An AI Just Joined the Marketing Department: 7 Real Case Studies, Broken Down One by One

A while back, I stumbled onto a set of case studies — and the more I read, the harder it was to sit still.

They came from a company called DP6. DP6 is a digital marketing consultancy that helps brands make decisions with data. In late 2024, they ran an event called Marketing Data Science — which is exactly what it sounds like. There, Rafael Ennes, a senior Data & AI manager at the company, gave a talk on a single theme: how data and GenAI amplify potential.

The talk laid out 7 real ways GenAI is being used in marketing. Not concepts on slides — projects that DP6 and its sister companies under The Brandtech Group have actually run for major brands.

My verdict was one word: Impressive.

Today I'm taking these 7 use cases apart and walking you through them one by one.

1. You Ask, It Answers

What are marketers most starved of?

Not ideas. Data.

Picture this. You run the marketing department, and you want to know: which of our campaigns actually drove the most revenue for our flagship product line? The answer is sitting in the company's data warehouse. But you can't write SQL. So you file a request, wait for it to get scheduled, wait for the report. And by the time the report lands, the window for acting on it has already closed.

So what is a Reporting Chatbot?

In plain terms: you station an AI at the door of the data warehouse as your translator. You ask in everyday language, it goes in and translates, then comes back with answers as text, charts, or tables.

Just like that, the barrier to data comes down. Teams that don't speak data can pull their own numbers; decisions no longer wait for the weekly report — you ask, you get. As a bonus, the data analysts are freed too, no longer on call every day as every department's human query interface.

Data used to belong to the data analysts. Now data belongs to everyone who can type.

2. The Millimeter War on the Digital Shelf

Second — let's get closer to the money.

You've compared prices on Google Shopping, right? Tapped a product inside Amazon Ads results? Behind every single listing hangs a string of content: title, description, images, attributes of every kind.

Shoppers can't see your warehouse. This is all they see.

Which makes this content worth taking seriously. The trouble is, scale turns it into a disaster. A few hundred SKUs, and every platform has its own rules — review and fix them by hand, one listing at a time? See you next year.

That's the job Feed Optimizer does: GenAI automatically audits those titles, descriptions, images, and attributes — optimizing what needs optimizing, adjusting what needs adjusting. And it plays by the rules: the specs of Google Shopping, Amazon Ads, and the rest — not a single one violated.

Both sides win. Shoppers see listings that actually speak to them — clicks, conversions, and sales climb. Sellers save hours of manual review and get products live much faster.

3. The Bad Review Is the Best Product Manager

Third — my personal favorite.

Your app lives in the app stores, and every day users leave reviews there. Praise, rage, suggestions — and the occasional rant that comes with a free redesign proposal attached.

How did most companies handle these reviews? Nobody read them. Or one poor soul got assigned to read them, went numb doing it, squeezed out a monthly report — which nobody read either.

And yet it's pure gold in there.

This use case is called App Review Benchmark: GenAI turns thousands of messy complaints into structured answers. What are users angriest about? Getting lost in the navigation, screens that take forever to load, pricing that stings? Which issues hurt users most and deserve to be fixed first? You can even line the reviews up against your competitors' and see where each product's bad reviews cluster.

User feedback used to lie scattered across a comment section nobody could dredge. Now it files itself into line and lands on your desk, sorted by damage, highest first.

A bad review is the best product manager. The only question is whether you're listening.

4. So How Is AI Introducing You to Everyone Else?

Fourth, the direction flips. The first three were about how you use AI. This one is about how AI talks about you.

Have you noticed? Over the past couple of years, when people hit a problem, the first move has shifted from opening a search box to just asking an AI. And whoever the AI mentions in its answer is the one who may win that business.

Which raises the question: when a user asks an AI "which one should I buy in this category?", how is that AI introducing you? Are the selling points it lists the same ones you're trying to land? Compared with your competitors, whose corner is it fighting in?

Share of Model is a solution built by Jellyfish to answer exactly that. It uses GenAI to analyze what the major models "know" about your brand: which of your attributes they've retained, which of your competitors' attributes they've retained too — put the two side by side, and the gaps show.

Once you can see it clearly, you know where to steer your content and positioning — and you get the chance to bend the version of you inside the AI's head, little by little, into the one you want.

People used to fight for position on a search results page. Going forward, there may be one more thing to fight for: share of the AI's memory.

You've probably heard of Share of Search. Share of Model I was seeing for the first time, and it floored me — sheer brilliance.

5. Giving Creative a Rearview Mirror

Fifth, back to the creative itself.

When a campaign wraps, what does the debrief cover? Click-through rate, conversion rate, ROAS (return on ad spend). Everyone looks at the numbers. But hardly anyone asks: why was this ad actually good? Or, why was it bad?

This use case is called Creative Insights: GenAI goes back through past campaigns and matches the creative assets' many attributes — one by one — against what the placements called for at the time and what the results actually were.

Once the matching is done, the next campaign's brief is no longer a guess — it grows out of what actually worked last time. Which visual elements, which phrasing, work best on which audience — the team has the ledger. When the creative matches the audience's taste, ROAS naturally has an easier time climbing.

In plain terms, this is a rearview mirror for creative. A car without one still drives — it just keeps hitting the same pothole, over and over.

6. Creative Finally Has an Assembly Line

Sixth tackles a particularly painful problem: capacity.

For big brands running campaigns, nothing is scarier than the word personalization. 10 audience segments, 5 markets, 3 languages — how many versions does one creative split into? Do it by hand and capacity buckles instantly.

Pencil Pro automates that production line with GenAI: text, image, and video — several classes of generative models working together, batch-producing creative assets to brand guidelines.

Capacity goes up without quality slipping. Shorter delivery cycles make campaigns more agile. And each audience sees a version that suits them better, instead of the same ad blasted at everyone.

Some worry: does an assembly line make creative cheap? My take: what gets copied in bulk is the asset, not the insight. Insight is always scarce.

7. The Foundation: First, You Have to Feed It

The last one is the foundation under all the others.

You might ask: these tools all sound great, but where do big companies actually get stuck trying to use them?

Stuck at the feeding.

Here's an analogy. An AI is like a brilliant intern on day one — mind racing, but knows nothing about your company. Put them straight to work and they'll make things up with total confidence.

What's the reality for a major brand? Creative ships to dozens of markets, each with different compliance requirements and a different performance history. And all of that information lies scattered across documents, databases, and spreadsheets — even humans can't always gather it all, let alone an AI.

Brandtech Brain exists to fix this. Technically, it's RAG (retrieval-augmented generation), fine-tuning, prompt engineering — that family of techniques. In human terms: gather the company's scattered knowledge into a form AI can read, then, while the AI works, feed the relevant, current, company-specific pieces back to it in real time.

The effect is immediate.

There's less making things up, because what it generates now stands on your own data, not free association. Personalization gets stronger, because strategic business information now enters the model — customization at scale becomes possible. And there's less hand-feeding and back-and-forth error correction — which is what finally makes scaling hold up.

In one line: GenAI's ceiling depends on what you feed it.

Finally, a Parting Thought

Look back at these 7 use cases and you'll notice they're all saying the same thing.

GenAI hasn't turned marketing into a different industry. What it does is take the parts of marketing that were always make-do — "too labor-hungry, so we settled" — and pull them apart and rebuild them, one by one.

We made do with half-seen data, made do with half-heard reviews, made do with copy-paste creative, made do with scattered knowledge. Now we don't have to make do.

One point from that talk stuck with me: the real value of an AI project only surfaces after it scales into production. The excitement of a demo doesn't count.

I couldn't agree more. The cases are laid out right here. The tools are all here too.

The only question left: when is your marketing department going to redo those "make-do" parts?

Here's to getting there first.