An industry coalition of ad tech businesses, media businesses, and buy-side companies have today launched ‘Ad Context Protocol’ (AdCP), a new open standard for agentic advertising infrastructure. The protocol sets standards for the ways in which different AI agents communicate with advertising platforms, and with each other, when buying and selling ad inventory. It’s designed to work with existing tools, helping AI agents work more efficiently with existing programmatic infrastructure, but further down the line, it’s set to facilitate a whole new way of transacting, where buy-side agents and sell-side agents talk with each other to buy and sell ads.
The initiative’s founding members are Yahoo, Optable, PubMatic, Scope3, Swivel, and Triton Digital, while it’s also being supported by The Weather Company, Classify, Magnite, MiQ, LG Ad Solutions, Kargo, Butler/Till, Raptive, and Samba TV, among others. AdCP is governed by a neutral working group, and not owned by one single company. The founding members say it is designed for open contribution and broad adoption.
At launch, AdCP includes protocols for audience activation, helping advertisers to discover and activate audience segments directly from publishers, and media buying, enabling agents to execute and manage campaign line items across platforms. The founders say the toolset will expand as the project evolves.
Built for now, and the future
Buyers, sellers, and ad tech companies are rapidly rolling out agentic tools and capabilities for advertising, as a means of cutting out manual work, but also creating more efficient and effective ways of surfacing inventory, finding audiences, and trading ads.
For example, a buyer might want to use an AI agent to seek out a particular audience on a publisher’s properties using a natural language prompt, rather than finding or building one in a demand-side platform (DSP). Or a publisher might want to use an agent as an easier means of selling inventory types which programmatic tools don’t cater to.
A range of these types of tools are already being built. But in order to work, they need to be able to plug into the existing set of technologies which are commonly used across the industry. So the initial task for AdCP has been to ensure that AI agents can speak with these existing technologies using a common language. This ability to connect into existing infrastructure is key, allowing new agentic technologies to quickly benefit from the reach and interoperability which is crucial for digital advertising.
Further down the line, however, AdCP will provide a way for agents to communicate and transact without the need for existing infrastructure. Scope3’s co-founder and CEO has been vocal about his vision for advertising’s AI-based future, where agents acting on behalf of buyers go out in search of inventory which is appropriate for a given campaign brief, applying the brand’s criteria around brand safety and suitability.
Anne Coghlan, Scope3’s COO, outlines an example. “A brand could create an agent and tell it, ‘I want to find women that are interested in rock climbing in the UK’, and send that request out to the ecosystem,” she said. “Then publisher and platform sales agents could respond saying ‘this is the inventory that I have, this is the pricing’, and respond with relevant audiences. All that would be enabled through the AdCP protocol, with the benefit that the advertiser doesn’t need to have individual integrations with every single publisher in the world.”
Not all in the industry agree that’s the direction we’re headed. But the presence of several established players in the current programmatic ecosystem suggests there’s growing belief in an agent-to-agent future for digital advertising.
OpenRTB for agentic advertising
Regardless of how commonly these sorts of agent-to-agent transactions eventually become, AdCP should help accelerate the speed with which they’re able to get off the ground. The project’s founding members draw a parallel with OpenRTB’s role in accelerating the growth of programmatic.
And companies backing the project say there are significant gains to be had from agent-to-agent transactions.
David Olesnevich, chief product officer, The Weather Company, said that by using selling agents, publishers can more effectively convey the full value of their inventory to the buy-side.
For example, one of The Weather Company’s biggest assets is its bank of weather data and its deep audience understanding. There’s lots of value here for advertisers, but some of that value is lost when the data needs to fit into rigid programmatic tools. With agentic technologies, these same restrictions don’t exist. The same could be true for niche formats which aren’t catered to by today’s technology.
“Premium assets can just get wrapped up into generic impression inventory and commoditised,” said Olesnevich. “Now we’re able to represent the true value of those premium products.”
For advertisers meanwhile, AI agents open up opportunities to access portions of the market which aren’t available through programmatic marketplaces, without the need to set up direct deals either. And Joe Hirsch, CEO of Swivel, said that buyer agents will massively simplify product discovery.
“If you think about programmatic today, if you were a first-time media buyer and you wanted to spend $100,000 on programmatic, the probability of being successful is near zero,” said Hirsch. “Your budget would immediately be consumed by essentially whoever in the bid stream has the greatest volume of ad requests, the barrier to entry in programmatic is pretty high.”
A buy-side agent meanwhile would do the work of figuring out which audiences an advertiser needs to reach, and where they can be found. “That means there is a chance that when a first-time buyer buys agentically, they could be successful right away.”
Follow VideoWeek on LinkedIn.
