In January 2025, agentic media trading was being discussed at industry events largely in theoretical terms. It had only been two months since the launch of Model Context Protocol (MCP), a standard which allows AI tools to connect to external tools and systems, and Google was only just heralding the dawn of the agentic era for AI as a whole. Media-specific use cases still felt some distance off.
At least, that’s how it felt to some of us at the time. Over the course of the year, theoretical discussions were fleshed out into industry initiatives and protocols — notably Ad Context Protocol (AdCP) and IAB Tech Lab’s Agentic RTB Framework (ARTF) — laying the groundwork for agentic trading.
Now, at the start of 2026, we’re starting to see agentic media tools coming to market. One early mover was Yahoo DSP, which used this year’s CES exhibition to showcase new agentic AI solutions embedded in its platform. Rather than just building its own agents, Yahoo has taken what it describes as a ‘Yours, Mine, Ours’ approach. That is to say, customers can interact directly with Yahoo’s agents, they can use their own agents to interact with Yahoo, and/or they can work through agent-to-agent connections.
Specifically, advertisers and partners can use the MCP to connect third-party AI tools with Yahoo DSP, allowing them to activate campaigns through those third-party tools — the ‘Yours’ part of the offering. On the ‘Mine’ front, Yahoo has built a troubleshooting agent which identifies common pacing and delivery issues, flagging fixes and executing with approval. Finally the ‘Ours’ solution covers audience exploration, whereby users’ agents can access Yahoo DSP audience metadata through APIs or MCP, and present it to the user based on its own settings and configuration.
Goodbye, UI
Adam Roodman, GM of Yahoo DSP, says the company is continuing to work on new agents within its own platform, as it looks for areas where agents can add value. “We have a horizon of the middle of this calendar year to have virtually all use cases covered,” he said. “But I also know that whenever we think we have every use case covered, some new use case pops up, either from a client or someone on our sales and service team.”
These use cases won’t always be flashy. In fact, often it’s the unsexy parts of the workflow where agents can make a big difference, cutting out laborious manual tasks. “Think about billing reconciliation,” said Roodman. “So many hours are poured into that on the buy-side and the sell-side, that’s a lot of routine work which an agent can cut out.”
But AI agents won’t just be time savers, either. By changing the ways buyers interact with their DSPs, agents could completely change the competition dynamic in the space.
“From a workflow perspective, we really do believe that agentic allows a DSP to change its definition,” said Roodman. “Historically, a DSP was a workflow that allowed you to select certain items and navigate a web-based UI, and get outcomes as a result. With agentic, you don’t even need to use our agents, and you may never even see the Yahoo DSP’s UI.”
In that world, DSPs will compete more strictly on performance, and be incentivised to invest more in the things which really drive outcomes, such as data, bidding, and measurement solutions.
“There are some DSPs that have done really well for themselves because they’ve gotten large groups of traders really familiar with their own UI, which makes it hard to switch,” said Roodman. “But if you think about using a prompt-based, personalised chatbot for your own company, you don’t have an issue with pointing that at any particular DSP. So now you can evaluate them based on their ability to drive performance, not just your familiarity with their UI.”
As far as AI tools go, it’s really their ability to deliver on those DSP fundamentals which clients are interested in.
“At CES, our customer conversations were agentic forward, but they jumped right into all those things at the core of the DSP,” said Roodman. “They wanted to hear about our commerce media footprint, they wanted to talk about our in-flight outcome product for offline conversion. Our customers still think about outcomes first. They’re excited about agentic, but they’re excited because they see how it gives them better access to performance products and better access to data.”
What wins in CTV?
Competition in the DSP space continues to heat up, particularly around CTV, which Roodman says is a core part of Yahoo DSP’s positioning. “We fit in the enterprise space mostly today, serving large advertisers with a CTV-first mindset,” he said. “I’d say that’s been the case for the last three to four years.”
That’s a surprise to some, says Roodman, given the company’s heritage linked to its owned-and-operated property. But while display is still an important part of the picture, it’s really the data spine which comes from that owned-and-operated media which provides a competitive advantage, and that data spine has plenty of value in the CTV space.
Inventory access also has an important part to play, and Yahoo DSP has had success on that front, signing notable deals with Netflix, Roku, and Spotify last year, while also getting access to Disney’s live sports and entertainment inventory. These sorts of deals are obviously very sought after for DSPs, but aren’t handed out to everyone, so what makes the difference when it comes to sealing them?
“Having gone through the process with Netflix and others like it, they’re smart and they’re incredibly diligent in talking with their customers,” said Roodman. “They ask their customers where they should make their next move, so the platforms they choose to work with is a reflection of that customer feedback. So it’s like the ultimate testimonial that we’ve built a really good CTV product.”
Follow VideoWeek on LinkedIn.
