Marketer, Do You Really Know How to Use AI?

A number I saw recently made me do a double take.

Carvana, a used-car retailer, uses AI to generate a personalized video for every single customer who visits their site. Not one video for everyone — one video per person. The visuals, the script, the recommended car models — all different.

How many videos?

1.3 million.

1.3 million videos, every single one unique. Think about it: if you relied on humans to write scripts, shoot footage, and edit, how many people and how many years would that take? Five years ago, the mere thought would have been impossible.

But today, if you peek behind Carvana's dashboard, those videos are still being generated in an endless stream.

That's what generative AI does in marketing.

It's Not Just "Writing Copy"

When most people hear "AI in marketing," the first image that pops into their head is: oh, use ChatGPT to help me write a few social media posts.

True. But that's only a tiny piece of the picture.

Spotify did something else entirely. They used AI to automatically translate podcasts into different languages. A show recorded by an English-speaking host can now reach listeners in Latin America, Japan, and the Nordics in their native language. The market just cracked wide open.

Think about what that means. Producing global content used to mean either hiring translation teams or re-recording everything — costs were staggering. Now? AI tore down the language wall for you.

So here's the real question: what can AI actually do in marketing?

Let me break it into three layers.

Layer One: Ready to Use, Right Out of the Box

The lightest layer — call it "off-the-shelf tools."

What does that mean? It's the stuff you can use by simply opening a browser and signing up for an account. ChatGPT helps you draft that first email; Adobe's Generative Fill swaps out a poster background in three seconds.

What's the advantage of this layer? Speed.

A marketing copywriter who used to write 10 product descriptions a day can now write 100. What do you do with the time saved? Work on things that actually need a human brain.

But this layer has a ceiling: the tools are generic. They don't know your brand, your customers, or your product's personality. What they give you is roughly the same as what they give your competitor.

Layer Two: Feed It Your Own Stuff

So smart companies take the next step.

They feed their own data into the model. What kind of data? Past customer conversations, purchase records, ad performance results, brand voice guidelines. Once the model learns all this, what it produces carries your distinctive DNA.

IBM has gone deep on this layer. They have a family of enterprise-grade models called Granite, specifically trained on data from legal, financial, and academic domains. The language in these fields has to be extraordinarily precise — there's no room for guesswork. A general-purpose model can't handle this work; you need a specialist.

McKinsey shared a case study. Kellogg's — yes, the cereal company — used AI to scan social media for trending breakfast recipes, looking at what people pair their cereal with. Once the AI had that data, it automatically generated creative assets and social posts. Which flavors are trending, which combinations spark the most discussion — all of it becomes instantly visible, and it all becomes ammo for the next ad campaign.

The core of this layer? Your data becomes your moat. Someone else's model might be better than yours, but without your customer data, what it produces will never be as precise as yours.

Layer Three: Reinvent the Entire Pipeline

Climb one more level, and you reach the heaviest layer.

You tear down the entire marketing chain and rebuild it from scratch, with AI running through every step. From the first moment a consumer lays eyes on your brand, to the moment they place an order, to after-sales support — all connected.

For example: a customer browses a pair of shoes on your website, doesn't buy, and leaves. The old logic: wait a few days and send a discount email. The new logic: AI instantly analyzes that person's browsing path, time spent, and historical purchase preferences, then generates a fully personalized message within minutes — even pairing it with an image they're more likely to find irresistible — and pushes it out.

The IBM Institute for Business Value (IBV) conducted a survey showing that more than half of CMOs plan to use their own company data to build a foundation model from scratch. Not minor tweaks — a full teardown and rebuild.

That's the difference across the three layers: use it off the shelf, feed it your data, reinvent the entire chain.

Why Is Marketing the Best-Fit Function for AI?

Think about it — marketing, as a function, produces data by its very nature.

Every ad placement, every social post, every customer service call, every click and every pause — it's all data. And most of it is unstructured. The messages customers leave on WeChat, the reviews they post on Xiaohongshu (China's lifestyle social platform), the comment barrages flooding livestreams — it's all text, voice, and images.

Traditional tools can't handle this messy, tangled input. AI can.

AI can sift through signals that look chaotic and extract patterns. A customer has browsed your maternal and baby product pages three days in a row and asked about coupons in a comment. AI flags it: this is a high-intent prospect — push a limited-time offer now for the highest conversion rate.

This kind of judgment used to be called "experience." Veteran marketers relied on gut instinct. Now AI makes the call on data — faster than you, more accurately than you, and without needing sleep.

The Truth Behind 67% and 86%

IBM and Momentive.ai jointly conducted a survey of CMOs across various companies.

67% of CMOs said they plan to adopt generative AI within the next 12 months. 86% said they must adopt it within 24 months.

That ratio is essentially the entire industry going all-in.

But here's what's interesting. What are most companies actually using AI for right now? Cost-cutting. Automating product descriptions to save headcount, automating customer service to save headcount, mass-producing creative assets to save headcount. It's all about "saving money."

Those genuinely using AI to drive innovation and find new growth? Rare.

It reminds me of something.

Every time a new technology wave arrives, everyone's first instinct is to "use it to cut costs." Cost-cutting matters, of course — but it's defense, it's the baseline. What truly separates the winners is those who use new technology to do things that were impossible before.

Did Carvana pull that off? 1.3 million personalized videos — is that a question of "saving a videographer's salary"? No. It's "something that was fundamentally impossible before, now made real."

Cost-cutting is defense. Building something new is offense. A company that only plays defense won't lose — but it won't win, either.

Micro-Segmentation: From "A Group" to "An Individual"

Another shift AI brings to marketing could prove even more profound than efficiency.

We used to talk about market segmentation by age, income, and geography. Women aged 30–40, tier-one cities, monthly income above 20,000. After slicing the pie, you'd run the same ad campaign for that entire group.

What can AI do? It segments audiences down to the individual level. Not a group — every single person.

So someone coined a term: "micro-segmentation."

What does it mean? AI examines a person's purchase history, browsing behavior, social preferences, and even their recent mood (inferred from the tone of their comments), then generates a piece of content exclusively for that one individual. Not choosing between Option A and Option B — selecting the best fit from ten thousand options.

Personalization at that scale is something humans simply cannot do. The human brain can't juggle that many variables.

But AI can.

Before You Dive In, Think Through Three Things

I've talked up a lot of benefits, but in reality, plenty of companies have stumbled. AI isn't something that makes you powerful the moment you buy a tool. There's something you have to get straight.

First: data quality.

AI's output depends entirely on what you feed it. Garbage in, garbage out. If your historical customer data is messy, biased, or incomplete, AI will amplify those flaws ten thousand times after learning from them. That's why many companies invest heavily in data cleaning before deploying AI, sometimes even hiring dedicated data engineers for the job. Cut corners here, and everything downstream is wasted effort.

Second: privacy.

Using customer data for personalization sounds great. But do you actually have the right to use people's data to train a model? Regulations differ across countries. Handle it poorly, and the fine is the small problem — losing trust is the catastrophe. Trust takes three years to build and three seconds to destroy.

That's why IBM, when discussing this topic, keeps hammering two words: transparency and explainability. You need to clearly articulate how your model was trained, what data it used, and how it makes decisions. If you can't explain it, customers won't trust you.

Third: brand consistency.

AI-generated content is sometimes perfectly fluent but reads nothing like "you." Your brand's tone, attitude, and personality — AI won't automatically absorb these. They require constant tuning. Take your eye off it, and you'll get an article whose style reads like a stiff corporate memo one day and an internet-obsessed teenager the next. Your customers will be thoroughly confused.

My Take

This field is evolving fast. Anything I write today might need updating in three months. But there's one direction I believe won't change anytime soon.

AI won't replace marketers. But marketers who use AI will replace marketers who don't.

That's not an empty platitude. Those 1.3 million Carvana videos weren't AI's own idea. Someone envisioned the use case, fed the data in, built the pipeline, and validated the results.

AI is an amplifier. The depth of your insight determines the magnitude of what it amplifies. Conversely, if you have no insight, all AI amplifies is noise.

So back to the original question: do you really know how to use AI?

It's not about whether you can use ChatGPT. It's about whether you've thought about where to start in your current business, which layer of tools to use, and what core problem you're solving.

That question deserves serious thought from every marketer.