the AI says something you would never say.

What actually changed

Microsoft now builds the campaign. A new advertiser can go from sign-up to a Performance Max campaign built from their website, brand guidelines and product images. In the same update Microsoft made its Conversions API generally available, which sends sales and leads from your own systems straight to the ad platform instead of relying only on a browser tag.

Microsoft also lets you write the rules. In a pilot, advertisers using Performance Max or Search with AI Max can set up to 40 “messaging guidelines” — themes, claims, products and offers the AI should stress or stay away from when it writes your copy. Microsoft does not claim the rules guarantee compliance. They steer; they do not police.

Supermetrics made the human the switch. Campaigns Claude builds through its connector are created paused and do not go live until a marketer approves them. Edits to live campaigns can require approval too, account by account, and changes are logged so they can be undone. Its co-founder put it simply: “The goal isn’t to remove marketers from the process.”

When making the ad costs nothing, judging the ad is the whole job.

Why this matters more for some businesses than others

If you sell t-shirts, a clumsy AI headline costs you a few clicks. If you are a dentist, a law firm, a lender or a medical practice, it can cost you far more. These are professions with rules about what you may promise. “Painless.” “Guaranteed results.” “Lowest rate in California.” “We win.” An AI writing thousands of variations will, sooner or later, write one of those sentences, because they are exactly the sentences that perform.

Your licence does not care that software wrote it. Neither does your reputation. The ad ran under your name.

The uncomfortable partThe same speed that makes AI ads attractive is what makes an unchecked mistake expensive. A bad brochure reached a few hundred people. A bad ad set can reach thousands by lunchtime.

What has not changed

Forty years in, I can tell you the things that decide whether advertising works have not moved an inch. A clear offer. A landing page that answers the question the ad raised. A phone that gets answered. Measurement that counts customers rather than clicks.

AI makes the ad faster. It does not make a weak offer strong or a slow follow-up fast. If anything it exposes them sooner, because you can now spend money faster than you can learn from it.

What to do this month

1. Write your own rules before any AI writes your ads. One page: what you always say, what you never say, which claims your profession forbids, which offers are current. That page is your version of Microsoft’s 40 guidelines, and it works on every platform and with every agency, human or not.

2. Keep a person on the approve button. Whatever tool you use, new campaigns should start paused and a named human should approve them. If a vendor cannot tell you who approves what, that is your answer about the vendor.

3. Fix your measurement first. AI optimises toward whatever you tell it counts. If it counts form fills and not booked patients or signed clients, it will find you a great many form fills. Connecting real sales or appointments back to the ad platform matters more now than it ever did.

4. Spend the time you save on the offer and the follow-up. The hours AI gives back are worth most when they go into the parts it cannot do: what you are offering, and how fast you answer the person who responds.

Want a second pair of eyes? I review AI-run ad accounts for practices and firms in regulated fields: the rules, the approvals and whether the measurement counts the right thing.

Talk to Siamak

The long view

Every time a part of marketing became cheap, the advantage moved to the part that stayed hard. Printing got cheap, and the message mattered more. Websites got cheap, and trust mattered more. Ads are getting cheap now, and judgment matters more.

Let the machine build. Keep your hand on the switch.

Common questions

Can AI create and launch ad campaigns on its own now?
AI can now build campaigns. Microsoft Advertising builds a new advertiser’s first Performance Max campaign from their website and brand materials, and Supermetrics lets Claude create and edit campaigns on Google Ads, Meta, Microsoft Advertising, TikTok, LinkedIn and Snapchat. Supermetrics creates new campaigns paused, so a person must approve them before they run.

What are Microsoft Advertising messaging guidelines?
Messaging guidelines let advertisers set up to 40 rules about the themes, claims, products and offers that AI-written ad copy should emphasise or avoid. They are in pilot for Performance Max and Search campaigns with AI Max, and Microsoft does not claim they guarantee compliance.

Is it safe for a dental practice or law firm to let AI write its ads?
It can be, with controls. Regulated professions limit what an ad may promise, and AI-written copy can produce claims a practice is not allowed to make. Write down the claims you never make, require human approval before anything goes live, and review new ad variations regularly.

What should a business do before using AI to run its advertising?
Write a one-page set of messaging rules, make sure a named person approves new campaigns, and connect real sales or appointments to the ad platform so the AI optimises for customers rather than clicks.

Sources

Every factual claim above links to its source, because a claim about a fast-moving field that you cannot check yourself is not worth much:

AI can now build your ad campaigns. Who approves them?