Your Media Plan Is Becoming a Machine Brief

There was a time when media planning involved a room full of people arguing about where to spend the client's money.

Someone wanted television. Someone wanted search. Someone had a very convincing PowerPoint explaining why the answer was social media.

Then came programmatic advertising.

The arguments didn't disappear. They just moved further away from the actual decisions.

Having spent years working in digital strategy and media, I've watched advertising become progressively more automated. We gained extraordinary targeting capabilities, better measurement and the ability to change campaigns while they were running.

We also created a system where very few people could explain precisely why an advertisement appeared where it did.

Now AI is taking that arrangement one step further.

And this time, the machine isn't simply buying the media.

It's increasingly helping decide what the campaign should be.

The brief is changing hands

For years, AI in marketing was largely presented as a production story.

More headlines. More images. More audience variations. More tests.

Useful, certainly. But rather unimaginative.

The bigger change is happening elsewhere.

McKinsey describes an emerging agentic advertising economy, where automated systems increasingly connect discovery, recommendation, purchasing and measurement.

That means machines are moving closer to decisions previously made by planners, buyers and strategists.

Which audience matters? Which message should they receive? Which placement deserves another impression? Which customer interaction is worth paying for?

These are no longer exclusively human decisions.

Check it. 👇

The question isn't whether AI can produce more advertising. It's whether anyone has told it when to stop.

The efficiency problem nobody wants to discuss

Digital marketing has always had an awkward relationship with waste.

We celebrate efficiency while generating enormous quantities of activity that may contribute very little.

Another impression. Another retargeting sequence. Another creative variation. Another auction.

The logic is familiar: if testing becomes cheaper, test more.

AI makes that logic almost irresistible.

But cheap production is not the same as efficient production.

The advertising industry is developing ways to measure its environmental footprint. Ad Net Zero's Global Media Sustainability Framework is one example.

That's useful progress.

But measuring the footprint of a campaign after it has run doesn't necessarily change the decisions that created the footprint.

And this is where my background in carbon accounting collides with my years in digital advertising.

One discipline asks whether the campaign delivered a return.

The other asks what resources were consumed to deliver it.

Neither question is sufficient on its own.

The real advantage isn't teaching AI to buy more efficiently. It's teaching it what isn't worth buying.

A different kind of media brief

Imagine two instructions.

The first: maximise conversions within a CHF 100,000 budget.

The second: maximise profitable conversions while limiting unnecessary impressions, avoiding redundant creative production, scrutinising supply paths and measuring emissions alongside commercial performance.

Those are different assignments.

The first gives the system a financial objective.

The second gives it a definition of responsible performance.

That doesn't mean every campaign needs a complicated environmental model. It means marketers should become more deliberate about the constraints they give automated systems.

How many variations genuinely improve learning?

How much frequency is useful before it becomes repetition?

Are we reaching new customers or paying repeatedly to interrupt the same people?

And can we demonstrate that the additional activity produces additional value?

AI will become exceptionally good at following commercial incentives.

The difficult part is deciding which incentives deserve to exist.

For marketing teams, that may be the most important strategic skill of the next decade.

Not writing better prompts.

Writing better rules.