AI Moved Into the Marketing Department: The Money Saved and the Cars Wrecked — You Can't Keep Pretending Not to See Either

A while ago, a friend in marketing grumbled to me: the boss has decreed that every campaign's imagery next year gets an AI-generated version first.

I said: and then?

He said: and then I can't sleep. Not because I fear losing my job — because I fear the wreck.

That sleeplessness, I think, is representative of the moment. So today let's lay it all out: how exactly has AI rewritten marketing? Is the money genuinely saved, or just moved from the left pocket to the right? And the brands that crashed — at which step exactly did they go wrong?

First, the Ledger

Marketing has never lacked data. What it lacked was the appetite to chew the data up and swallow it.

That's the appetite AI supplies. An insight report that used to take two weeks now takes an afternoon; a poster that used to go through twenty rounds of revisions now goes through twenty rounds of prompts. McKinsey put out a report: roughly 42% of organizations are already using generative AI in marketing and sales functions. SurveyMonkey's survey goes further: 88% of marketers use it every single day.

88%. Sit with that for a second. This is no longer a debate about "whether to use it" — it's a debate about "how well."

The market pie is growing too. The global AI market in 2025 is about $47.3 billion, projected to hit $107.5 billion by 2028. More than doubling in three years.

There are hard-cash examples as well. Klarna, the Swedish fintech company, has said that using AI to generate images and creative assets saves it roughly $10 million a year in marketing costs.

Ten million. Dollars.

And here's the interesting part: the bulk of the money Klarna and its ilk saved didn't come from layoffs. It came from fishing people out of mechanical work and pointing them at what AI can't do: judgment, taste, and asking the fatal question of an AI image — "does this look like us?"

But Every Coin Has Its Other Side

Savings stories told. Now let me tell you three wreck stories.

First, LEGO. A Ninjago promotional video generated by AI was executed so sloppily that consumers started yelling outright.

Second, Coca-Cola. In 2024 they remade their classic 1995 "Holidays Are Coming" Christmas ad with AI. Homage is homage, but social media wouldn't buy it: no human warmth — just a shell of sugar syrup.

The third is the roughest: Google. During NBC's 2024 Olympics broadcast, Google ran a 60-second spot, "Dear Sydney": a father uses Gemini AI to help his daughter write a letter to her idol, track star Sydney McLaughlin-Levrone. The intent was to flex. The audience exploded instead — words from the heart to your idol, and you let AI ghostwrite them? The backlash got so loud that Google pulled the ad.

And Toys R Us: at the 2024 Cannes Lions festival it screened its first brand video made with OpenAI's Sora. Viewers' one-line verdict: creepy. Less a brand story than a brand accident.

Have you noticed? Across these four cases, the crash point is astonishingly consistent.

What the audience resents was never AI itself — it's that gesture of "pretending to be heartfelt."

The Christmas ad, the letter to an idol, the story of a childhood brand — these are precisely the things that should never be outsourced to a machine. What you save is production cost; what you overdraw is other people's trust in you.

And there's an even harder ledger inside: artists' rice bowls. AI image generation has directly displaced a lot of illustration and design work. Companies want to save money — understandable; but people who have drawn for half a lifetime — on what grounds should they be "learned from" by machines for free? This ethics question has no standard answer right now. But pretending the question doesn't exist is definitely not the answer.

So What Now?

My judgment is simple: use AI as an employee; don't enshrine it as the boss.

Concretely, three things.

First, draw the division of labor. Drafting, reformatting, running data, producing first passes — hand those to AI. Judgment, creativity, and the brand's breath of life must stay in human hands. AI handles the hands; humans handle the heart.

Second, never dismantle the human review gate. Every piece of content about to go out must pass a human once: any bias? anything offensive? does it look like us? What the LEGOs of the world lacked wasn't technology — it was that one last look before publishing.

Third, transparency. If you used AI, say so, openly. Hide it, and once you're caught, the scolding will be ten times worse than if you'd been honest. Trust takes decades to build and one ad to collapse.

Back to the sleepless friend from the beginning. He later told me he'd worked it out: AI won't take his job — but a colleague who knows how to use AI will.

Crude words, sound reasoning.

AI is rewriting the marketing industry — that much is true. But it can't solve every problem, especially the parts that concern the human heart. Give the machine its machine work, and give people back to people.

That's how you keep eating this bowl of rice for a long time.