Horizon Media Builds the Plumbing for Agentic Ad Buying

Horizon Media just threw its agentic hat into the increasingly crowded ring by adding agentic ad buying into its intelligence platform, Horizon OS Blu. The agency has built a software layer into its platform, so its ad-buying agents can interact with agents from adtech partners (including Innovid, Magnite, Vidmob, Smartly, and others) as well as agents built by media companies (such as Fox, Disney, TikTok, and NBCUniversal). Clients like SharkNinja are currently using Horizon’s agentic layer to connect with retailers, according to the company. Those agents are designed to combine a variety of data—including publisher insights into audiences, campaign performance data, and measurement—to make media buying faster and smarter. Traditionally, an ad buyer like Horizon would build an audience they want to target with ads. But staff from media companies like Fox or NBCU must then work with the buyer to determine how to match that audience to their environments. But AI agents, armed with data from both the ad buyer and ad sellers, can resolve that question quickly. “Eighty percent of our audience-building queries used to take five to 10 days to create,” said Katie Comerford, president of Horizon commerce and client transformation. She said that it can now happen in hours or minutes. “Speed to insight to action is the point of all of this,” she added. In Horizon’s system, the actual purchase doesn’t happen autonomously; human staffers must make a decision from a menu of options, from making the buy to asking for more information. Horizon's AI agents are currently internal, said chief product and data officer Domenic Venuto, but Horizon is looking at the possibility of deploying them to clients. "We're talking to some clients right now who want to activate certain workflows within their working models—so they could be calling the agent themselves,” he said. Meanwhile, AI agents built by adtech and media partners can currently plug into Horizon’s agent orchestration layer. Initial agentic work will probably be modest: “The first ‘handshake’ is about establishing trust, and then grows to larger enterprise-to-enterprise workflows,” said Ryan McConville, NBCU’s evp and chief product officer of ad products and solutions, in an email. Do agentic workflows untether clients from their partners, or tie them tighter? Horizon emphasizes portability to differentiate its nascent agentic system from those developed by holdcos. "Our goal is to be an open ecosystem, plug into our clients’ tech stack, and actually try and help them reduce their total cost of ownership,” Comerford told ADWEEK. “Maybe they don't need certain data sets. Maybe they don't need some of the other capabilities, and we can be more efficient at it as well." Horizon’s press release on the news swings at big agency competitors. According to Samantha Rose, evp and head of integrated investment and programmatic, the ability for partners to build agents on top of Horizon’s platform “stands in contrast to how holding companies typically operate: consolidating technology ownership and forcing clients into proprietary stacks regardless of fit.” But the data that powers the technology in an agentic world is arguably more important than the technology itself. Agentic workflows between companies will have to become more prevalent before we understand whether they bind clients closer to their partners or allow more freedom of movement between them. The portability of the data that powers the agents’ decision-making could ultimately be what prevents vendor or agency lock-in. AI agents are poised to transform the ad landscape It feels like everybody in the advertising world—from brands and content owners to vendors and agencies—is splurging on building out agentic capabilities. Fox, for instance, recently told ADWEEK that it has built an “end-to-end” agentic ad platform. NBCU also touted its agentic capabilities during the upfronts. Dentsu invested heavily to bulk up its platform in anticipation of agentic collaborations. Expect a deluge of agentic news from ad industry companies over the next week. “This is the year agentic AI moves from technical proof of concept to full production,” said McConville. The industry ramp-up has been so sudden that, in recent months, it’s gotten to the point where agentic collaboration is now a requirement of the RFP process, said Comerford."We are seeing more and more RFPs come through asking about agentic capabilities, and our ability to co-collaborate with clients’ agents,” said Comerford. If agentic solutions take hold, it could cut out a lot of the intermediaries whose job it is to shuffle data and inventory from ad buyer to ad seller. These include staffers involved in the human handoff between agency buying teams and ad sales teams. And it includes adtech vendors that also facilitate data-driven ad buying. "I think that there are some big existential questions that this raises: how we access suppliers, how we partner with publishers, the role of SSPs and DSPs,” said Venuto. “All of that is up in play." AI agents aren’t the end goal; the system that supports them is Agentic transformation is still too nascent to have transformed anything yet, and it’s to be seen whether many of the AI agents being released across the ad industry will perform both consistently and at scale. NBCU’s McConville pointed out that agent interactions need to be codified around trusted industry standards. Those standards haven’t been developed yet. And, he added, the industry needs to agree on how to measure success. But that is not stopping the investment. Companies are approaching agentic transformation with a much different philosophy compared to traditional enterprise software, where platforms and point solutions require long installation processes and can fail expensively. The monetary and time cost of building and deploying an AI agent to automate a task is, by contrast, minuscule. AI agents should not be thought of as point solutions, Comerford said, but as part of a larger system. In this system, companies can build AI agents on an as-needed basis to solve specific problems as they arise. In that sense, firms like Horizon will have to build a lot of agents, many of which won’t work, to figure out what does. “If we aren't failing in what we're testing and trying with agents, we're probably not testing and trying enough,” said Comerford. Added McConville: “What we’re doing today is intentionally a starting point, so we can understand where autonomy adds value and how it can evolve responsibly over time.”Correction June 18: Katie Comerford's title has been updated.

Artificial Intelligence

Horizon Media Builds the Plumbing for Agentic Ad Buying

The agency enhanced its intelligence platform to support AI agents and speed up the ad buying process

UNLOCK FULL ACCESS

Ryan Joe

Agencies

When It Pays for an Agency to Break the Rules

WPP Promotes Adam Gerhart to Global Chief Client Officer

Stagwell Wins ServiceNow in the Largest Multi-Agency Deal in Its History

Matt MacDonald Leaves VCCP for Omnicom's Merkley+Partners

With Fewer Big Holdcos, Arthur Sadoun Says Client Conflict Is No Longer a Dealbreaker

Havas Grows In North America as It Lags In Other Regions

Publicis Thanks Omnicom For Market Consolidation As Q3 Growth Hits 5%

Award-Winning Agencies Talk TikTok

ADWEEK Agencies Advantage: Brands Stall the Shift to Performance Pay