In this three-part series, media industry veteran Clemenger BBDO chief media officer Stuart Bailey introduces a concept he believes is reshaping the industry: the Agentic AI Asymmetry Dilemma. Examining the impact on publishers, clients, and the ecosystem, Bailey argues that the winners in this new era won’t be the smartest, they’ll be the most connected.
I wanted to call out a structural problem forming in our industry, and it needs a name.
I am calling it the ‘Agentic AI Asymmetry Dilemma’.
This is an ecosystem problem, and I wanted to write a series that highlights the problem, as well as some practical solutions.
The first of these articles is focusing on publishers, and how they have been caught sleeping by the sudden and aggressive advent of agentic capabilities, being driven by the major agency groups.
Agencies are building proprietary agentic systems that autonomously plan, recommend, activate, and evaluate media, using data and logic that publishers cannot see. Publishers are trying to optimise for systems they don’t understand, using signals they can’t verify, measured against criteria nobody has shared with them.
The asymmetry is structural. It is compounding. And the market has no natural mechanism to correct it.
I want to be clear: this is not a criticism of agencies. Building proprietary AI capability is a legitimate commercial strategy, and the data that trains these systems was hard won. Agencies are under enormous pressure from clients demanding savings and efficiencies. They are racing to lead in a new world, and they need the rest of the market to catch up, not slow them down.
In fact, it is equally incumbent on publishers to invest, to build capability, and to show up as genuine partners rather than passive participants.
We Have Been Here Before
This dynamic should feel familiar, because we have lived through a version of it already.
A powerful platform is built. It accumulates data, reach, and algorithmic intelligence at a scale no individual publisher or agency can match. And then it becomes the system everyone depends on, operated according to rules only its creator fully understands.
We called them walled gardens. And one of their defining characteristics was the asymmetry they created. Advertisers pour money in. Data goes in. Detailed performance intelligence and the logic by which decisions are made do not come out in equal measure.
The industry spent over a decade debating this. I am not going to relitigate it here.
But there is a fundamental and important difference this time. The entities leading the charge with agentic AI are not just the platforms, they are intermediaries. Agencies have a view of the whole market: walled gardens, open web, publishers, and everything in between. They have the greatest potential to build truly connected agentic workflows that draw intelligence from across the entire ecosystem.
And crucially, they have a vested interest in doing exactly that. Because the more sources these models learn from, the better they become. The asymmetry is not in their long-term interest either.
The Winners Will Be Collaborators, Not Champions
There is a tempting framing of this moment in which the winners are simply the smartest. The companies with the most sophisticated AI.
I think this framing is wrong.
The value of an agentic system is not in its internal intelligence. It is in what it can access. The companies that forge the best interconnected AI workflows, across agencies, publishers, platforms, clients, and measurement partners, will accumulate a structural advantage that compounds with every connection they add.
The question is not how smart your AI is. The question is how well your AI connects to everyone else’s.
Publishers will need to invest. They will need to invest in infrastructure, expertise, and more than anything they will need to invest in understanding how agencies and clients operate, and how their structures and products can answer a modern marketer and agencies problems.
Current market conditions have created an intense drive into short termism. Publishers have not been immune to this, as they strip out product and so called “luxury” roles so that everyone has a number against them, and is driving short term sales, over long term business health.
This needs to change as every planning cycle that passes without alignment is another cycle in which agency AI systems learn without publisher input. The gap does not stay the same size. It grows.
In this Industry we have spent twenty years building moats. The next twenty years will reward bridges.
Trusted Repository or False Prophet
The scale and complexity of modern media have long exceeded what any individual can fully comprehend. It’s no surprise that many marketers feel overwhelmed. In this environment, agentic
AI is no longer a luxury, it is fast becoming the only viable way to hold that complexity together.
This does not diminish the need for deep expertise, strategic thinking, or human judgment, quite the opposite. These capabilities are more critical than ever. AI may serve as the repository of information, but it still relies on human input to shape it, train it, and ultimately validate its outputs. When this partnership works well, it creates the conditions in which human judgment can be applied to the questions that matter.
But there is another, equally plausible version of the future, one we should be honest about.
Agentic AI built behind proprietary walls, trained on incomplete or biased data, and deployed without transparency will not become more trustworthy simply because it operates faster or at greater scale. It risks becoming a false prophet: delivering confident answers grounded in flawed inputs, optimising for what it can measure rather than what truly matters, and concentrating power in the hands of its creators.
The difference between these two futures will not be determined by the technology itself, but by the choices we make today.
What This Requires
The Agentic AI Asymmetry Dilemma is structural. But structural problems have structural solutions.
The solution is not to assume the market will self-correct. It won’t.
The solution is to build the connections, the standards, and the collaborative practices that make the asymmetry unnecessary.
The winners in the new world of agentic AI will not be the smartest. They will be the ones who collaborated best, who understood that in an interconnected system, the strength of your
relationships is the measure of your intelligence.
The Agentic AI Asymmetry Dilemma is real. It is growing. But it does not have to stay.
If there is one thing I have learned about this industry and this country, it is that when we put our minds to something; when we decide as a collective to solve a problem, we can be an unstoppable force.
This feels like that kind of problem. And that kind of moment.
