Everybody has a walled garden essentially.
As artificial intelligence begins to make autonomous decisions across the machinery of modern advertising, the industry's central question has shifted from efficiency to ownership. At a pivotal summit in May 2026, agency leaders made clear that agentic AI is not a tool to be borrowed but a system to be possessed — driving a fundamental reorganization of who controls data, supply chains, and the invisible logic of where ads appear. The movement reflects a deeper human instinct: when the rules of a system change, those with foresight reach not for the output, but for the levers.
- Agentic AI is forcing agencies to confront a landscape so fragmented and opaque that one summit attendee described it simply as everyone having their own walled garden — and no one sharing the keys.
- The industry is abandoning its passive 'default on' posture, pivoting to deliberate 'default off' buying models that give agencies precise control over which publishers, exchanges, and supply routes are permitted to operate.
- A sharp internal divide has emerged over AI adoption itself — some agencies are racing to integrate autonomous systems while others refuse to let them near sensitive client data, exposing a governance gap the industry has yet to close.
- Rather than licensing AI capabilities from external platforms, agencies are investing in proprietary, in-house systems — accepting the resource strain in exchange for control they believe no vendor relationship can guarantee.
- Publicis' $2.2 billion acquisition of LiveRamp has sent a signal across the holding company landscape, accelerating a wave of M&A recalibration as rivals reassess how much of their critical infrastructure they've left in someone else's hands.
The advertising industry is undergoing a fundamental reorganization — not a story about robots replacing people, but about who gets to own the machinery those robots operate within.
At Digiday's Programmatic Marketing Summit in May, agency leaders described a programmatic landscape grown increasingly opaque: publishers with walled gardens, murky auction dynamics, and DSPs that offer visibility only on their own terms. The response has been a decisive shift from passive 'default on' buying — where ads run everywhere unless blocked — to a deliberate 'default off' model, where agencies choose exactly which publishers, exchanges, and supply routes are permitted to participate. It is a move toward curation, leverage, and control.
The deeper tension, however, surrounds AI itself. Agencies are experimenting with AI-assisted reporting, audience workflows, and campaign management, but remain sharply divided over governance. Some are fully committed; others refuse to allow autonomous systems near sensitive client data. The consensus emerging from the summit is that governance frameworks are not optional — they are the prerequisite for any meaningful adoption. And rather than outsourcing AI capabilities to external platforms, agencies are increasingly building proprietary systems in-house, accepting tighter resource constraints in exchange for ownership they cannot otherwise guarantee.
Publicis' $2.2 billion acquisition of LiveRamp crystallizes the broader strategic logic. The deal is nominally about identity resolution and data collaboration, but its real significance is infrastructural: by acquiring one of the industry's most critical data onboarding platforms, Publicis is purchasing control over a foundational piece of the advertising stack itself. Rival holding companies are already reassessing their own dependencies, and a wider wave of M&A recalibration appears to be underway.
What is taking shape is a market defined less by humans versus machines than by who controls the systems those machines inhabit. Agencies are positioning themselves as operators, not passengers — and the platforms, exchanges, and data providers that fail to adapt may find themselves excluded from the curated ecosystems agencies are quietly, deliberately building.
The advertising industry is in the midst of a fundamental reorganization, and it's not about replacing people with robots. It's about who gets to own the machinery.
At Digiday's Programmatic Marketing Summit in May, a clear picture emerged from conversations with agency leaders: the rise of agentic AI—systems that can make autonomous decisions across media buying workflows—is forcing agencies to rethink their entire operational structure. This isn't a software upgrade. It's a power grab. Agencies want control. They want it over their data, over their supply chains, over the systems that decide where ads run and how much they cost. And they're willing to restructure their entire organizations to get it.
The frustration is palpable. Attendees described a programmatic landscape that has become increasingly opaque and fragmented. Publishers have their walled gardens. DSPs have theirs. Auction dynamics are murky. One participant summed it up bluntly: "Everybody has a walled garden essentially." The result is that agencies are shifting from a "default on" model—where ads run everywhere unless explicitly blocked—to a "default off" approach, where buyers decide precisely which publishers, exchanges, and supply routes are allowed to participate. It's a move toward curation, toward deliberate choice, toward operational leverage.
But the real tension lies in how agencies handle AI itself. Some are diving in headfirst. Others won't touch it. The divide is sharp, particularly around data handling and privacy. Agencies are experimenting with AI-assisted reporting, audience workflows, and campaign management, but many are deeply wary about handing autonomous systems control over sensitive client information. One attendee framed it as a governance challenge: "We need to have governance frameworks for those." Another was more direct about the split: "You either have some folks who are all in on it… and you have some that will not touch it."
This tension is driving a strategic shift. Rather than outsourcing AI capabilities to external platforms, agencies are increasingly building proprietary systems in-house. Custom, private, controlled. One attendee captured the preference: "Building something in-house, custom and private, is really the best version of that." The constraint is real—resources are tight—but the direction is clear. Agencies want to own their own infrastructure.
Publicis' $2.2 billion acquisition of LiveRamp crystallizes this broader movement. On the surface, the deal is about identity, data collaboration, and AI readiness. But it's really about infrastructure ownership. By acquiring LiveRamp, one of the industry's largest data onboarding platforms, Publicis is buying control over a critical piece of the advertising machinery itself. Rival holding companies are already reassessing their own relationships with LiveRamp and rethinking their dependencies on external infrastructure. The deal has accelerated a wave of strategic recalibration across the industry.
The market dynamic that's emerging is less about humans versus machines and more about who controls the systems those machines operate within. Agencies are positioning themselves as the operators, not the passengers. They want tighter control over supply paths, clearer visibility into auction dynamics, and ownership of the AI systems that manage their workflows. The economics of advertising are being restructured in real time, and the agencies that move fastest to consolidate control over data, governance, and infrastructure will likely emerge with the most leverage. For the rest of the industry—platforms, exchanges, data providers—the question is whether they'll adapt to this new reality or find themselves locked out of the curated ecosystems agencies are building.
Citas Notables
Building something in-house, custom and private, is really the best version of that.— Agency attendee at Digiday's Programmatic Marketing Summit
You either have some folks who are all in on it… and you have some that will not touch it.— Agency leader describing the divide over AI adoption
La Conversación del Hearth Otra perspectiva de la historia
So when you say agencies want "control," what exactly are they controlling for? What's the fear?
The fear is opacity. They're sitting in the middle—between clients who want results and a supply chain that's become a maze of walled gardens and hidden auction dynamics. They can't see what's happening, and they can't explain it to their clients. Agentic AI makes that worse because now you're asking them to let an autonomous system make decisions in that same opaque environment.
But wouldn't an AI system be better at navigating that complexity?
In theory, yes. But agencies are realizing that if they outsource the decision-making to someone else's AI—a DSP's AI, a platform's AI—they lose visibility and accountability. If something goes wrong, they can't explain it to the client. So they're choosing to build their own systems, even if it's harder.
Is that realistic? Can mid-market agencies really build proprietary AI?
Not all of them. That's the real divide. The big holding companies like Publicis can acquire LiveRamp and build internal capabilities. Smaller agencies are stuck. But that's also why some are betting on independent solutions—vendors that can give them control without requiring them to build from scratch.
What does Publicis' LiveRamp deal actually change?
It signals that the game is about infrastructure ownership now, not just media buying efficiency. If you own the data layer, you own the relationship with the client. Competitors are watching and asking: do we need to own something similar? It's going to trigger a wave of M&A.
And the agencies that don't consolidate?
They become dependent on whoever does. That's the real risk. Control becomes a competitive advantage, and those without it lose leverage.