How Does AI Actually Help Marketers Make Money? I Studied 10 Real Case Studies

A while back, a friend who works in marketing complained to me.

He said his boss told him to "do AI marketing" this year. He excitedly bought a suite of tools, used them for a month, and produced nothing.

I asked him: What do you think the problem is?

He thought about it and said, "The tools are fine. I just don't know what to do with them."

That hit the nail on the head.

Most articles about AI marketing are either too vague — all "reshaping the future" and "disrupting the industry" — leaving you with nothing actionable. Or they're too shallow — just listing ten tools and calling it a day, as if installing ChatGPT means you've made it.

What's missing?

Real cases. Real examples of someone actually using AI to do something specific, and the results they got.

I spent a few days digging into 10 case studies. There's a mom-and-pop tamale shop, a giant like IBM, a lingerie brand, and a startup that ghostwrites for CEOs.

And I discovered one thing: Nobody who's truly good at AI is using it to "replace humans." They're using AI to amplify themselves.

Let me walk you through them.

One Short Video, 22 Million Views

Let's start with my favorite.

There's a shop in Los Angeles called The Original Tamale Company. They sell tamales — a husband-and-wife operation, very small.

In 2025, their marketing manager noticed a viral meme spreading on social media. Most people would've just scrolled past it, but he asked himself a question:

"Can I use this meme to drive traffic to my shop?"

He fed the meme's context to ChatGPT and had it write a 46-second script. Then he spent 10 minutes with simple tools to finish the video. The content showed a person falling from the sky and crash-landing right at their shop's front door.

Funny, unexpected, and relevant to the food they sell.

Three weeks, 22 million views, 1.2 million likes. Foot traffic to the shop went through the roof.

What strikes me most about this case?

It's not how powerful AI is. It's this marketing manager's judgment: he knew which meme could connect to his business, and he knew how to quickly turn a trending topic into content. AI only did one thing for him: it made him fast.

While others were still debating "should we try this," he'd already posted the video.

This is the first truth about AI in marketing: It doesn't create opportunities. It amplifies the speed at which you seize them.

IBM's 1,600-Person Design Team — Why Do They Still Need AI?

Now let's talk about a big company.

IBM has a 1,600-person design team. You read that right — 1,600 people.

But in 2023, they partnered with Adobe on an ad campaign. They used Adobe Firefly to generate over 200 original images, which spun into more than 1,000 versions, deployed across social media in markets worldwide.

The results? Engagement was 26 times higher than their previous comparable ads. And 20% of those who engaged were C-level decision-makers.

Wait. A 1,600-person design team still needs AI?

Yes.

Because IBM needed to quickly adapt the same creative concept for audiences across different industries and regions. Having humans draw each one by hand just wasn't fast enough. IBM's VP Ari Sheinkin said something I think hits the mark perfectly:

"The pressure on enterprises to deliver omnichannel personalized experiences is unprecedented. Generative AI has given us a path to scale."

What's the takeaway for small teams?

You don't need a 1,600-person design department to produce multi-version, multi-audience creative content. This kind of work used to require hiring an ad agency. Now AI has lowered that barrier. One person with the right tools can do the work of an entire team.

A Skincare Brand Saw Conversion Rates Jump Nearly 4x

A.S. Watson is the world's largest beauty and health retailer. They built an AI skincare advisor.

Users come in, fill out a questionnaire, and upload a selfie. The AI uses computer vision to analyze your skin — type, tone, texture, 14 different metrics. Then it recommends a personalized skincare routine.

The result: Users who went through the AI advisor had a 396% higher conversion rate, a 29% higher average order value, and spent 4x more overall.

396%. You read that right.

Why so dramatic?

Because the biggest pain point in buying skincare is "I don't know what's right for me." At a physical counter, a consultant can look at your skin and guide you. Online? You're just guessing.

What A.S. Watson did was essentially take the in-store consultant experience and put it online — but faster and more accurate.

That's what a good AI application looks like. It doesn't solve a technology problem. It solves a user decision problem.

Wherever your users are hesitating, that's where you deploy AI.

Verizon's Approach Is Even Smarter

Telecom carrier Verizon did something in 2024.

They used GenAI to predict the reason for 80% of incoming calls, then routed customers to the appropriate service agents in advance.

The result: each customer spent 7 fewer minutes in-store, and they intercepted roughly 100,000 customers who were about to churn.

But there's a detail here that I think is critical.

Verizon didn't "use AI to replace customer service." They used AI to arm their customer service reps — putting the right information in the right hands at the right time to make the right call.

The improvement in customer experience came from humans being empowered by AI.

A lot of companies think "AI" and immediately think layoffs. Verizon's approach: first figure out how to make your frontline employees stronger.

A Content Team Compressed 20 Hours of Work Into 20 Minutes

Adore Me is a DTC lingerie brand. They faced a problem: product descriptions needed to be multilingual, SEO-optimized, and maintain brand voice. They simply didn't have enough people.

Their approach was clever: they built three AI Agents on the Writer platform.

One was dedicated to writing product descriptions, tuned to brand tone and SEO. One handled Spanish translations for the Mexican market. One wrote stylist note drafts for human stylists to refine.

The results: product descriptions went from 20 hours per batch down to 20 minutes. Stylist note writing time was cut by 36%. Mexican market localization went from months to 10 days. Non-brand SEO traffic grew by 40%.

20 hours to 20 minutes.

This is the power of Agents. What's an Agent? Think of it as "an AI employee that does one specific thing." You set the rules, feed it the materials, and it works tirelessly within that scope.

The content team of the future might not be one editor leading ten writers. It might be one editor with five Agents.

A CEO's LinkedIn Following Grew from 7,000 to 11,000

Vector is a company that makes graphics software. Their CEO, Joshua Perk, wanted to stay active on LinkedIn but didn't have time to write.

Their marketing lead, Jess Cook, did something clever. She pulled down every post, email, and speech the CEO had ever made, analyzed his word choices, sentence rhythms, and how he opened and closed — then fed all of it into a custom AI model.

Then she'd do a 15-to-30-minute interview with the CEO every week, asking a few questions: "What did customers say to you this week?" "What surprised you recently?" "What's the most common misconception you hear?"

She'd transcribe the interview recordings and feed them to the AI. The AI would generate 3 to 5 LinkedIn post drafts in the CEO's voice. An editor would polish them, and the CEO himself would give final approval.

Jess said something I really love:

"AI gets me 80% of the way there. Then 15 minutes of editing takes it to 99%."

The result? The CEO's LinkedIn following grew from 7,000 to 11,000, and demo requests from leads quadrupled.

What's the essence of this case?

AI won't have ideas for you. It only helps you turn existing ideas into content faster.

If you don't have anything in your head, AI will just produce empty words. If you have real substance, AI helps push it from an 80 to a 99.

Heinz Asked AI to Draw Ketchup, and It Blew Up

Heinz did something clever. They used DALL·E 2 to have AI draw "ketchup."

The prompts were things like "ketchup street art," "Renaissance-style ketchup bottle," "ketchup in space." They didn't even mention the name "Heinz."

The result? The bottles AI drew looked almost exactly like Heinz's classic design — the bottle shape, the label colors, the visual signature.

Then Heinz opened this up to the public, inviting people to submit their own prompts. They turned selected AI images into actual ads, social media content, and even limited-edition bottle designs.

850 million impressions. Social engagement 38% higher than previous campaigns. ROI was 25 times their media investment.

What's brilliant about this case?

Heinz didn't use AI as a tool. They used it as a mirror.

AI was telling the whole world: when people think "ketchup," what comes to mind is Heinz.

This isn't a stunt. This is using AI to run a stress test on brand equity.

You could try this too. Feed your category keyword to GPT or Claude and see what comes back. If the output has nothing to do with your brand, that's not AI's fault — it means your brand doesn't have enough recognition in that category.

Virgin Holidays Uses AI to Write Email Subject Lines

Virgin Holidays used to come up with only 2 to 3 subject lines per marketing email. Doing it manually was slow, and the results were underwhelming.

They started using Phrasee. After Phrasee thoroughly learned their brand voice, it automatically generated a large volume of subject lines and continuously learned from each campaign's data, getting sharper with every round.

Email open rates jumped by 2 percentage points. Sounds small? For a travel brand that sends massive volumes of email, those 2 percentage points translate to millions of dollars in revenue.

Here's an important lesson: Generic AI gives you linear returns. Purpose-trained AI gives you compounding returns.

Phrasee didn't just write good subject lines once and stop. It learned from data after every campaign, making the next round better. Once that flywheel starts spinning, it gets more valuable the more you use it.

HubSpot Turned "Segments" Into "Mind Reading"

HubSpot's old email nurture workflow looked like this: divide leads into broad groups like "marketing leads" and "sales leads," then send the same email to everyone in each group.

Then they brought in AI. The AI analyzed each user's form behavior and site browsing patterns to predict what they actually wanted, then matched them with the most relevant content.

Conversion rates rose 82%. Open rates grew 30%. Click-through rates grew 50%.

From "segment-based marketing" to "intent-based marketing" — this isn't a technology upgrade. It's a paradigm shift.

You used to send the same email to 1,000 people. Now you send 1,000 different emails to 1,000 people. Before, it was "I think you might be interested." Now it's "AI tells me you're thinking about this right now."

A Creator Used AI to Make a Video with 1 Million Views

Karen X Cheng is a content creator. She took on a branded partnership with Insta360.

When most creators get a brand deal, they just film a product review. She didn't.

She filmed herself passing through a series of "portals" in different settings, with each portal rendered in a different AI art style — watercolor, oil painting, sketch. She shot it on an Insta360 X4 360 camera, then used AnimateDiff to transform the footage into different animated styles.

The video racked up over 1 million views and 50,000 likes on social media. The comments section was full of people asking "how did you make this?"

Then she released her creative process. The process naturally showcased Insta360's features and capabilities.

The product placement was invisible. Because it wasn't an ad — it was a piece of art.

Karen's case tells brands one thing: partner with creators who are already using AI in their work. They know how to combine technology with storytelling, and their audiences are more receptive to new things.

One more thing — don't lowball creators just because "AI makes creation faster." Good creators using AI produce work that's better than what you'd get from a traditional team at ten times the price.

AI isn't a cost-cutting tool. AI is a force multiplier.

Nike Had Young Serena Williams Play a Virtual Match Against Peak Serena

Nike used AI to simulate a match between a young Serena Williams and peak Serena Williams.

This wasn't a game. It was a narrative tool Nike used to tell the story of Serena's 20-plus-year career. The video was live-streamed on YouTube, drawing 1.7 million viewers. Organic reach was 1,082% higher than Nike's previous organic content.

Nike didn't launch a new product. They used AI to do one thing — tell you a story that could never happen in real life.

This is inspiring for anyone making content. If you're wondering "how do I get more backlinks" or "how do I get media to cover me" — the answer might be: use AI to find an angle that media can't resist talking about.

After Studying These 10 Cases, I Identified 5 Patterns

Pattern 1: AI doesn't replace humans. It amplifies them.

In every successful case, AI wasn't the main character. The main character was the person who spotted the opportunity, made the judgment call, and executed. The Tamale shop marketing manager's eye for a meme, Jess Cook's ability to deconstruct a CEO's voice, Karen X Cheng's aesthetic sense for visual storytelling — these are the irreplaceable parts. AI just made them faster, more prolific, and broader in reach.

Pattern 2: Personalization at scale is no longer a luxury.

A.S. Watson's 396% conversion rate, HubSpot's 82% conversion lift, IBM's 26x engagement — the core is the same: turning "one-size-fits-all" into "personalized for every single person." Personalization used to require an army of people. Now it runs on data and algorithms.

McKinsey has a study showing that fast-growing companies earn 40% more revenue from personalization than slow-growing companies.

This isn't a question of "should we do it." It's a question of "when do we start."

Pattern 3: Train AI on your own data.

Vector fed the CEO's historical content to AI. Adore Me baked their brand voice into Agents. Phrasee continuously learns from each campaign's data — what's the common thread?

None of them just used a generic AI out of the box. They invested time in making AI "one of their own."

What generic AI gives you, anyone else can get too. Only AI trained on your own brand data becomes your competitive moat.

Pattern 4: The best AI is the one users don't notice.

Verizon's customers experienced "shorter wait times" and "problems getting solved." They had no idea AI was behind it. The Original Tamale Company's audience experienced "this video is hilarious." They didn't care whether AI was involved either.

Consumers don't care what technology you used. They care whether you actually helped them.

So don't make AI the selling point. Make it the infrastructure.

Pattern 5: AI focused on a specific task compounds.

Virgin Holidays' Phrasee gets more accurate the more it's used. HubSpot's intent prediction model gets stronger with each iteration. Adore Me's Agents sound more and more like the brand itself with every training round.

Generic AI tools give you one-time returns. Purpose-built AI tools give you returns that keep growing. That's the difference.

What Happens Next?

Let me share a few predictions.

Search is changing.

As people get used to finding information through ChatGPT, Claude, and Google's AI overviews, traditional SEO isn't enough anymore. You need AI platforms to think of your brand when your category comes up.

What does this mean? It means media coverage, industry articles, and word-of-mouth content are becoming more important — because AI platforms rely on these "authoritative sources" to judge brand relevance.

You can't just do website SEO anymore. You need presence across the entire content ecosystem.

LLM hype will peak.

The phase of unlimited AI hype will pass. What comes next is the "real returns" phase — which scenarios actually saved money, actually made money, actually improved efficiency. Companies that raised money on AI concepts without solving real problems will struggle.

Agents will start replacing Apps.

Gartner predicts that by 2027, mobile app usage will decline by 25% as AI assistants handle tasks that users used to need apps for.

What does this mean? Users may no longer open your website or your app. They'll tell an AI assistant "book me a flight" or "buy me something," and the AI assistant will make decisions for them.

If your product can't be understood, recommended, and invoked by AI assistants, you're invisible on a growing traffic channel.

One Last Thing

After studying these 10 cases, my deepest takeaway is this:

AI won't eliminate marketers. But marketers who use AI will eliminate those who don't.

The gap isn't in tools. The gap is in judgment — whether you can identify which scenarios are right for AI, how to train AI to become "one of your own," when to let AI take over, and when a human should make the call.

This kind of judgment doesn't come from reading a single article.

But if you found one case among these 10 that's relevant to your business, try it today — even if it's just using ChatGPT to write 10 email subject lines for A/B testing.

Take that first step. Everything after that gets clearer.