90% of marketing leaders are doubling down on generative AI — but very few actually know how to use it

A while back, a friend of mine who runs an e-commerce shop invited me to dinner.

All through the meal, he kept his eyes glued to his phone. I asked what he was doing — couldn't he even eat properly? He said he was reviewing product copy written by AI. His store carries more than three thousand SKUs. It used to take two editors a full month of heads-down writing; now AI spits out all of it in a single afternoon, with the editors just reviewing everything at the end.

He looked up and asked me: is marketing, as a trade, about to be turned upside down?

Turned upside down is a stretch. But there's a set of numbers that might make you sit up a little straighter: in a survey of marketing leaders, 90% said they plan to increase their investment in generative AI over the next two years.

That isn't some company's promotional line. It's the consensus marketing leaders themselves arrived at.

Which raises the real question.

With this many people rushing in, what exactly are they rushing toward? Is it convenience? A trend to follow? Or is there a real return to be counted?

I've turned this over in my head for a long time. Let's talk it through today.

First, let's define our terms

What is generative AI?

You might say: isn't that just ChatGPT?

Close — but it's more than that. It's a class of AI that can create something out of nothing: you give it instructions, it produces output. Text, images, video — all of it works. Writing product copy, designing posters, editing short videos, replying to customer messages — it can lend a hand with all of it.

The previous generation of AI was a true-or-false machine — judging whether an email is spam, judging whether this person is likely to buy. Today's generative AI is an essay-question machine: you set the question, it writes the answer.

That difference matters enormously to marketers.

Because more than half of a marketer's daily work is essay questions.

What is it actually good at?

Let me break down three of the most practical battlefields.

The first: content production.

No pain is more familiar to marketers than the blank page. Staring at the screen, cursor blinking, sweating out a product description for an entire afternoon.

Now? A first draft comes out in seconds. You're no longer starting from zero — your job is to revise: reshape it into your brand's voice, strip out the inaccurate claims, edit until it has a human touch.

McKinsey has run the numbers: marketing teams using generative AI for content can see productivity gains of up to 40%.

40%. Not from working overtime — from handing off work like writing product descriptions and summarizing customer reviews.

My friend's three thousand SKUs run on exactly this logic. AI takes over the grunt work, and the editors only guard the key new products that genuinely need storytelling.

The second: customer engagement.

Have you browsed Sephora's online store? Its AI assistant can recommend the right shades based on your skin tone and what you've bought before. Airlines are using it too — flight updates, baggage issues, AI answers directly, no more waiting on hold.

What makes these tools powerful is that they know you. A returning customer walks in, the AI knows what they bought last time, and the recommendations adjust accordingly. Done well, what the customer feels isn't being fobbed off by a machine — it's being taken seriously.

The third: ad delivery optimization.

What was the old rhythm of media buying? Run for a week, pull the data, hold a meeting, adjust the creative. Now AI watches the dashboard in real time: this headline isn't working — swap in the next one immediately; this audience is responding well — tilt the budget over right away. A loop that used to take days to close now closes in minutes.

Those are the three battlefields. Notice what they have in common?

What AI does is all "quantity" work. Only when humans are freed up can they do the "quality" work.

But everything has a flip side

If I stopped after all those benefits, this would be an advertorial.

Generative AI has a pile of real problems, and any one of them can wreck an otherwise brilliant campaign.

The most famous one is called hallucination.

What's a hallucination? It's AI making things up with a straight face. You ask it to write product copy, and it invents a feature the product doesn't have at all; you ask it to cite data, and it fabricates numbers down to plausible decimal places. The tone is so confident it never gives you a reason to doubt it.

Gartner ran a survey: nearly 70% of organizations experimenting with generative AI rank hallucination as their number-one concern. Almost seven in ten. That isn't paranoia talking — it's people who've already been burned.

The second problem: bias.

AI learns from the data it's fed. If the data hides stereotypes, it learns stereotypes. Job-ad wording quietly skews toward certain groups; in the generated images, some groups are always absent. Nobody intended any of it, but the harm is already done.

The third problem: privacy.

Personalization is what AI does best, and personalization runs on customer data. How much you collect, how you store it, how you use it — every step is a minefield. Europe's GDPR and California's CCPA regulate exactly this.

Consumers are digging in too. In Cisco's 2024 privacy survey, 92% of consumers said they want more control over their own data, and nearly 60% have already switched brands over privacy concerns.

Nearly six in ten. Using AI for personalization while scaring your customers away — that's a bad trade.

The fourth problem: copyright.

These models were trained on data scraped from across the web. Does it include copyrighted material? When generated copy or images collide with someone else's work, who owns the liability? Several lawsuits have already been fought, and it still isn't fully settled.

And there's the question many people don't say out loud: what about my job?

The World Economic Forum's 2023 report estimated that 83 million jobs worldwide could be swallowed by automation over the next five years. Sounds terrifying.

But the same report has a second half: 69 million new jobs are expected to be created.

The ones swallowed and the ones newly created aren't the same people. What separates them is the willingness to relearn.

So how should you actually use it?

With this many problems, should you hold off for a while?

My judgment is exactly the opposite. Precisely because the problems are many, few people can use it well — and that's why the window is still open.

The key is one discipline.

Coca-Cola once ran a campaign called Create Real Magic, inviting consumers to use AI to generate artwork featuring brand elements. Sounds a little unhinged, right? But every piece that was finally published went through human review.

That's where the line sits.

There's a widely circulated saying in the industry called 70/30: let AI carry 70% of the heavy lifting, and humans hold the 30% that turns content from "usable" into "good".

I like that split, but my own summary is simpler:

AI handles scale; humans handle judgment.

AI can write twenty Instagram captions in a minute — a capability ten years of practice wouldn't give you. But which one fits your brand's tone, which one will offend people, which one is lying — only you know.

Take the subject line of an email. AI hands you two hundred candidates. But whether "our biggest sale ever" suits your brand, whether "exclusively curated for you" is what your customers want to hear — the machine doesn't know. You do.

That's human work. It can't be taken from you.

How to start, concretely? My advice is refreshingly simple: don't rip out your entire tech stack, and don't convene a half-year planning summit.

Pick the single most annoying repetitive task and hand it over first. Headline testing, or first drafts of product descriptions. Run it for a while, look at the data, then expand.

And there's one thing many people overlook: the data itself has to be clean.

Garbage in, garbage out. If your customer database has a missing address here and a mistyped email there, even the smartest AI won't help. What's interesting is that AI can work in reverse to clean your data — automatically spotting wrongly filled fields and filling in missing information, creating a positive loop. Better data, better campaigns; better campaigns, better data they leave behind.

Finally, let's talk money

After all that, back to the crudest, hardest question of all: money.

The return on generative AI in marketing can be measured along four dimensions.

Efficiency. Salesforce's 2024 State of Marketing report says 68% of marketers using generative AI report significant time savings, with many freeing up 10 to 20 hours a week.

Let's run the numbers. Take a midpoint of 15 hours a week, 50 working weeks a year — that's 750 hours. At 8 hours a day, nearly 94 working days.

For a ten-person marketing team, that's several extra people's worth of capacity out of thin air. That isn't a feeling — that's arithmetic.

Conversion. Adobe's 2024 Digital Trends report found that brands using AI for personalization see conversion rates up to 30% higher than those using traditional segmentation. Deloitte's research backs this up: 61% of consumers are more willing to buy from brands that offer personalized experiences.

Revenue. McKinsey's conclusion is even sharper: companies that do personalization well generate 40% more revenue from it than the industry average.

Retention. This is older data from Bain & Company: a 5-percentage-point increase in customer retention can lift profits by 25% to 95%. AI can predict which subscriber is about to churn and get ahead of them with a retention offer built for that one person. Keeping a customer is far cheaper than acquiring one — every marketer can do that math.

Four dimensions, four sets of numbers. None of it is wishful thinking.

Back to that dinner

Halfway through the meal, my friend glanced at his phone again and suddenly said: what scares me most isn't that AI works too slowly — it's that it works too fast, so fast I can't keep up with what it's writing.

I think he'd hit on something.

The faster the tool, the more valuable human oversight becomes. That was true of the printing press, and it's just as true of generative AI.

Marketers who know how to use AI aren't replacing their colleagues — they're replacing the days spent wrestling with a blank page.

As for you: may your team have both AI's speed and human judgment.