The Song No One Picked in Your Ad

A while back, I heard something that made me sit and stare at the ceiling for a long time.

An independent ad agency called Broadhead made a video ad for a client that makes wearable devices.

The process went exactly as usual: concept, shoot, edit, deliver, launch.

So what happened?

Someone on the team noticed the ad now had some background music — an AI-generated score.

The dozen or so people on the team had never picked that track. No one knew when it slipped in.

So where did it come from?

The answer: the platform added it.

That's right. The platform added it on its own, without so much as a heads-up to anyone.

Now, before you say "it's just some music, what's the big deal" — let me explain why this matters.

Because this is exactly what I want to talk to you about today. What does it mean when AI acts on its own?

What does "AI acting on its own" mean?

Plainly: it means the platform can ignore you and go change your ad all by itself. Before, it didn't work that way.

It used to be different. When you ran an ad, changing the copy, swapping the visuals, tweaking the layout — that was all done by people. If you didn't nod, nothing moved. You kept an eye on it, and it stayed out of your way.

Now it's changed. The ad backends of platforms like Meta and Google are stuffed with automation settings. From scoring a video with music to giving it a bit of "visual polish," the moment you flip a toggle, the AI just runs with it — no need to wait for you.

In one sentence: before, it was you telling AI what to do; now it's AI deciding for you.

What's the difference?

It comes down to a single line: before acting, does it ask — or doesn't it?

Here's an analogy. Say you hire an editor. In the first way of working, he cuts the video and shows it to you, and you ask for changes until you're happy — that's called consulting you. In the second way, he cuts it and posts it online directly, without giving you so much as a heads-up. If you were dealing with the second kind, wouldn't a chill run down your spine?

Giving AI power and giving AI boundaries are two very different things.

So you might ask: does that mean the feature is completely useless?

No, it's useful. It saves manpower, it's fast, and it's cheap. No one can pretend not to see those benefits.

The problem comes down to four words: it acts without asking.

So the real question to think through is this: which things do we hand over to AI, and which must stay for you to call?

How do you protect yourself? Here are three ways. They're all plain common sense, but they work.

First, lock down the boundaries before launch. Go through the toggles in the backend one by one: this one can add music, that one must be confirmed by a human. Write it down clearly, make it explicit. Don't take the easy way out and just switch everything on.

Second, keep a gate at the critical points. Anything that touches the brand or touches money must force a human approval step. You want AI to be fast, but you can't let it run out of control.

Third, go back and check after launch. Every so often, look at what it's changed and whether anything it shouldn't have touched got touched. It's the same logic as keeping accounts: if you never check the books, you won't know where the money went.

Plainly put: AI is the employee who gets things done, not the boss who calls the shots. It can run fast, but the hand that makes the final call has to be yours.

Look back at that ad. Later, Broadhead went through it, tightened the boundaries, and added another human gate. That "self-starter" agent had, so to speak, been put on a leash.

Technology only gets stronger. You can't stop it, and you shouldn't.

But the stronger it gets, the more important it is to remember one thing: the only thing that should always be in charge is you.

Don't wait until one day you open the backend and find a song no one picked suddenly playing in your ad.

That's my view — and I may not be right. But I hope you never hear that stranger's song in your own ad.