This week: Over the past few weeks, paid media has received another wave of automation.

Search matching. Creative generation. Landing-page selection. Bidding. Targeting. Campaign management.

The proposition from the platforms is increasingly simple:

Give the system more freedom and let it find the result.

That can remove an enormous amount of manual work.

But every decision we hand over creates another question:

Google's AI Max can expand beyond existing keywords, generate or customise ad text and select different landing pages based on the searcher's apparent intent.

New Google Search campaigns now have AI Max selected by default, although advertisers can change individual settings and controls.

TikTok is moving further.

Its Agentic Hub describes AI Skills capable of autonomously managing campaign budgets, rotating fatigued creatives and adjusting targeting using campaign data.

These are not all the same technology.

But they point in the same direction:

Advertising automation is not new.

Smart Bidding has been making auction-level decisions for years. What is changing is the surface area of the decisions being automated.

The system may increasingly influence:

Who receives the ad → what they see → where they land → how much is bid → what gets optimised next.

The advertiser still sets important boundaries. But inside those boundaries, the platform is being given more room to interpret intent and decide how to respond.

That changes the media buyer's relationship with the account.

The question becomes less:

“Which adjustment should I make?”

And increasingly:

“Which decisions am I comfortable allowing the system to make for me?”

Audit automation by decision, not by feature name.

Instead of asking whether AI Max, Smart+, Advantage+ or another automation is enabled, ask:

What can this system decide that a person previously decided?

Then identify which of those decisions are reversible, which are observable and which still require human approval.

Both report:

New-customer CPA = ~$42.

For the first business:

First-order contribution margin = ~$13.
Repeat purchase is low.

A $42 acquisition cost could destroy the economics of the sale.

For another business:

Customer LTV = ~$848.

Gross margin = 70%.

Retention is strong.

A $42 acquisition cost could be excellent.

The number is identical.

Its meaning is completely different.

AI can make a logical decision against an incomplete objective.

If CPA rises from $37 → $42, the system may read that as worse performance.

But what if those higher-CPA customers:

Then the higher CPA may actually be better for the business.

The issue is not always bad optimisation.

Platforms can use richer conversion values, offline outcomes and first-party signals.

The problem starts when important business information never reaches the system.

AI ON / AI OFF.

It is:

How much authority should this system have under which conditions?

THE SIGNAL

The shift is not that AI is replacing media buyers.

It is that automation is gaining more authority while business context still sits outside the ad account.

Platforms may know CPA, clicks and conversions.

They may not know margin, LTV, inventory pressure or why one customer is worth more than another.

So the risk is not always bad optimisation.

It is good optimisation against the wrong objective.

Operationally, teams should test automation by decision type, feed richer business signals where possible, and keep approval rules around high-impact changes.