‘Agentic with a small a’: CMOs are adopting AI more slowly than it’s evolving

“That’s the client preference at the moment,” said Wesley ter Haar, co-founder and chief AI officer at S4 Capital.

At the holdco, that presence shows up in workflow. Humans still anchor about 85% of the process. Teams initiate the brief, guide the models through execution and make the final calls. Even in areas where automation is more accepted — trend spotting, competitive monitoring and anomaly detection — the system flags, the human acts.

And the caution isn’t confined to large language models or flashy generative tools. It bubbles up just as clearly in decisions about agentic workflows, predictive bidding systems and advanced optimization engines.

“We want to be seen to be using it responsibly, and in doing so, thinking more about how it advises what we do, as opposed to executing on our behalf,” said Diageo’s head of culture, entertainment, media and digital Mike Cheetham at a Tesco event last month. “We will get to a point where we start to think about how AI could execute on our behalf.”

Weeks later, Cheetham’s comment was echoed on the latest earnings call from Procter & Gamble: “AI and gen AI capability help our teams discover consumer-relevant insights at every step of the consumer path to purchase, grounded in a unifying brand idea,” CEO Shailesh Jejurikar said.

Notice the boundary both companies draw. AI is central to insight generation and creative refinement. It is not yet positioned as the decision maker. It informs. It augments. It accelerates. It does not independently allocate budgets or set strategy.

“We’re using AI in some image generation for some creative campaigns, usually lower tier campaigns,” said Ivan Dashkov, Puma’s head of emerging marketing tech, who added that it represented a “very small percentage” of social assets.

Granted, some global advertisers, including Mondelez International, Unilever and Procter & Gamble, are pushing further. But the step from expanded automation to fully autonomous systems that manage advertising end to end remains significant.

“Something that we’ve been experimenting with is more on AI-based media optimization,” said Paras Shah, senior director of digital media at Georgia-Pacific. “Over the last, I’d say, year and a half, that’s an area where, you know, a lot of what we’ve been trying to do is figure out how we can empower our talent to leverage AI more in our digital media team. Because we can, it can help us get to a better outcome.”

The same reservations show up at the technically fluent advertisers with in-house ad teams. They are building agents internally but largely for contained tasks rather than end-to-end control, according to Robert Webster, founder of AI marketing consultancy TAU. The appetite for autonomy exists. The appetite for liability is another matter.

As Webster explained: “We’re already working with clients on those projects now, and often they’ve had to go through an internal process where they decide whether they want to build it themselves or use an alternative from their agency.”

These concerns aren’t new. What’s changed is their scale. A year ago they were background noise beneath the AI hype cycle. Now, they’re mainstream, surfacing just as the tech becomes more autonomous and capable of day-to-day decision-making.

Some of that anxiety is self-preservation. As AI absorbs more parts of the workflow, marketers are recalculating their own reliance. But the deeper tension isn’t AI’s capability, it’s governance. Businesses welcome systems that accelerate execution. They’re less at ease when those systems start reshaping accountability and power.

High level AI questions

None of this suggests CMOs are indifferent to AI. If anything, they’re leaning in — commissioning pilots, reallocating innovation budgets and peppering agency pitches with pointed questions.

As Scott Shamberg, CEO at independent agency Mile Marker, said: “In the pitches that we’re getting, there’s certainly more AI questions but we’re not getting ‘How is it used?’, ‘What does it do?’, ‘What’s the adoption rate?’. We’re getting more high level questions. ‘How do you think about it?’ How can it impact our business?’ versus anything that’s agentic specifically.”

Several structural factors are shaping that posture.

Many marketing organizations still operate with fragmented customer, media and transaction data, limiting the context available to AI. Once those partial inputs begin shaping spending decisions rather than just insights, financial exposure rises.

That exposure brings a transparency challenge. Systems that operate as black boxes are harder to defend when their outputs influence budgets, especially in markets where media supply, brand safety conditions and other variables shift quickly. The stakes rise just as explainability thins.

Then there’s the market itself. Opaque supply paths and principal trading already make it difficult to trace where value and decision-making sit. Layering automated systems onto that infrastructure makes oversight more complex, not less.

All of this produces a structural asymmetry. AI systems can influence outcomes at scale, but the mechanisms for oversight, explanation and liability still sit with individuals and teams. Control becomes partially automated, while responsibility does not. That misalignment supports automation in contained tasks, but makes organizations slower to delegate judgment and budget authority.

“I don’t think anyone’s let the agent take the wheel, so to speak, from an investment management [point of view],” said Sean Gilpin, CMO at Hyundai.

Nowhere is that tension more evident than in programmatic advertising. 

A pragmatic approach

Calls for “more AI” often miss a basic reality. Programmatic frictions are not simply prediction problems waiting for better models. They are structural: opaque supply paths, principal trading incentives, measurement distortions and fragmented identity frameworks.

The irony is that programmatic has never lacked automation. For more than a decade, machine learning systems have optimized bids, paced budgets and scored impressions in milliseconds. The market already runs on AI. What it struggles with is transparency, accountability and incentive alignment. 

“The smart advertisers aren’t rushing to deploy the technology,” said Gerry D’Angelo, former vp of global media at P&G and now a senior advisor at McKinsey. “They’re looking at their current business processes and where AI can genuinely improve them — and most of those opportunities are about efficiency. Think campaign taxonomies, ad tracking, standardizing naming conventions, resizing creative or automating repetitive trafficking tasks that junior teams still handle manually. That’s where AI can clearly add value — agentic with a small ‘a’.”

Large language models intensify the debate but they do not resolve those tensions. They are not built for low-latency bid decisions — and even if they were, smarter automation would not automatically close governance gaps. Layering another black box onto an already complex marketplace can make oversight harder, not easier.

The hesitation, then, is not technological conservatism. It is structural realism. Advertising’s constraints are rooted in computational limits and more in incentives. AI can optimize within that system. It does not, on its own, redesign it.

“No doubt, AI is going to revolutionize everything we do but so far it is difficult to scale, expensive to put in place, and fails to deliver measurable value in 95% of cases,” Publicis Group CEO Arthur Sadoun told analysts on the company’s earnings call earlier this month. “Consumer adoption of AI is better and faster than company adoption to realize the true potential of artificial intelligence.”