When AI Grows Into Marketing: The Victories You Can't See

I had coffee the other day with a friend who runs a consumer-goods business.

He didn't complain about how tough business was. Instead, he threw me a line: "The hardest part this year isn't how much the budget got cut. It's that the money's gone and you can't even hear it land."

I froze for a second.

That hit something. Media has fragmented to the point where the old playbook — throw one TVC (TV commercial) at the wall and watch it ripple — just doesn't work anymore. But his next half-sentence was even more interesting: "The competitors who are really surging aren't really running ads anymore. They're tuning their AI."

Tuning their AI.

Not in the "spin up ChatGPT, write two lines of copy" sense. The sense of building machine-learning capability into every link of the marketing pipeline.

I went home and dug through the case studies from the past couple of years. After reading them, I was left with one feeling: AI has been reshuffling the deck for a while now — the noise is just buried deep.

Let me walk you through a few sets of stories. Once you've heard them, you'll probably understand how far this game has gone.

Group One: AI Made the Expensive Business of "Making Content" Cheap

Let's start with the most direct — and the most easily underestimated — use case.

Cut costs. Speed up. Compress the kind of assets that used to require hiring an agency and grinding for half a year down to a few days, or even a few hours.

Klarna — the Swedish buy-now-pay-later company. They did something pretty ruthless: they plugged their entire marketing production line into generative AI.

The result?

Image production cycle compressed from about 6 weeks to 7 days. In three months they cranked out over 1,000 custom images. Over the course of a year, about $10 million showed up on the books — direct savings, the bulk of it from cutting out external agencies and vendors.

Wow.

Behind that number hides a counterintuitive truth: a lot of people think what AI saves is "labor." But the real bulk is the "friction cost of external collaboration." That whole agency workflow — brief, quote, schedule, three rounds of revisions — got flattened overnight.

A similar script replayed itself at Mango. The Spanish fast-fashion brand built an entire campaign for its teen line, "Sunset Dream," shot entirely with generative AI. The process is interesting: they shot the real clothes first, then used AI to grow the models, the scenes, the whole mood. The clothes are real; the image was grown by AI.

Why do it that way? Because they treated this campaign as one capability puzzle piece in their 2024–2026 strategy — the goal being to make "AI growing into the production line" actually run end to end. The PR spectacle was a by-product, not the point.

The winners aren't the ones who treat AI as a gimmick. They're the ones who turn it into an assembly line.

Group Two: AI Turned Consumers Into "Creators"

Cost-down is about saving your own skin. The second playbook is more ruthless — it gets your users to do the marketing for you.

Burger King ran a "Million Dollar Whopper" contest. The rules aren't complicated: users build their own burger by picking ingredients through the app or a mini-program. And then? AI instantly generates a poster-grade finished-product shot of your burger, plus a custom ad jingle AI composed just for it.

You finish building a burger and you've got a finished asset in your hand, ready to post to your social feeds.

Think about it — what's always been the biggest pain point of UGC? Users don't shoot it well. BK's move crushed the "barrier to participation" down to the floor.

Warner Bros. used the same move from another angle to hype up Barbie. They partnered with PhotoRoom on a selfie generator: upload a photo, and in seconds you get a movie-poster-grade "Barbie version of you."

How crazy are the numbers? After launch: over 13 million cumulative uses, and around 13 million social shares.

Every user became an ad machine with a built-in distribution channel. And what they posted looked the way your official aesthetic intended — templated, so even at their worst, they couldn't look that bad.

The most interesting one is Cadbury's Diwali campaign in India.

They got Bollywood superstar Shah Rukh Khan — but not to front one unified ad. They used deepfake plus voice synthesis so that small shop owners could type in their store name and get a video of Khan personally calling out their shop name as an endorsement.

More than 2,500 small merchants — each one got their own, Bollywood-star-studded, exclusive ad film.

Reach: 140 million people.

That's powerful.

The shared logic of this playbook: hand users the tools, and turn them into content factories. Creativity is no longer scarce — your users will come up with 2,500 variations for you.

Group Three: AI Pulled Off "I Get You" Without Leaving a Trace

Groups one and two solved "production" and "distribution." Group three is the piece that actually holds up the business model: personalized recommendation.

The granddaddy of this space is Amazon. As far back as 2003, they made the recommendation engine their core engine: your browsing, your cart, your orders, your time-on-page — all fed into machine-learning models that reverse-engineered "what you're most likely to buy next."

This logic was copied by countless others, but two cases pushed it to "textbook" level.

Netflix.

Their recommendation system drives 80% of viewing on the platform.

80%.

In other words, when you open Netflix and scroll past a show three seconds in, that show was filtered for you by AI. And they're doing something even finer: the same show gets a different poster depending on who's looking. If you lean toward romance, the poster plays up the love story; if you're an action fan, it foregrounds the explosions. Just this one move lifted click-through rates by 20–30%.

Even better — AI also took local shows like Money Heist and Lupin, originally only big in Spain and France, and precisely pushed them to fans of the same taste all over the world. Reverse cultural export, powered by machines recognizing "no matter what country you're in, if you have this taste, you'll like this."

Spotify went even further.

They turned "I get you" into an annual cultural ritual — Spotify Wrapped. Every end of year, every user gets their own personal "year in listening" report: your favorite artists, how many minutes you listened, your music personality type.

Over 60 million shares worldwide every year. One person posts it, three or five friends follow suit, and the snowball rolls.

On the ad side, their Ad Studio runs on the same logic: based on your mood, device, time of day, and listening habits, it generates different audio ads in real time. Ad recall hit 2.7× that of non-personalized versions.

When the data gets accurate enough, "recommendation" stops being a sales pitch and becomes a service.

Group Four: AI Is Also Redefining "Brand Trust"

The first three groups are stories about business efficiency. But this AI blade can swing even deeper — it cuts straight into brand values.

Dove was one of the first brands to turn "AI ethics" into a marketing issue.

They noticed something: the "default output" of generative AI tends to reinforce stereotyped, unrealistic beauty standards. For instance, if you write "beautiful woman" in a prompt, AI will most likely hand you a fair-skinned, slim, large-eyed Western European face.

So Dove did two things.

First, they published a Real Beauty Prompt Playbook — a guide teaching creators how to write prompts to get more inclusive, more realistic image outputs. They handed people a tool, not just a slogan.

Second, they publicly committed: they will never use AI to simulate the image of a "real woman."

Together, these two moves turned an "ethical stance" into a "differentiated asset." In an era when AI content is flooding everything, "this is real" itself becomes the scarce commodity.

Heinz took another wild path. When DALL·E 2 first came out, they ran a minimalist experiment: let users type in any ketchup-related prompt they wanted, and see what AI would draw.

The result was astonishing: no matter how you wrote the prompt, the bottle AI drew by default looked like Heinz's signature long-necked glass bottle.

850 million impressions.

Why did it blow up like that? Because it struck a counterintuitive nerve: when even the machine's subconscious treats you as the synonym for the category, how deep must your brand moat already be.

Nike's "Never Done Evolving" went even more artistic. Serena Williams was about to retire. Nike, with AKQA, used AI to train two virtual versions of her — the 1999 one (when she won her first Grand Slam) and the 2017 one (when she won the Australian Open) — and had them play a match against each other across an 18-year gap.

1.7 million people on YouTube watched this match that never happened. Organic views, compared to Nike's usual content, were up 1,082%.

Honoring a legend used to mean cutting a retrospective film. AI lets you re-stage "legend" itself.

So, Who Are the Real Winners?

The stories end here.

You'll notice one thing: none of these four groups wore "we use AI" on their sleeves. None of them treated AI as the selling point itself.

Look closely, and every winner actually did one specific thing to the fullest:

• Klarna erased the friction of external collaboration;
• Burger King handed users creative tools so they'd produce ad assets themselves;
• Netflix turned "I get you" into the default experience;
• Dove drew a bottom line for "real" that nobody else dared to touch.

The real winners of AI marketing aren't the ones who adopted it earliest. They're the ones who first figured out "how AI reshapes my relationship with consumers."

My friend — the one griping that "the money's gone and you can't hear it land" — his problem isn't the budget. It's that he's still playing the 2018 playbook: one TVC, a few in-feed ads, a couple of KOL posts on Xiaohongshu (China's Instagram-like lifestyle platform), and then sitting in the back office staring at ROI.

That playbook's pipeline is broken. From the moment a consumer sees the ad, gets interested, decides, buys, and shares — there's a funnel between every step, and every step is leaking. The real value of AI is welding this funnel back into a closed loop.

Whoever welds this chain together first leaves their competitors no time to react.

Lexus's ad — the world's first with a script written by IBM Watson — was called "Driven by Intuition." They had AI study 15 years of award-winning car ads, find "what narrative arc moves people most," and then let Oscar-winning director Kevin Macdonald shoot it.

After-the-fact research showed the share of people who perceived Lexus as an "innovative brand" rose 13%.

This ad itself isn't the endpoint. It's a signal: when even AI can learn what "moves people," the core question of marketing stopped being "how much budget do I have" a long time ago. It's "how well do I understand the user."

Back to that coffee that afternoon.

I told my friend at the end: "Your biggest competitor this year probably isn't the brand next door. It's the old mental model in your own head — the one that says 'AI is just a tool.'

It stopped being a tool a long time ago. It's a new worldview."

May you see it clearly, one step ahead of everyone else.