As with any digital ad campaign, the important thing is to reach streaming audiences who will convert. Roku’s self-service Ads Manager stands ready with powerful segmentation and targeting — plus creative upscaling tools that transform existing assets into CTV-ready video ads. Bonus: we’re gifting you $5K in ad credits when you spend your first $5K on Roku Ads Manager. Just sign up and use code GET5K. Terms apply.

Use code GET5K now

There is a particular sound the ad-tech ecosystem makes when it is confronted with something genuinely new. It’s not excitement, not alarm — more like the nervous throat-clearing of an industry that suspects it’s about to be asked to explain itself.

For that last few years, AI has been invited into the room like a decorative guest, expected to sit politely on top of the existing infrastructure without making any suggestions that might actually affect the floor plan.

Anne Coghlan walked into that same room and immediately asked why the floor was sloping.

She has this way of stating the obvious so plainly it becomes subversive. At one point she remarked, almost offhandedly, that “we’ve been watching AI be sprinkled on top of all of the existing ad-tech paradigms,” and you could practically hear the oxygen leaving the building.

Sprinkled — the word people use when something is cosmetic, insufficient, a gesture rather than a fix. With that one sentence she pushed aside a decade of industry claims that AI had been deeply, meaningfully integrated. What she was really saying was: you’ve been frosting the cake without checking whether it’s actually baked.

What she and the consortium behind AdCP attempted instead was something the ecosystem has avoided since RTB first took over the keyboard: they questioned the architecture itself. They didn’t talk about modernizing programmatic or reimagining bidding or any of the familiar buzzword acrobatics. They asked a more fundamental question — what if the system isn’t broken in the way people keep insisting? What if it’s broken in the way they’re too exhausted to articulate?

That exhaustion shows up in the numbers. When Anne described how her team discovered “24,000 resellers for one site,” she didn’t deliver it like a revelation. It came out as a kind of field note from a naturalist who has wandered too deep into an ecosystem with its own internal logic. She even laughed a bit — the stunned, what-else-can-you-do kind of laugh — as she admitted they assumed it was a bug. It wasn’t. It was the supply chain behaving exactly as it had been incentivized to behave.

And that is how AdCP emerged: not from ideology, but from structural clarity. As Anne put it, gently but unmistakably, the goal wasn’t to create yet another improvement to the existing ad-tech machinery. It was to “solve all problems in a new way,” which is the kind of phrase that, coming from anyone else in this space, would sound like bravado. With her, it came across more like a diagnosis — not a promise, but a blueprint.

One of the things that makes talking to Anne refreshing is that she does not have the Silicon Valley habit of describing AI like it’s auditioning for sainthood. She talks about agents the way a seasoned editor talks about a summer intern: useful, promising, and very clearly not allowed near the expensive equipment unsupervised.

“I think about agents right now as our smart interns,” she explained, and the metaphor fits so well you wonder why the industry spent so long pretending agents were anything else. Interns need direction. They need supervision. They are capable of surprising competence and equally surprising confusion. And, crucially, they do real work — the kind humans dislike doing but cannot avoid.

Her emphasis on guardrails is constant. It’s baked into her speech patterns. When I joked about agents eventually changing my Wi-Fi password — because in ad tech, imagining the disaster modes is half the entertainment — she shut it down instantly, noting that “your Wi-Fi password is serving you ads,” which might be the most devastating one-liner any ad-tech executive has delivered this year. But she followed it with something more important: “built into the protocol is this idea of a human in the loop.” Not because humans are better, but because they are accountable in ways machines cannot yet be.

The brilliance of AdCP is that it treats AI not as prophecy but as labor. It acknowledges what AI can do and — more importantly for the health of the ecosystem — what AI should not be expected to do. Agents can negotiate, reason, interpret, confirm. Humans can define, decide, evaluate, feel. The system works only if both parts show up as they are, not as the marketing departments of the world imagine them to be.

In other words, it’s the first protocol in ad tech that refuses to cosplay.

There is a moment in every industry where its ambitions outgrow its infrastructure. For ad tech, that moment arrived quietly — sometime between the rise of immersive environments and the realization that the existing pipes were built for rectangles. Anne articulated it cleanly when she said, almost rhetorically, “could you buy a drone show experience… in the programmatic world? You can’t.” Not because no one has tried, but because the system was never designed to carry that kind of creative intent.

The industry keeps trying, of course. It stretches taxonomies that were already overextended. It maps increasingly exotic experiences onto fields meant for banner ads. It adds more intermediaries, more compliance wrappers, more layers of abstraction — all in the hope that the old pipes will somehow accommodate the new water.

Anne’s diagnosis, once again, is simple: they won’t.

AdCP assumes a world where advertisements behave less like units and more like performances — contextual, dynamic, collaborative, intricate. It allows the buyer to express nuance without sending a PDF, and allows the seller to describe their environments in sentences instead of taxonomies. It creates a space where “creative and media working together really, really well” isn’t a fantasy; it’s the design.

This was the piece even the IAB Tech Lab seemed reluctant to confront as it launched its parallel agentic framework, ARTF. The Lab’s instinct — one part institution, one part reflex — was to build a system sturdy enough to maintain central gravity. AdCP’s instinct was to decentralize the reasoning layer, to let agents talk to each other without requiring a central authority to translate every syllable.

This philosophical split is the heart of the debate:
whether the future of ad-tech AI belongs inside the walls of the Tech Lab or outside them.

And it’s telling that the Lab’s response wasn’t to ignore AdCP, but to build something adjacent to it. That is how incumbents acknowledge challengers without admitting that they’re acknowledging challengers.

One of the quiet pleasures of talking to Anne is watching her describe the dysfunctions of the ecosystem with the precision of someone who has observed every variety of absurdity. She doesn’t roll her eyes — she doesn’t need to. The ecosystem rolls them for her.

When I asked whether programmatic’s original sin was “data hoarding, fake metrics, or pretending to care about ad comps,” she didn’t flinch. She simply said the ecosystem “forgot who it was meant to serve.” And in that moment, the entire supply chain shrank to a single point of failure: no one built it for the outcomes they claimed to optimize. They built it for the intermediaries who learned to thrive inside the noise.

Her view of the ecosystem’s excesses is equally blunt. “AI slop is the evolution of made-for-advertising,” she said, and honestly: she’s right. The mess has metastasized. And if half the ecosystem goes extinct because it added no value? “Happy days,” she said, without breaking cadence.

This is what happens when a system ages without accountability: the middle expands, the ends suffer, and everyone pretends the math still works.

AdCP is her answer to that imbalance.
Not a correction.
A replacement.

Speaking with Anne about how agents actually communicate, you begin to see why AdCP feels so different from the last decade of “innovation.” She doesn’t describe a process; she describes a conversation.

“My agent goes to you and says, ‘I want to buy this from you — what do you have?’” she explained, as if she were describing two coworkers comparing notes in a hallway. The seller’s agent responds, clarifies, describes context, outlines constraints, and confirms alignment. The entire exchange behaves the way humans behave when the system surrounding them isn’t actively fighting them.

“It’s almost like two humans conversing,” she said.
“Only difference is they are machines… and they have been given the rules of engagement by us.”

That line contains the whole philosophy:
machines talk; humans govern; the mess evaporates.

And in a way, it also explains the Tech Lab’s anxiety — not the loud kind, but the silent, procedural kind. ARTF tries to bring agentic workflows into the Lab’s universe, to containerize them, to ensure the infrastructure remains under central stewardship. AdCP releases that logic into the wild. Two models, two futures, two implicit claims about where authority should live.

The story is not which one will win.

The story is that the debate is finally honest.

And Anne — quietly, decisively, almost annoyingly calmly — is the reason for that.

You’ve read the free stuff.

You’ve laughed.

You’ve nodded along like a strategist pretending the deck was their idea. But here’s the thing: ADOTAT+ isn’t for the dabblers. It’s for the people who want to know why brands make you feel things before you even realize it’s working. It’s where cognitive science meets sarcasm, and where “brand purpose” gets stripped down to the PR perfume it really is.

Ulli Appelbaum drops frameworks that actually have footnotes. Jeff Greenfield talks like Carl Jung if Jung had worked in adtech. It’s brand science, not brand fluff — and the only therapy session where the data talks back.

So if you’re tired of PowerPoint prophets and marketing mythology, join ADOTAT+.
Because insight this sharp shouldn’t be free — and neither should your brain.