AI on the Supply Side: How PubMatic Is Challenging Ad Tech's Old Guard

REDWOOD CITY, CA – June 22, 2026

In the intricate, often opaque world of digital advertising, a decisive battle just took place, and the results could signal a fundamental shift in how billions of dollars in media are spent. Performance marketing firm Level Agency recently pitted PubMatic's new AI-driven platform, AgenticOS, against its trusted, incumbent Demand-Side Platform (DSP). The outcome was not a marginal victory; it was a rout. By every meaningful metric, the new system dominated, delivering more than double the reach for every dollar spent.

This isn't just another incremental improvement in a hyper-competitive field. It's a proof of concept for a new architecture in programmatic advertising—one that moves intelligence from the buyer's side deep into the supply chain. The test results validate a strategic bet by PubMatic that by using a fleet of autonomous AI agents operating directly where ad inventory lives, it can offer advertisers a level of efficiency and speed that challenges the very structure of the market, including the dominance of the industry's walled gardens.

A New Benchmark for Performance

For any advertiser, but especially for sectors like education that demand hyper-local, time-sensitive precision, the goal is simple: reach the right audience, at the right time, for the lowest cost. Level Agency structured a test to find the most efficient path to that outcome. With identical budgets and objectives, AgenticOS and a traditional DSP were run in parallel for an education client's campaign.

The data, tracked by Level's own reporting, was unambiguous. AgenticOS delivered 351 qualified impressions for every dollar, compared to the incumbent's 164. This greater than 2x improvement in reach per dollar was consistent across mobile, online video, and the highly coveted premium Connected TV (CTV) space. For premium CTV, the platform achieved 53% lower weighted average costs-per-thousand-impressions (CPMs). The quality of that engagement was also higher, with video completion rates jumping by 54% overall.

"At Level, every technology decision starts with one question: will this move the needle for our clients?" said Patrick Van Gorder, Chief Partnership Officer at Level Agency. "What AgenticOS delivered changed how we're thinking about where the open internet can compete for client budgets, and that's exactly the kind of adaptive advantage and innovation we're always looking for."

Perhaps most strategically significant was the speed of execution. AgenticOS activated the campaign in minutes and began retargeting audiences at scale within days—a process that typically requires a 30 to 60-day ramp-up period with DSP-led campaigns. This agility allowed Level to pivot creative from a social campaign and immediately deploy it for a date-specific campus open house, an activation that would have been difficult to execute with such speed through traditional channels.

Inside the Machine: The Agentic Advantage

The dramatic performance gap isn't magic; it's architecture. Traditional DSPs operate on the buy-side, making decisions based on data signals they receive from the supply side. This model has inherent latency and often relies on incomplete information. PubMatic's AgenticOS flips this paradigm. It operates within the supply chain via Activate, its direct activation platform.

Instead of a single buyer agent, AgenticOS coordinates a sequence of over 20 specialized AI agents that manage the entire campaign lifecycle. These agents handle everything from audience discovery and inventory filtering to activation and optimization, all before the auction even happens. They work with a full, unfiltered view of inventory and data signals, allowing for more intelligent and efficient decision-making. For Level Agency, this entire stack was accessible through a simple, plain-language prompt defining their campaign objectives.

"The best media strategies are the ones that don't require constant intervention," noted Anjlee Majmudar, VP of Programmatic at Level Agency. "When your supply pool is already controlling frequency, filtering inventory, and qualifying audiences before the auction, you get scale that manages itself — and that frees you up to make better decisions faster."

This approach directly addresses a long-standing inefficiency in programmatic advertising, where a significant portion of every dollar is lost in the complex chain between advertiser and publisher. By simplifying the ecosystem, PubMatic aims to put more of the advertiser's budget into "working media," a goal articulated by Andy Bryant, AVP of Advertiser Solutions at PubMatic, who stated the platform was designed to make "the open internet as a legitimate competitor to the walled gardens."

Shifting the Balance of Power

Bryant's comment cuts to the core of the strategic implications. For years, advertisers have flocked to the walled gardens of Google and Meta for their vast user data and efficient, integrated ad systems. The 'open internet'—the sprawling ecosystem of independent publishers, blogs, and streaming services—has struggled to compete on efficiency and ease of use. Technologies like AgenticOS represent a concerted effort to change that dynamic.

By delivering superior ROI and radical speed, PubMatic is making a compelling case for advertisers to allocate more budget to open internet properties. The success of Level's hyper-local, time-sensitive open house campaign demonstrates that the open internet can now effectively compete for performance-driven budgets that were previously the exclusive domain of social platforms. This is not just a theoretical possibility; Level Agency made a mid-flight decision to expand the program, allocating a new, time-sensitive campaign to AgenticOS alongside its existing social buy.

This shift is occurring as the broader industry rapidly embraces AI. Recent IAB research indicates that two-thirds of digital video buyers are already using or plan to implement agentic AI campaigns in 2026. PubMatic is positioning itself at the forefront of this wave, not just by building its own agents but by creating an open system. With the recent launch of Decision Fabric, it is allowing third-party AI models to run natively within its platform, extending the architectural advantage to a wider ecosystem of partners.

From Test Case to Market Force

The success with Level Agency is not an isolated incident. PubMatic has been systematically rolling out AgenticOS with global partners, including WPP Media, Butler/Till, and MiQ. Campaigns for clients like Geloso Beverage Group and Interbev have shown similar patterns of increased efficiency and improved performance. With over 20 operational agents already available and new ones rolling out monthly, the platform is rapidly evolving from a promising concept into a scaled, market-ready force.

The company is also expanding its inventory, recently launching a Creator Marketplace for CTV advertising that connects independent creator media with agentic demand. By processing trillions of bid requests daily across thousands of publishers and streaming apps, PubMatic has the infrastructure to support this new model at scale. The initial results are in, and they suggest that the hidden currents of the digital economy are beginning to flow in a new direction, powered by autonomous agents operating at the heart of the supply chain.