Prebid held a public webinar on May 14, 2026, dedicated entirely to the question of how the open-source header bidding organization is positioning itself inside the emerging world of agentic programmatic advertising. The session drew participants from both the sell side and the buy side - an intentional design choice that gave the discussion unusual breadth for a technical briefing.

The event covered the state of Prebid's Agentic Working Group, the protocol layer being built around the Ad Context Protocol (AdCP), early commercial signals from buyers already testing the system, and a frank reckoning with what the industry still lacks before agentic transactions can scale.

The protocol layer: AdCP and the seller agent

Most of the engineering effort underway at Prebid is happening at the protocol level, specifically around AdCP. According to John Rosendahl, Director of Product Management at Optable, and Olha Paramonova, VP of Ad Tech at Sigma Software - who jointly presented the Prebid Product Management Committee update - the current priority is defining how buyer and seller agents communicate, standardizing requests, responses, and negotiation logic, and creating a shared language for agent-to-agent interaction.

Prebid took formal ownership of the seller agent within the AdCP protocol earlier this year. The organization announced stewardship of the Prebid Sales Agent on January 29, 2026, after the repository had been incubated within AgenticAdvertising.org. The move gave publishers a neutral, open-source path into agentic infrastructure that does not require proprietary rebuilds or vendor lock-in.

The seller agent, as described in the webinar, serves three core functions. It represents publisher inventory in agentic environments. It translates publisher data - audience composition, contextual signals, pricing floors - into structured formats that machines can process. And it acts as the interface between supply and buyer agents during negotiation and transaction.

This is a meaningful departure from how programmatic has worked under OpenRTB. According to Rosendahl and Paramonova, standard RTB operates through fragmented, bid-by-bid transactions. Agentic protocols, by contrast, enable direct interaction between buyer and seller agents, enabling more holistic and outcome-driven negotiation. The shift is conceptually similar to the early days of header bidding, but with broader implications for how inventory is packaged and valued.

What publishers are missing

The webinar did not shy away from the current gaps. According to Rosendahl and Paramonova, publishers are not structurally ready for agentic buying. Their systems were built for human sales, not machine-to-machine communication. The biggest single gap is the translation problem: converting internal knowledge about audience, context, and inventory value into structured data that buyer agents can understand and act on.

Specific tools that do not yet exist include mature publisher-side tooling, a standardized method for packaging inventory for agents, and widely adopted workflows. These are not edge cases - they are foundational requirements for the seller agent to work at scale.

Interoperability is a central concern in the roadmap. The goal is to ensure that different agents from different vendors can work together without any single platform controlling the ecosystem. Asked whether Prebid intends to make its seller agent compatible with existing standards from the IAB Tech Lab, the speakers confirmed the aim is to stay interoperable with existing specifications while also developing new frameworks like AdCP that are better suited to agent-to-agent communication. The position is to work alongside current standards where possible, without being constrained by frameworks that were not designed for agentic interaction.

On business models, the expectation is a SaaS-style approach similar to existing Prebid ecosystems. Core infrastructure is expected to remain open, with monetization happening through services, tooling, and value-added layers on top.

Early commercial signals: spend in the low six figures

The webinar provided a rare quantitative data point. According to Rosendahl and Paramonova, initial agentic spend is currently in the low six-figure range. A gradual ramp is expected through the rest of the year. Multiple agencies and brands are already testing agentic setups, though not all have been publicly disclosed.

The scale expectation, however, is notably ambitious. According to responses provided during the session's Q&A, the trajectory is described as exponential rather than gradual - moving from low six figures toward potentially massive scale as infrastructure matures and more buyers participate.

The buy-side view: WPP and the transparency gap

Ingmar Zach, President of Product Data and Intelligence at WPP, presented the buyer perspective. His framing centered on efficiency and outcomes. Buyers are increasingly using automation to optimize campaigns toward cost-per-acquisition or return-on-ad-spend targets, navigate fragmented supply paths, and react to performance signals faster than human teams can manage.

The structural critique Zach raised is pointed: many current solutions - especially those from large platforms - operate as closed systems. They deliver results, but they limit visibility into how decisions are made. According to Zach, what buyers need is better access to structured, high-quality supply signals, more transparency in auction dynamics, and open frameworks that allow flexibility in how buying strategies are executed.

This is precisely the gap that Prebid's agentic framework is designed to address from the supply side. The tension Zach identified is not new. Industry experts have raised similar concerns about closed platform dynamics since AdCP launched in October 2025, with analysts questioning whether agentic protocols would primarily reinforce existing intermediaries or genuinely redistribute value toward publishers and buyers.

Scope3 on misalignment and feedback loops

Brian O'Kelley, founder and CEO of Scope3, presented the case study and marketplace insights section. His central argument: misalignment between buyers and sellers remains a major source of inefficiency in programmatic. According to O'Kelley, sellers optimize for yield without visibility into buyer goals, while buyers optimize for performance without understanding supply constraints. Missed opportunities accumulate because neither side has access to the other's signals.

Agentic systems offer a structural fix through continuous feedback loops. Seller-side logic can adapt floors or packaging based on demand behavior. Buyer-side systems can respond dynamically to changes in supply quality or pricing. But this only works, O'Kelley argued, if both sides interact through a shared framework - one reason Scope3 has been a driving force behind AdCP since the protocol's founding. Magnite built seller agent capabilities directly into its SpringServe platform and completed its first agentic test in December 2025 with Scope3 acting as the buyer agent, an early signal of how the protocol is taking shape in live infrastructure.

The RTB question: coexistence, not replacement

One of the more technically substantive questions in the Q&A addressed the relationship between agentic transactions and real-time bidding. The concern was direct: if buyer and seller agents negotiate directly and more continuously, does real-time bidding become obsolete?

According to the Prebid speakers, RTB is not going away. Agentic infrastructure will sit alongside it as an additional transaction model - analogous to how the waterfall still exists in parallel with header bidding today. Agentic transactions open up more direct, outcome-driven interactions, but auctions will continue to play a role.

This framing is consistent with how industry veterans have explained the distinction elsewhere. Ad industry veterans have noted that AdCP is not competing with real-time bidding but rather addressing a different set of transaction types, particularly programmatic direct and portfolio-level buying. The protocols exist in different parts of the stack.

The Google question and ecosystem openness

A pointed question during the webinar asked what happens when Google builds its own agent - specifically, whether a Google Ad Manager seller agent would simply absorb the agentic layer in the same way GAM dominates publisher ad serving today.

The response was notably pragmatic. According to Prebid, if major platforms like Google build agents, those agents simply become part of a broader ecosystem. The key architectural difference with agentic systems is that they are not as tightly constrained as today's programmatic pipes. Multiple agents can coexist rather than everything being locked into one system. Whether that holds in practice depends heavily on whether AdCP and similar open protocols achieve sufficient adoption before large platform agents establish de facto standards through market power.

Trust, fraud, and provenance

The technical risks in an agentic world are not trivial. One Q&A exchange addressed non-technical requirements: trust models, reputation systems, anti-fraud measures, and privacy frameworks. All of these have been persistent problems in programmatic advertising for years. In an agentic environment, where decisions are automated at speed and scale, those problems do not disappear - they accelerate.

According to the webinar, much of the governance work will come from AgenticAdvertising.org, with Prebid building against the outputs. Provenance was specifically highlighted as a signal expected to become more important. In a world where automated agents are making decisions, establishing reliable chains of custody around inventory - where it comes from, who authorized it, under what conditions - becomes a foundation for the trust that makes transactions possible.

The adagents.json specification, which Brian O'Kelley published at AgenticAdvertising.org on March 30, 2026, is one structural attempt to address authorization at the publisher level - extending the logic of ads.txt into a format that agent systems can parse with the granularity needed for placement-level, geographic, and temporal authorization.

Why this matters for ad buyers and sellers

For the marketing community, the significance of Prebid's agentic work is partly about neutrality. The organizations building agentic advertising infrastructure today include large platforms with substantial interests in controlling the transaction layer. Prebid, as a member-funded open-source organization formed in September 2017, represents a countervailing force: a shared infrastructure project with no proprietary platform interest in the outcome.

The practical implications for buyers extend to supply path transparency. Agentic systems that operate through open protocols allow buyers to understand how decisions are being made - and to build buying strategies on top of frameworks they can inspect and influence. The closed-system critique Zach raised at WPP is precisely what Prebid's seller agent is designed to address from the other side of the transaction.

For publishers, the stakes are structural. The risk in an agentic world is that, without representation in the protocols that govern how machines negotiate for inventory, publishers become passive price-takers in a system designed by and for buyers and large intermediaries. PubMatic launched AgenticOS on January 5, 2026, with WPP Media, Butler/Till, and MiQ as early participants. Optable's Audience Agent was embedded into AgenticOS in March 2026, enabling buyers to discover and activate publisher first-party data without that data leaving the publisher's environment.

Prebid's bet is that ownership of the seller agent, and stewardship of the open-source infrastructure around it, gives publishers a seat at the table in negotiations that will increasingly happen between machines.

Timeline

Summary

Who: Prebid.org, the open-source header bidding organization, hosted speakers including John Rosendahl (Optable), Olha Paramonova (Sigma Software), Ingmar Zach (WPP), and Brian O'Kelley (Scope3).

What: A public webinar examining Prebid's agentic advertising roadmap - specifically the development of a seller agent for publishers within the Ad Context Protocol, early commercial spend data, buy-side requirements around transparency, and the structural gaps preventing agentic transactions from scaling.

When: The webinar took place on May 14, 2026.

Where: The session was held online. The broader infrastructure work spans the global programmatic advertising ecosystem, involving publishers, agencies, supply-side platforms, and standards organizations.

Why: As agentic AI systems begin to influence how advertising inventory is bought and sold, publishers risk losing representation in protocols designed primarily by and for buyers and large platforms. Prebid's stewardship of an open-source seller agent is a direct attempt to ensure publishers have structured, machine-readable ways to communicate the value of their inventory - before that layer of the stack is defined without them.