Dwi Wanna
By Seb Joseph | May 22, 2026
The promise of agentic AI has been the singular drumbeat of the advertising industry for the past 12 months. Across the global marketing landscape, holding companies have unveiled proprietary AI operating systems, ad tech vendors have rushed to market with autonomous campaign-creation tools, and consultancies have published elaborate frameworks. CMOs have taken to global stages to declare that we have entered the age of autonomous marketing.
Yet, beneath the glossy surface of press releases and keynote speeches lies a starker reality. According to Dr. Daniel Hulme, WPP’s chief AI officer, the industry is currently mired in what he candidly describes as the "teenage sex phase" of artificial intelligence—a period defined by the illusion of widespread adoption where "everyone thinks everyone else is doing it, but when they actually look, they’re not."
The Illusion of Deployment: A Chronology of Expectation vs. Reality
The industry’s current infatuation with agentic AI follows a familiar, if somewhat disappointing, historical trajectory. In the annals of advertising technology, this gap between consensus and operational reality is a recurring theme.
- The Programmatic Era (2010s): Programmatic media buying was promised as the total revolution of the industry. While it eventually transformed the ecosystem, the reality lagged a decade behind the initial surge of industry marketing.
- The Data Clean Room Phase (Early 2020s): As third-party cookies began to crumble, clean rooms were pitched as the universal solution for addressability. Years later, they remain a nascent, often fragmented solution, still struggling to unlock the full potential of the open web.
- The CTV Measurement Gap (Mid-2020s): The promise of digital-grade accountability for television remains a persistent point of friction, with many marketers still unable to effectively reconcile cross-platform data.
Agentic AI—systems capable of autonomous decision-making, task execution, and iterative learning—is merely the latest entry in this chronicle of "teenage sex" moments. When agency traders, media planners, and campaign managers are pressed on their actual adoption levels, the responses are almost universally uniform: “We’re exploring it,” “We’re currently in a pilot phase,” or “We’re still building the business case.”
The "Shit Show" of Scaling Autonomous Agents
What differentiates agentic AI from previous technological shifts is the unique texture of the implementation gap. It is not necessarily that companies are being dishonest, but rather that the industry is collectively avoiding the uncomfortable truth of what deployment at scale actually requires.
Dr. Hulme’s warning to attendees at the IAB U.K.’s AI growth summit was blunt: "The reality is that companies will deploy an army of agents across the organization, and forgive the technical term, but it’s going to be a ‘shit show,’ because most of those agents are not going to be capable of doing their jobs."
The primary bottleneck, according to Hulme, is not the initial programming of the agent, but the testing. "At least 80% of the energy that you need to build agents is testing," he noted. In a complex, high-velocity environment like digital marketing, testing is not a one-time setup phase; it is an ongoing, labor-intensive commitment that most organizations are currently under-resourced to handle.
The Second-Order Gap: Why Historical Data Fails
One of the most critical challenges for marketers moving toward agentic systems is the "second-order gap." This concept represents a fundamental paradox in AI-driven advertising: the moment a model is trained on historical campaign data and deployed to make autonomous decisions, it begins to alter the very ecosystem it is designed to operate within.
If an AI agent is trained to predict consumer behavior based on past media performance, its intervention creates a new reality. Consumers react to the AI’s new strategies; competitors pivot their own tactics in response; media prices fluctuate. The environment is no longer the one the model was trained on.
"The problem is that buying and selling didn’t exist in the past," Hulme explained. "I’ve now changed the behavior, and marketing is the same thing. If I build a ‘magic oracle’ and predict human behavior, I haven’t changed that behavior. Now I predict, and the models are out. You cannot predict the future based on the past."
Without rigorous, real-time testing, a model risks becoming obsolete almost immediately upon deployment. Marketers are not just testing whether the agent follows its initial instructions; they are testing whether those instructions remain relevant in a market that has already moved in response to the agent’s own influence.
Moving Beyond "Rule Following"
A critical takeaway from the current state of the industry is that most of what is being marketed as "AI" is not actually agentic. It is, in fact, a sophisticated form of automation—or, as Hulme puts it, "very fast, very sophisticated rule following."
True agentic AI requires a system that makes decisions, observes the outcomes, and adapts its own logic—a loop that most current ad tech stacks are not built to facilitate. The industry is currently utilizing the "weakest possible version" of AI: systems that simply replicate what humans can already do, albeit faster.
Strategic Implications for Marketers
For brands and agencies looking to navigate this landscape, the path forward requires a shift in focus:
- Stop Chasing Quick Wins: Hulme argues that low-hanging fruit and simple automated tasks can be outsourced to third-party vendors at a fraction of the cost. Investing heavily in building internal tools for basic automation is a misallocation of resources.
- Focus on Differentiation: Instead of attempting to automate commodity tasks, companies should focus their AI efforts on the problems that genuinely differentiate their business. If an AI project does not solve a unique business challenge, it is likely a distraction.
- Invest in Infrastructure for Testing: Organizations must prioritize the development of robust, automated testing environments. Without a framework that can handle the constant flux of market responses, the deployment of agentic AI will likely result in operational chaos rather than efficiency gains.
- Accept the "Feedback Loop" Reality: Marketers must shift their mindset from "prediction" to "observation." Because agentic AI changes the environment it operates in, success is measured by the system’s ability to adapt to its own impact, not by the accuracy of its historical modeling.
Conclusion: The Road to Maturity
The advertising industry is currently in a state of transition. While the "teenage sex phase" of agentic AI has provided plenty of headlines and inflated expectations, it has also highlighted the significant work that remains to be done.
The promise of autonomous, agentic marketing is real, but it will not be delivered by simply plugging in a new software layer. It requires a fundamental shift in how organizations think about data, testing, and the nature of decision-making. As the hype inevitably cools, the companies that will win are those that stop pretending the technology is already mature and start doing the hard, unglamorous work of building systems that can truly learn, adapt, and operate in an unpredictable, ever-changing market.
For now, the advice for CMOs is simple: look past the marketing, ignore the buzzwords, and focus on the testing. The "shit show" is coming for those who don’t, but for those who build for reality rather than expectation, the potential for genuine transformation remains vast.
Muslim
