AI Can Tell You What's Wrong. It Still Can't Tell You What's Next.

I'm supposed to know what's next. It's in the job description. This week, mostly, I didn't.

I went looking for clarity in my feeds and found an argument instead. Half of Cannes was convinced AI is the future of creativity. The other half wanted everyone to step away from the machines and make something that felt human. Both positions came out of the same festival, sometimes the same stage, on the same afternoon.

Then I did something I've never done to figure out where advertising is going: I read a papal encyclical. In May, Pope Leo XIV published his first, Magnifica Humanitas, which argues that AI must be "disarmed" and steered toward the common good. I've read a lot of trend reports. This was a new one.

By the end of the week the industry had settled on a tidy conclusion. Human creativity matters more than ever. Authenticity is the moat. Protect the humans. The Cannes juries said it, the Pope said it in his way, my feed said it on a loop.

I don't buy it. Or rather, I think it's true in a way that's useless.

"Protect human creativity" is the comfortable thing to say, and it loses every argument that actually matters. It loses to the spreadsheet, because AI is faster and cheaper and the pressure to use it isn't going anywhere. No CFO has ever been moved by the word "human." And it loses on its own terms, because "human creativity" is far too broad to mean anything. Most of what people make in this industry is also competent and forgettable. A bad ad made by a person is still a bad ad. Made by hand is not a quality guarantee, and it is not a strategy.

So if the line isn't human versus machine, what is it?

The work that breaks through has two things going for it, and neither is that a person touched it. It's culturally precise, tuned to a specific moment with a specificity that feels almost rude. And it carries emotional capital: it's willing to make you feel something risky, not just notice something clever.

Old Spice took a product people associated with their grandfather and made it the most talked-about creative of a Super Bowl. Domino's said, on the record, that its own pizza wasn't good enough. When SickKids and Cossette launched VS in 2016, the campaign turned a children's hospital from victim into fighter and made a lot of people uncomfortable on purpose; it ran seven years, raised $1.7 billion, and had donors tattooing the logo onto their skin. This year, Andre Is an Idiot, a documentary about a man's colon cancer, was sold less as a film than as a reason to book a colonoscopy, using everyday objects shaped like a backside. None of these survive a focus group. None of them are things a model recommends, because a model recommends what has already worked.

You can see the same instinct in who brands choose to bet on. This spring, Nothing made Charli XCX its first global brand ambassador and a shareholder. Its CEO was blunt about the logic: the tech industry had spent a decade making everything quieter and more uniform, and she had spent hers doing the opposite. They weren't buying reach so much as unpredictability, the exact quality an algorithm exists to smooth away, and the same one that made her a star in the first place.

What all of these have in common is harder to admit. You could not have measured any of them in advance. Whatever made each one land was invisible to the dashboard until after it had already landed.

That is the real limit of AI, and it's also the opening it leaves. AI is a machine for the measurable. It's extraordinary at the part of the job that can be counted, predicted, and optimized, which is most of the job and a growing share of it. What it can't do is move first on something the data hasn't seen yet. It can't decide that a colonoscopy joke is the right way to sell a film, because nothing in the data says so until it does.

None of this started with AI. We spent the last decade building tools to remove exactly that kind of guess, and we called it rigor. The obsession with data is older than the technology. It came from an ordinary, very human impulse to minimize risk, to never have to defend a call we couldn't prove. AI didn't make us cautious. It handed the instinct an engine and offered to do the measurable part for almost nothing.

Which leaves us the one skill we automated ourselves out of practicing: betting on the immeasurable. It's the appetite we've been trained to distrust, the call we can't justify yet and the dashboard only confirms in hindsight.

The predictable answer is to protect human creativity. The real one is the nerve to make that bet: cultural relevance, emotional capital, the punch a spreadsheet never sees coming.