Jochen Schlosser, chief technology and product officer at Adform, told ExchangeWire's Trader Talk TV in an episode published on September 23, 2026 that the two agentic advertising protocols competing for the industry's attention do largely the same job, that very little money passes through either of them, and that the advertisers they are designed for would ideally never notice which one their platform has built. The 26-minute whiteboard session set the IAB Tech Lab's Agentic Advertising Management Protocols (AAMP) against the Ad Context Protocol (AdCP) and asked a question that has hung over the sector since October 2025: why does ad tech need two?

In Short

The head of technology at Adform, a European company that makes software for buying and selling online ads, went on a video show and explained how two competing sets of rules for AI ad-buying software compare. Both rulebooks try to let AI agents find ad space, agree prices and set up campaigns, but hardly any ad money flows through either one yet, and every ad platform that supports them has to pay engineers to do so. His view is that you would not have to pick a side, because your ad platform would translate between the two, which leaves the platforms deciding how much of that cost ends up in their fees.

A whiteboard, Britpop and a transcript with gaps

Ciarán O'Kane, co-founder and adviser at ExchangeWire and a general partner at FirstPartyCapital, hosted the episode, which ExchangeWireTV uploaded to YouTube. O'Kane promoted it on LinkedIn, casting the contest as the ad tech version of Blur against Oasis. The video runs in six chapters, the last of which, on adoption, starts at 21:50.

Schlosser said this was his third appearance on the programme. On an earlier one, in May 2025, he discussed Microsoft's decision to close its Xandr DSP. He is not a disinterested observer. Adform sells a demand-side platform, an ad server and a supply-side platform from one system, and it booked EUR 103.4 million in 2025 revenue, 81.9 per cent of it from buy-side fees. It sits at both ends of the transactions these protocols would standardise, and pays the engineering bill for supporting them.

The job both protocols claim

Both protocols remain little used, according to Schlosser, though he placed them beyond incubation. Their purposes overlap almost entirely: simpler campaign setup, simpler negotiation and help in discovering supply, whether packaged as an insertion order, a programmatic deal or a programmatic guaranteed buy.

He divided the benefit by client sophistication: efficiency for mature advertisers, access for less mature ones. A mid-sized advertiser wanting connected television, in his example, would not know every publisher to call - its agency would. The protocols put publishers and SSPs behind agentic interfaces, letting a buyer query for inventory it would not otherwise find, since much premium supply never reaches the open market. Otherwise, he argued, buyers use OpenRTB to bid rather than to activate negotiated packages.

What buyers want is help rather than hands-off automation. People "don't crave for autonomy they crave for assistance," he said. The demand he described is for a native agent inside the platform, or one reached through a buyer's chosen large language model, that surfaces inventory without a complicated interface. According to ExchangeWire's episode description, clients want conversational agents that find inventory and bypass complex setup menus.

Same functions, different foundations

Schlosser colour-coded the functions of each protocol in green on the whiteboard and found little to separate them: discovery, creating inventory opportunities, packaging, negotiation and targeting appear on both sides. What differs is the foundation.

AAMP, in his description, is "meant as an evolution of the existing protocol ecosystem". It builds closely on OpenRTB, while also allowing activation through OpenDirect, the IAB Tech Lab specification for direct programmatic buying. Everything already attached to OpenRTB comes with it: curated audiences, classic frequency capping and the targeting already built into DSP and SSP stacks. On top, IAB Tech Lab is adding a layer for package discovery, negotiation and setup. Continuity has a price. A developer connecting to AAMP needs to know what an entity is in OpenRTB, how deals work and how identifiers are handled.

AdCP takes a more abstract, execution-agnostic stance. It describes what an execution layer must do rather than prescribing which one to use: if a campaign needs targeting, whatever sits underneath must return a proper targeting interface. OpenRTB can fulfil an AdCP request, but so can something else. Schlosser pointed to AdCP's Trusted Matchcomponent, which can perform frequency capping away from the DSP, using a different identity system, on individual SSPs or directly on the publisher side. That brings freedom, and with it more implementation work in the execution layer. The buyer-facing definition layer is lighter; how the rest gets built is left to sellers.

Audience data shows the consequence. In a DSP built on AAMP, audience vendors sit in the platform's own marketplace. Under AdCP, a buyer states its audience requirement and sends it to publishers or SSPs, which must return traffic that fulfils it - so the same audience vendor needs an integration with each of them. "You're pushing functionality to another part in the ecosystem," Schlosser said.

Some trading optimisations, such as cost-per-acquisition buying, are not yet possible in AdCP, he added, though he predicted they would arrive. When the two protocols started, the overlap between them was small. It is now, he said, getting bigger.

Where the standards body draws the lines

Schlosser's framing - AAMP as the heavier, OpenRTB-rooted option and AdCP as the lighter abstraction - resembles the layered picture that IAB Tech Lab's chief operating officer, Shailley Singh, set out to dismantle in a document PPC Land reported on August 20. Singh listed 13 business functions both frameworks address and said AAMP does not require every transaction to clear through OpenRTB, citing OpenDirect-based direct deals, the Deals API, private marketplaces and linear television buying. The same document characterised Trusted Match as a single targeting plane.

The accounts partly converge, since Schlosser also named OpenDirect as an AAMP route. They diverge on his claim, set out below, that channels outside OpenRTB fall beyond AAMP's reach. On the larger point, that the overlap is wide and widening, both agree.

Two protocols, one patient

O'Kane pressed the obvious question: why two, when one would seem sensible? Schlosser answered with a medical metaphor. After 20 years, he said, parts of the stack "have become a bit fat", and a diet might not harm. "At the same time, the question is, do you need to kill the patient in order to create a new skeleton and then flesh it out again?" AdCP, in his telling, starts over while potentially remaining compatible; AAMP holds that "the patient is actually pretty healthy" and extends its limbs toward agentic capabilities.

AAMP already reaches digital out-of-home, connected television, display and native, and is moving step by step into broadcast, according to Schlosser. But it requires OpenRTB, which several large players do not use. AdCP's pitch is broader. A buyer wanting one platform to run search, social and everything else - Schlosser's placeholder for an unbuilt channel was "ads on Mars" - could in principle do so if each channel built a system compatible with AdCP's descriptions. The obstacles are familiar: different measurement criteria, different transparency requirements and separate systems for search and social. "This has the potential to break the silos," he said, while conceding that whether anyone believes the vision is a different question.

O'Kane brought the discussion back to adoption. "There's very little money running through both these systems, right?" he asked. Schlosser's reply was a single word: "Absolutely."

A core layer with ports

How would platforms cope with two? Schlosser expects many DSPs and SSPs to build what he called a core layer: a component that takes a brief - KPIs and goals, budget, dates, inventory and audience descriptions - and passes it through connectors to whichever protocol a seller speaks. His illustration was hypothetical, and he flagged it as guesswork: Disney might one day use AdCP for insertion orders and guarantees, Paramount might choose AAMP, and the same brief would reach each over its protocol. No such choice by either company was described. It costs extra effort, he said, but he did not expect any platform to rule out AdCP entirely.

50 milliseconds against 20 to 100

Would another layer add latency? Schlosser separated two phases. Setup and negotiation can take a day without anyone minding: most campaigns are not built in two hours, and much negotiation still needs human sign-off, because no publisher runs its premium deals strictly off a rate card. Agents, he said, "can only do rate card fast". Everything else is routed to approval anyway, so whether setup takes two hours or four hinges on people rather than protocols.

Execution is different. Once a campaign is live, systems must pace budgets, avoid overspend, reach the audience and report, fast and at volume. AdCP's original premise was to avoid that overhead, but added functionality has brought added technology. Trusted Match, according to Schlosser, now carries a service-level agreement of 50 milliseconds. OpenRTB's window he put at 20 to 100 milliseconds, while Adform's internal systems must reply within 2 milliseconds. "This is not like we're talking a different planet," he said. Real-time frequency capping, a later addition, pulls AdCP into the same territory. "So the same requirements are hitting different protocols and the downstream execution layer ends up with similar solutions," he said. The latency figures are his, offered on air, not drawn from a published specification.

Who pays for discovery

Asked who pays, Schlosser said it depends on the task. He struggled to see a fee for discovering a pre-negotiated catalogue that already sits on Magnite and is already visible in Adform's interface through a deal synchronisation API. "Discovering a catalog, I can't imagine anyone charges for that," Schlosser said. Finding deals, packaged or not, amounts in his view to a new interface and a more efficient workflow.

Data is another matter. URL targeting is almost free, according to Schlosser. Standard contextual targeting costs about six cents and custom contextual targeting about 20 cents - figures he gave without stating the unit. "So the more custom you go, the more you pay."

Google, O'Kelley and the question of motive

O'Kane then raised motive: why build AdCP when IAB Tech Lab already sets standards for the ecosystem? He described AdCP as largely Brian O'Kelley's vehicle - the captions render the name as "Brown Kelly" - while noting that O'Kelley says it is not his. AdCP arrived on October 15, 2025 with six founding members - Scope3, the company O'Kelley leads, alongside Yahoo, PubMatic, Swivel, Triton Digital and Optable - plus 23 further participants and nine core tasks built on the Model Context Protocol. Governance sits with the nonprofit AgenticAdvertising.org. Scope3 cut engineering and sales staff in February 2026, five months after an earlier round, and has since dropped its sustainability branding. O'Kane's LinkedIn post, for its part, opened with the "recent pivots (so many pivots) and rebrands in ad tech".

Schlosser declined to speculate: "I leave that up to you on the motivation." His practical reading was narrower. A company heavily invested in OpenRTB will find it easier to ramp into AAMP, while a CTV publisher not yet trading through OpenRTB, wanting to test insertion orders without selling by phone, is more likely to set up AdCP.

Google came next. O'Kane said the company had just released its own agent for standard insertion orders and had "basically won their case to not be broken up". The transcript does not name the product, and PPC Land's searches did not locate a corresponding release. The court outcome is also more complicated than a win. Judge Leonie Brinkema rejected the Justice Department's request to divest AdX on September 2, and the 106-page opinion unsealed on September 16 imposed six years of behavioural remedies with worldwide reach, including interoperability between Google's ad tech and Prebid. Schlosser had characterised that outcome to Digiday on September 18 as six years of algorithmic hide-and-seek.

His answer on Google was that it does not change the calculation. Complexity and cost are what count, and "at the end this will be a discussion on margin of the platforms". Asked whether heavyweights such as Google might lean into either protocol, he said: "I have to admit I do not know." Google was also absent from AdCP's founding roster, along with Amazon, The Trade Desk and Microsoft, a gap ExchangeWire's chief operating officer Lindsay Rowntree had noted on a podcast.

350 days or 50

The episode's most concrete numbers concerned engineering effort. According to Schlosser, a custom deals API integration with a single SSP would cost Adform 250 to 350 days of work, as the captions render the range; built against a shared standard, perhaps 50. The transcript garbles the specification he cited, apparently an IAB Tech Lab deals interface. IAB Tech Lab's Deals API reached final status as version 1.0 on February 6, 2026, after opening for public comment on December 5, 2025.

For agentic integrations the gap widens. A custom agentic connection with every SSP "on the planet" would take thousands or tens of thousands of days, he estimated. "When I build the standard, even if it's just two, it will cost almost nothing to connect." A seller registers an agent, the buyer agent discovers it automatically, and the two speak the same language. Whether that is French, English or German matters little, a metaphor he returned to at the close: "not every English word can be translated to French or German, but you will have 90% understanding."

Proofs of concept, not demand

On adoption, Schlosser cited version 2.3, with 3.0 expected soon, and a reference implementation. The whiteboard references make it hard to tie each version number to a protocol from the transcript, but 2.3 matches the AAMP release PPC Land reported on July 30, 2026, which added a pricing provenance field to stop buying agents inventing prices. A reference implementation was in use by March, when Kochava opened an AAMP workspace running through IAB Tech Lab's reference MCP server.

For the other protocol he did not have the numbering to hand. He described it as probably a little further along on package discovery, lighter to implement and quicker to prototype - producing many proofs of concept but little incoming demand, because the largest platforms have not built it. AAMP, by contrast, is seeing first integrations. He judged it more likely to grow because it fits existing systems and gives "the big armies of developers" something to build on. Demand will decide: if publishers put supply on AdCP, buyers would follow, "but usually the sellers do what the buyers ask them to".

He had spoken to a couple of SSPs over the previous two months. Their message, as he relayed it, was that the choice between AAMP and AdCP makes little difference, and that a quick, lightweight AdCP implementation already exists on their side. The rest of that remark is garbled in the transcript and cannot be reliably reconstructed.

His forecast allowed three outcomes: "either one will die or they will be merged or they will coexist". Either way, both will be heavy at the execution layer. OpenRTB stays, any alternative has to match it, and Trusted Match at 50 milliseconds is "also not cheap". Real-time pricing - bidding differently for a user with a 25 per cent likelihood to buy than for one at 5 per cent - "is not even in here yet". He doubted there was much complexity "that you can just cut", and expects any non-OpenRTB execution layer beneath AdCP to become "rich and heavy". For technology companies, he said, running two protocols is tolerable if both work as standards.

Little money, rising integration bills

The episode adds a platform operator's arithmetic to a debate PPC Land has followed since AdCP split the industry within weeks of its debut. IAB Tech Lab named its own umbrella initiative AAMP on February 26, 2026, and by March 11 its Agent Registry held 10 entries, every one an MCP server.

Money has lagged the specifications. DataBeat's June report found conventional programmatic buyers clearing at a $6.95 CPM against $6.13 for agentic buyers, a 13.4 per cent gap, with agentic demand entering 86 per cent fewer auctions. Magnite chief executive Michael Barrett put the most optimistic 2027 forecasts for agentic spend at $600 million to $700 million. Price pressure is arriving ahead of volume: rivals to The Trade Desk have been wrapping guaranteed deals in agentic tools at fees near 1 per cent. If agent-led discovery cannot be charged for, as Schlosser argued, his margin question applies to every platform in the chain.

Vendors are hedging. VIOOH's seller agent automated more than 100 curated DOOH deal package requests in the first half of 2026, with planned registration under both AdCP and AAMP. Joel Meyer, OpenX's chief technology officer and Prebid's chair, has said that from an engineering standpoint a single protocol is always preferable. Schlosser is more relaxed about duplication, but his position rests on a condition: that platforms such as Adform absorb the cost of speaking both languages. Advertisers stay indifferent only while that cost stays out of their fees. For publishers the choice is concrete - sellers, Schlosser noted, must pick with their technology partners, and one exposing inventory to both frameworks maintains two descriptions of the same supply.

Timeline

Summary

Who: Jochen Schlosser, chief technology and product officer at Adform, the Copenhagen-based advertising technology company that runs a DSP, ad server and SSP, interviewed by Ciarán O'Kane, co-founder and adviser at ExchangeWire. The protocols discussed are IAB Tech Lab's Agentic Advertising Management Protocols (AAMP) and the Ad Context Protocol (AdCP), governed by AgenticAdvertising.org.

What: A whiteboard comparison of the two agentic advertising protocols. Schlosser said both serve the same purposes - campaign setup, inventory negotiation and discovery - with AAMP building on OpenRTB and OpenDirect and AdCP defining an execution-agnostic layer. He put Trusted Match's service-level agreement at 50 milliseconds against 20 to 100 for OpenRTB, estimated a custom deals integration at 250 to 350 days of work against about 50 on a standard, agreed that very little money currently runs through either protocol, and forecast that one will die, the two will merge or they will coexist.

When: ExchangeWireTV published the episode on September 23, 2026, three weeks after a federal court declined to break up Google's ad tech business and less than two months after AAMP 2.3.

Where: On ExchangeWire's Trader Talk TV programme, distributed through YouTube and promoted by O'Kane on LinkedIn. The debate itself concerns the global programmatic supply chain, from connected television and digital out-of-home to channels outside OpenRTB such as search and social.

Why: Buyers are being offered two overlapping standards while little spend flows through either. Schlosser's view is that advertisers need not care which protocol a platform uses, but that indifference depends on DSPs, SSPs and publishers absorbing the cost of supporting both - and, as he put it, on how that cost is eventually settled in platform margins.