88% of Companies Use AI in Marketing. Fewer Than 10% Make Money From It.

A while ago, I had dinner with a few friends who run businesses.

AI was the undisputed star of the table. One said his marketing team hired two AI-savvy people this year. Another said they'd bought seven or eight tools and doubled the budget. Someone else flat-out said that next year, "AI penetration rate" would be written into the marketing department's KPIs.

As I listened, I asked one question:

Where's the money? Are you actually making any?

The table went quiet.

After a moment, someone said quietly: efficiency is definitely up. But making money... we've never really run the numbers.

That's what I want to talk about today. In AI marketing, where exactly does the money get made — and where does it get burned?

A Number That Should Make You Squirm

McKinsey's research says 88% of companies are already using AI somewhere in their marketing.

The same McKinsey has another set of numbers: fewer than 10% of them can see any real, dollar-and-cents impact on profit.

88%, and 10%.

Let that sink in.

The budget got approved, the tools got bought, the press release went out. Most companies are using AI to generate heat, not wealth.

So where's the problem? It's not that AI is useless. AI genuinely works in marketing, and in certain specific scenarios it works beautifully. The problem is that when most executives decide to invest in AI, they're going on anxiety and a vendor's slide deck — not on a cold, hard calculation.

Efficiency Went Up. The Business Didn't Get Faster.

First, let's pry two concepts apart.

What is productivity? What is ROI?

A lot of executives conflate the two.

IBM ran a survey in Q4 2025: 79% of companies said AI delivered efficiency gains, but only 29% of executives dared claim they could clearly account for their AI returns.

Those numbers sting.

Your team really is faster. A 1,500-word article that used to take 8 to 10 hours to write now takes under 2. Content output has multiplied several times over — email, ad creative, social media, everything is speeding up.

But are customers ordering faster? Has your acquisition cost dropped? Are existing customers sticking around longer?

If not, what you bought is "a simulation of efficiency," not a business upgrade. The team sprints inside the illusion of speed while the business stands still.

It gets harsher. MIT research from 2025 says 95% of generative AI pilots failed to deliver value. McKinsey's figure for the same year: only 6% of companies qualify as AI high performers — the bar being AI contributing more than 5% of profit.

95% fail, 6% excel. That's not a technology wreck rate. That's a decision wreck rate.

The Companies Making Money Win at One Small Thing

So what did that 6% do right?

McKinsey found a pattern: AI high performers are three times more likely to have senior leadership personally driving AI. Setting the direction themselves, looking at the data themselves, making the final call themselves.

Not tossing it to the tech team, not outsourcing it to an agency, and not hiring an "AI lead" and washing their hands of it.

Deloitte's 2026 analysis reaches the same conclusion: the strongest predictor of whether you can scale AI across the company is executive sponsorship and a formal governance mechanism. Not which tools you picked, not the size of the budget, not even the quality of your data at the start.

PwC puts it more bluntly: the companies that fail are the ones chasing the trend, without aiming AI at specific, measurable business processes.

The moment you outsource your AI strategy, the results stop belonging to you.

Where Is Real Money Made? Let Me Tell You a Few Stories.

Principles done. Now specifics. Where does AI actually make real money?

Paid media first.

This is where results show up fastest today. AI does three things: bidding, audience targeting, creative testing. The feedback loop is short, the data loop is closed, and whether you made money is plain as day on the dashboard.

Industry research puts it this way: AI-driven campaigns see 22% higher ROI, 32% more conversions, and 29% lower acquisition costs. They launch 75% faster and get 47% higher click-through rates.

Even more interesting is the market itself. AI-driven search advertising was roughly a $1 billion market in 2025, and is projected to hit $26 billion by 2029.

The market is moving without you.

Next, personalization.

Starbucks' Deep Brew runs personalized recommendations for 27.6 million loyalty members — every single person sees different offers. The result: average member spending rose 34%.

McKinsey's take: done well, personalization can cut acquisition costs by up to half and lift marketing ROI by 10 to 30 percentage points.

But there's a precondition here — a huge one: you have to use your own first-party data. Your member lists, your CRM, the behavioral data on your own platforms.

Feeding AI third-party data for personalization is like building a house on someone else's land. Not only is it unstable — it could be condemned at any moment.

Content production is another one.

93% of marketers say AI has accelerated content production. Teams working with AI see 41% more email revenue and far more consistent content calendars.

But I have to add a caveat. AI is a capacity amplifier, not a substitute for strategy. Cut human review in exchange for speed, and you'll get a pile of content that looks exactly like every other company's on the internet.

The cost of brand dilution never shows up in your content calendar report.

Predictive analytics is the underrated one.

92% of top marketing teams were already using AI for forecasting in 2025; 82% of CMOs say forecast accuracy has improved. Mature recommendation engines have delivered 150% conversion lifts and 50% higher average order values.

Yet it's also the least-deployed AI application in marketing. Thin competition. Rich pickings. The teams building this capability now are deploying with precision while their rivals are still arguing over quarterly attribution models.

Finally, customer service.

NIB, an Australian insurer, used AI to automate service workflows and saved $22 million. Across the board, AI self-service can cut service costs by 18% and shrink resolution times by up to 87%.

For B2B companies, this isn't a cost-cutting story. Let the bots catch the simple questions and free up your most expensive people — to close deals and nurture relationships.

This is talent reallocation.

Where Does It Go Wrong? A Few Stories Too.

Having covered the money-makers, we have to cover the wrecks. Most content about AI marketing won't touch this section, because the people writing it earn their living off "AI fixes everything."

Type one: strapping a turbocharger onto a bad strategy.

The strategy was already broken. You saw a competitor using AI, so you rushed to approve budget and buy tools.

McKinsey has studied digital transformation for decades: more than 70% of transformations fail, and the cause of death is unclear or misaligned goals — not technology.

AI won't fix a bad strategy. It will only make your bad strategy execute faster, and at greater scale.

Type two: buying a pile of disconnected tools.

Bain's 2025 research says AI projects stall because nobody owns them, there's no measurement discipline, and the processes were never redesigned. Companies with a fully integrated tool stack get twice the cost-efficiency gains of companies with loose point tools.

The self-check is simple. Have your team list the AI tools they use day to day: if there are more than six, and nobody can explain how data flows between them —

Congratulations, you're staging productivity theater. The tools are busy. The business isn't moving.

Type three: brand disasters.

Coca-Cola used AI for a holiday ad and got dragged by consumers straight into the trending topics. That AI-generated Willy Wonka experience, meanwhile, became a textbook case of a brand disaster.

Note: when AI content blows up, it's not AI's brand that takes the hit — it's yours. This is a board-level incident, not something marketing ops takes the fall for.

Any AI output that faces customers or touches brand voice needs a human-review gate. That's not bureaucracy. That's insurance.

Type four: over-personalization.

Past a certain point, personalization stops feeling like service and starts feeling like surveillance.

It's especially visible in B2B: blasting six contacts at the same customer with customized content at once earns you nothing but decision fatigue. CoSchedule's research says that in 2026, only 9% of marketers have actually made personalization a goal.

Where that line sits — for now, only human judgment can read it accurately.

Type five: pilot purgatory.

This is the most dangerous one, because it looks like progress.

BCG's 2026 CEO survey found: pilots blooming everywhere, but very few companies that can do the math. IBM's numbers: only 16% of AI projects have scaled company-wide, and only 25% delivered the expected return.

The companies that went all-in on marketing AI back in 2024 aren't running experiments in 2026 — they're harvesting the market. Every extra quarter you spend piloting hands a quarter of first-mover advantage to the rivals who moved first.

Four Gates

After seeing enough successes and wrecks, you can distill a simple test. Any AI marketing project, before the budget gets approved, goes through four gates:

Gate one, the Data Gate. Do you have clean data that you own to feed it?

Gate two, the Workflow Gate. Does it fit into existing workflows, or is it just another data silo?

Gate three, the Attribution Gate. Can you draw a line from the project to revenue?

Gate four, the Ownership Gate. Is there a named executive accountable for the results?

Fail any one of the four gates, and either the timing isn't right or the plan itself was designed wrong.

What Happens Next

Everything above is about "AI as a tool." The next wave is different: AI as an agent.

Gartner's call: by the end of 2026, 40% of enterprise applications will come with AI agents executing specific tasks — an 8x jump from 2025.

Applied to marketing, that means campaigns tune themselves in real time, leads score and route themselves, the content calendar schedules itself around demand signals, and the customer journey reshapes itself around individual behavior.

Companies that have already put more than half of their AI budget on agents see an 88% ROI realization rate, versus an average of 74% across all AI adopters. And that gap keeps widening.

But there's a flip side. Forrester warns that in 2026, companies will lose more than $10 billion to ungoverned generative AI. In PwC's survey, 60% of executives said responsible AI use improves returns — yet nearly half don't know how to implement it.

Governance isn't the brake. It's the ground that lets you floor the accelerator later.

Back to That Dinner

At the end of that dinner, my friend asked me: so tell me — what do I do now?

I said: don't buy any new tools yet. Run every AI tool you already have through the four gates. Anything that can't be connected to revenue — cut it or merge it. Then name one executive, just one, accountable for marketing AI returns, and review it once a quarter.

Then pick paid media for one pilot, and write down the hypothesis, success metrics, and review date in advance.

As for the rest — figure it out as you go.

The companies still ahead of their peers in 2027 won't be the ones that experimented the most. They'll be the ones that bet fastest on the right directions, built governance earliest, and had leaders who genuinely understood it.

Fewer tools, tighter discipline, deeper commitment. That's really all it is.

May you account clearly for every dollar you spend on AI.