The programmatic advertising ecosystem has long promised efficiency, yet for many buyers and publishers, the reality remains a sprawling, friction-heavy maze. Speaking at the Advertising Economic Forum in New York during the AI & AdTech Founder Showcase, industry leaders gathered to address the inertia holding the market back and explore how agentic AI could fundamentally reshape media trading.
Bob Lord, President of Horizon Media Holdings and CEO of Horizon Global, set the stage by challenging the current state of adtech, pointing out that despite trillions of dollars flowing through programmatic channels, efficiency remains elusive.
“What is so profound to me is the inertia in this adtech and martech world,” Lord said. “It is enormous. When you look at every other industry, technology has revolutionised everything from supply chain to manufacturing to asset management and banking. But in martech, not so much. It is still more confusing to the CMO and media providers than ever.”
Lord highlighted how original programmatic promises have stalled over time. “If I rewind 10 years, the promise of programmatic was to simplify everything,” he said. “Now we have got 3.5 trillion dollars going through programmatic and a CMO cannot even tell whether or not their dollar is effective.”
From supply chain friction to agentic orchestration
Addressing this systemic inertia requires moving beyond traditional layers of DSPs and SSPs, said Andrew Mole, CEO of PubX.AI. He explained how agentic infrastructure aims to bypass the tollbooths that currently drain value from media spend.
“We initially started on the supply side, trying to bring machine learning to publishers,” Mole said. “We saw the challenges they were facing within the existing programmatic ecosystem, where around 64 cents of every dollar goes to the actual media spend. We wanted to try and fix that on a publisher-by-publisher basis, but over the years we realised this is a macro problem.”
To tackle this macro issue, PubX.AI transitioned toward what Mole describes as agentic infrastructure, a system where autonomous buyer and seller agents communicate directly.
“We are not a layer over the DSP or SSP,” said Mole. “We connect the advertiser directly to the publisher. We have a buyer agent and a sales agent that interact, removing the existing infrastructure from the equation because that is where the massive transaction costs come from.”
Dispelling the fear of AI hallucinations in trading
A primary concern among marketers considering autonomous AI systems is loss of control or risk of AI “hallucinations” in execution. However, Mole emphasised that agentic protocols operate within strict, rule-based parameters designed to guarantee media quality and compliance.
“I think hallucinations are blown out of proportion within an agentic ecosystem,” Mole argued. “It is a protocol. Agentic systems use protocols like MCP, which give you the rules of how you operate. As long as the buyer agent observes those rules and the seller agent observes those rules, you do not have this problem with hallucinations because ultimately it is software speaking to each other within a set remit.”
This setup allows trading teams to shift from manual campaign configuration to higher-level strategic direction.
“The machine actually respects the brief,” Mole explained. “It respects the brief and can understand it at a scale that is beyond human capability. A trader on a standard DSP is limited by what they can manage. They are not set up to manage thousands of deals between different publishers because they simply cannot handle that volume, nor could they optimise it across all running campaigns.”
Mole added that autonomous agents can instantly discover micro-pockets of high-performing inventory: “If the agent finds a pocket of inventory that is 10 impressions and 10 amazing pieces of data, it can run that as a segment in the campaign and constantly monitor performance, moving spend around to deliver the best return.”
A conversational future for media planning
Rather than replacing human oversight, agentic adtech aims to change how traders and planners interact with technology, turning execution into an ongoing strategic dialogue.
“You have three levels of autonomy, and all of this is about governance,” Mole stated. “If you are a buyer, you put in your parameters, your geo, your targeting, and your intent for what you want to achieve. The agent optimises towards the fitness function of those two combinations to deliver results.”
This conversational model frees buyers from mechanical setup tasks. “As a planner or buyer, you can go into the agent and say, ‘Hey, actually we want to update the goal. We have hit our KPI and now we want reach.’ The agent can update and go find the inventory that meets that,” Mole said. “It transforms planning into a conversation. As a trader, you are not using your time setting things up on both sides; you are having a conversation with the agent.”
By cutting tech overhead and streamlining execution, the goal is to direct a far higher proportion of media budgets back into working media.
“When we started doing this, it was about getting tech through, which took a huge amount of work to get us to where we are,” Mole concluded. “Now, what we are trying to do is educate people on the promise of the new, putting more of the money back into the productive part of media.”
