You Just Invited AI Into the Marketing Department. It May Be Quietly Wrecking Your Brand.

A couple of days ago I was chatting with a friend who works in consumer goods.

He told me his company had just rolled out generative AI and put the marketing department on it — using large language models (LLMs) to crank out product copy by the batch. Thousands of pieces a week, replacing a dozen copywriters' salaries. He was grinning ear to ear as he told me. So proud.

I laughed too. Then I asked him: have you actually read any of that copy?

He paused.

"Not all of it."

I didn't push it. But my stomach dropped a little. Because I'd recently come across a freshly published business-school teaching case that was asking exactly this question — when enterprises fold generative AI into their marketing, are they really creating value, or are they quietly destroying it?

It's more complicated than you'd think. Let's talk it through.

What This Teaching Case Is Actually Asking

The case was designed for a business-school classroom, but the question doesn't stay in the classroom. It's a question every boss who wants to use AI for marketing has to answer.

It walks through four small scenarios across four industries: retail, FMCG (fast-moving consumer goods), luxury, and a B2B industrial-technology company. In each one, the company is using AI to do something different —

Some are doing personalized recommendation. Some are running consumer research on synthetic data. Some are letting AI produce creative design. And some are just letting AI generate marketing content straight up.

Four neat use cases. All sound lovely.

But the case lines them up side by side on purpose — and the question it's really pointing at is something else entirely.

Value Creation, and Value Destruction

And that's the heart of it.

What do I mean by value creation?

It's when AI saves you money, speeds you up, trims your costs. The work that used to take ten people a week now gets done by one person in half a day. The research report that used to take a research firm three months now gets cranked out by AI overnight.

Sounds fantastic, right?

But the case pushes back with the other half — AI is also destroying value at the same time.

Destroying what, exactly?

Your reputation. Your brand. The compliance risk that could land you in court.

Think about it.

A luxury brand — what does it run on? Decades of carefully built tone. That sense of restraint where everything is "not too much, not too little." Hand the copywriting to an AI with no training in that voice, and it will produce lines like "Big Clearance Blowout!" Just once is enough. The tone is gone.

A B2B industrial company selling tens of millions of dollars of equipment — why does the customer trust you? Because you're professional, rigorous, technically authoritative. You use synthetic data to spin up a "user persona," then craft your communications around that persona — but the persona itself is something the AI just made up. You're going to war following a fabrication.

A retail brand lets AI auto-generate thousands of product descriptions. Nobody reviews them. One of them flips the safety instructions backwards. Sold for three months. Then a safety incident.

That's value destruction. It doesn't happen slowly. It happens all at once, and it isn't reversible.

Those Four Scenarios Are Really Four Double-Edged Swords

Back to the four scenarios in the case. Let me break them down again for you — it'll be clearer.

Personalized recommendation. AI helps you push to each user the stuff they like to look at, and conversion rates climb. But the moment the recommendation goes a touch too far, the user starts thinking "how do you know what I was just thinking about?" — gets the creeps, turns around, and accuses you of invading their privacy.

Synthetic research. AI whips up a crowd of "virtual consumers" for you. You run interviews with them, test ideas on them — saves time, saves money. But these virtual people are, fundamentally, the AI's best guess based on existing data. Treat guesses as truth and every decision you make skews.

Creative design. AI hands you a dozen poster concepts in seconds. But every style is something the AI has already seen and recombined — not one of them is genuinely something "you haven't seen before." Brand uniqueness gets slowly diluted, round after round of "average."

Content generation. AI writes your copy, your social posts, your sponsored content. Piece after piece. But nobody cares about any of it. Readers can tell within two sentences: this wasn't written by a person. Then they quietly scroll past. And the warmth of the brand gets ground down to nothing, one post at a time.

See — every one of these has efficiency on the front, risk on the back.

The Hard Part Isn't Whether to Use It. It's How.

The smartest thing about the case is that it doesn't hand you a standard answer.

It just lays the four scenarios out and lets the students grow their own judgment framework inside the discussion.

What does that mean?

It means — there's no one-size-fits-all answer to this.

You can't say "never use AI, period." That's dumb. Your competitors are all using it. Don't use it and you die.

You can't say "AI for everything, full speed ahead." That's even dumber. One wreck and you can lose everything you spent the last decade building.

So what do you do?

The case doesn't spell it out. But after I finished reading it, I worked out a way to judge for myself. Let me run it by you — tell me if it holds up.

My Way of Judging: Ask Three Questions First

Question one: if this task goes wrong, is it fatal?

If a mistake just means a few missed sales, no big deal — let AI handle it. If a mistake means a product recall, a lawsuit, or headlines — that kind of work, AI only gets to do the draft, and a human has to sign off.

Question two: is the core of this task efficiency, or judgment?

If it's efficiency — like translating the same message into twenty languages, compressing a long doc into a summary, or turning raw data into a table — hand it to AI.

If it's judgment — like whether the brand should adopt this tone, whether this slogan is good enough for us, whether this user persona is accurate — that stays with a human.

Question three: is AI's output the endpoint, or the starting point?

If you ship AI's output as-is, you've handed the fate of your brand to an intern who hasn't woken up yet.

If you treat AI's output as a draft and let a human sharpen it from there — then AI becomes your force multiplier.

Strip these three questions down and it's really one sentence:

"Cheap" is never free.

Every bit of cost you save has already been booked as a risk on another ledger. The only difference is when that risk comes due.

Finally

Back to my friend.

He told me later that he read through that whole batch of AI copy from start to finish — and found three pieces where the "intended user" was wrong. It was a children's product, but the copy described it for adults. If those had gone out as-is, he might have been staring down a recall round.

He still uses AI. But everything AI writes now goes through him once before it ships.

I didn't say anything. But in my head I was thinking: this guy's starting to get it.

AI isn't the enemy, and it isn't a savior either. It's a colleague — wildly capable, but in serious need of supervision. Whether you can use it well doesn't depend on how strong the AI is. It depends on whether you, standing right next to it, know how to manage it.

Here's hoping you can keep yours in line.