Prebid enters the agentic arena. What does that mean for programmatic?
Everyone seems to be boarding the agentic AI train, so it was only a matter of time before Prebid climbed aboard.
The organization announced recently that it would take over the development of the Ad Context Protocol (AdCP) agentic sales agent code for publishers. It’s a significant move, given Prebid’s role as provider of open-source software that underpins much of programmatic advertising.
For a refresher: AdCP was developed by a group of tech companies to standardize agentic advertising, a model in which AI agents would communicate and transact with one another. But unlike most programmatic systems, AdCP is being built largely from the ground up, rather than on existing programmatic foundations.
It raises the question: What does Prebid’s move mean for programmatic?
Prebid Chairman Garrett McGrath claimed that this could pave the way for agents to conduct open-auction programmatic deals. Prebid’s treasurer, Phil Bohn, told AdExchanger that publishers “gain a clear path to agentic advertising capabilities without needing to build from scratch.”
If that’s the case, that means that publishers would have a friendly on ramp to agentic advertising. But some are dubious of those claims, and the programmatic ecosystem is already experimenting with agentic approaches in other ways that build on the existing infrastructure.
Some background helps. The rise of agentic has sparked (largely unhelpful) debate in the ad tech industry. IAB Tech Lab has released its Agentic RTB Framework (ARTF) which aims to standardize agentic workflow across the advertising supply chain, built on existing programmatic infrastructure and supporting existing AI protocols such as the Model Context Protocol (MCP) and Agent2Agent.
The framework uses containerization to enable faster communication and data exchange between technologies operating within the same data center. Here’s how Adam Heimlich, CEO of Chalice, an early backer of ARTF, described it at last week’s IAB Annual Leadership Meeting: “It’s a bento box that works at the edge with the advertisers’ choice of partners, performing the advertisers’ choice of tasks, with very low latency and absolute safety and control.”
Tech Lab CEO Anthony Katsur has been outspoken about what he sees as AdCP’s drawbacks. Most recently, he penned a piece for AdExchanger, published days after Prebid’s announcement, arguing that “ripping out battle-tested infrastructure and starting from scratch — as some emerging agentic protocols propose — is the slowest and most painful path you can possibly imagine.”
“Agentic systems cannot be trusted unless they can use the shared definitions, transparent interfaces and enforceable governance that enable trust and accountability,” Katsur wrote.
Momentum seems to be building behind ARTF. Most recently, PubMatic and Chalice announced that PubMatic would integrate a custom bidding agent — powered by its agentic operating system, AgenticOS, and developed within ARTF — with MCP. If that’s too many acronyms, the announcement described it like this: Human operators define objectives and guardrails, while AgenticOS continuously executes and optimizes across live programmatic transactions.
“Agentic systems cannot scale in isolation,” Heimlich said. “Interoperability between decision layers and execution infrastructure is what turns automation into accountable performance.”