Who actually switched to AI content, and did it work

The interesting question is no longer whether brands will use AI to make their content. It is how many already do, how fast that is moving, and whether the ones who switched are better off. All three now have published answers.

The adoption curve, from a survey that has run for years

The CMO Survey is run by Duke University's Fuqua School of Business with Deloitte and the American Marketing Association. Its Spring 2026 edition surveyed 308 marketing leaders. Because it has asked the same questions for years, the movement is the finding:

• AI now performs 24.2% of all marketing activities, up from 13.1% in 2024 — close to a doubling in two years.
• Generative AI specifically grew 220%, from 7.0% to 22.4% over the same period.
• Companies project AI will account for 55.9% of marketing activities within three years.

And the leading application is not analytics or targeting. It is making things: content creation, at 73.9%, ahead of personalization (65.4%), automation (48.9%), data analysis (46.3%) and targeting (45.2%).

Three switches with numbers attached

Klarna published its figures in a press release: a $6 million reduction in image production costs, external marketing supplier spend down 25%, roughly $10 million in annualised savings attributed to AI, more than 1,000 images in a quarter, and an image cycle cut from six weeks to seven days. It also reported cutting sales and marketing spend 11% in the quarter while running more campaigns, which is the part that makes this a growth story rather than a cost-cutting one.

Coca-Cola rebuilt its best-known Christmas campaign with generative AI in November 2025 — around 70,000 generated videos with about 100 people, fewer than the year before. The reception was mixed to hostile. It did it anyway, and it did it a second year running after taking the same criticism the first time. Read that as a company concluding the economics outweigh the reaction.

Meta is the platform rather than the brand, but it is where most of this content is spent. It reports $4.52 returned per dollar on AI-enabled Advantage+ products, 22% above business-as-usual campaigns, and an 11% higher click-through rate on campaigns using AI background image generation. Again: the platform's own figures for its own product.

The part most of these articles leave out

Adoption is not the same as competence, and the same survey says so plainly. No marketing technology activity scored above 5 on a 7-point performance scale, and those scores have not improved in two years. When asked what stops them getting value, marketers named budget (20.1%), technology integration and data architecture (19.1%), bandwidth and focus (14.1%) and talent (13.1%).

That is a precise description of a gap. The tools are everywhere. The ability to get a good result out of them is not, and it is not improving on its own. McKinsey's estimate that generative AI could raise marketing productivity by between 5 and 15 percent of total marketing spend, worth about $463 billion annually, is a ceiling, not a floor — and you only reach a ceiling if someone is steering.

What actually separates the brands getting a return

Across the published cases, the ones showing a real gain are not the ones that adopted earliest or generated most. They share something duller.

They changed a cycle, not a line item. Klarna's headline saving is money, but its operational change is six weeks to seven days. Saving 30% on the same eight campaigns a year is a procurement win. Being able to run forty is a different company.

They kept somebody accountable for the output. Coca-Cola's own framing was "human storytellers" directing the work, and the criticism it received was precisely about the places where that direction did not reach.

They did not announce it as the product. Customers do not buy a film because of how it was made. The method is an input. The only thing that travels is whether the work is any good.

So, should you switch

If you are a brand making content now, the evidence supports a fairly plain position.

Produce faster, and produce more. A seven-day cycle is now achievable, it is what your larger competitors have already moved to, and the volume it allows is the actual advantage — not the per-asset saving. Automate the business behind the content at the same time, because faster content pointed at a store that drops seven in ten carts just buys you a bigger leak.

And be honest about where the money should go. It should not go to the generating, which is nearly free and getting freer. It should go to whoever decides what is good, because the published evidence says that is the thing in short supply — and it is the only part of this that a competitor cannot buy off the shelf tomorrow.