The Trade Desk launches Kokai Zuma with agentic Koa AI and 32 percent CPA gains
A new AI assistant called Koa builds campaigns and audiences with minimal manual input inside Kokai.
Mariano De Vitto · September 2026
The Trade Desk has rolled out Zuma, the latest update to its Kokai platform, introducing agentic AI capabilities through a new assistant called Koa that can build campaigns and audiences with minimal manual input. According to AdExchanger, the company reported a 32 percent improvement in cost per acquisition across 62 campaigns that used these new AI-driven tools, marking one of the most concrete performance claims yet from the independent DSP as it pushes deeper into automated buying.
Kokai has served as The Trade Desk's flagship trading interface for the past couple of years, designed to give media buyers richer predictive signals and more efficient bid optimization than earlier versions of the platform. Zuma represents the next layer on top of that foundation, shifting from giving traders better data to actually acting on that data through automation. It is a natural evolution for a company that has positioned itself as the leading independent alternative to walled-garden ad platforms, where scale and automation increasingly determine whether buyers can compete on efficiency.
Koa functions as an AI agent embedded directly in Kokai, capable of interpreting campaign goals and translating them into audience segments and bidding strategies without requiring traders to configure every setting manually. This positions The Trade Desk alongside a broader industry shift toward agentic AI in programmatic buying, where platforms move beyond optimization suggestions and start executing tasks that previously required dedicated media buyers.
Agentic AI Enters Programmatic
The distinction matters. Traditional programmatic tools have long offered recommendations, flagging underperforming segments or suggesting bid adjustments that a human still had to approve and implement. An agentic system like Koa closes that loop, interpreting a stated goal and executing the underlying steps itself. For trading desks and agencies juggling dozens of active campaigns, that kind of automation can meaningfully reduce the hours spent on manual setup and ongoing optimization, freeing strategists to focus on higher-level questions like creative testing, channel mix and measurement.
For marketers, the Zuma update signals that The Trade Desk is betting heavily on AI reducing the operational complexity of programmatic campaigns, particularly for teams managing multiple audience segments across CTV, display and other channels. The 62 campaign sample size reported by AdExchanger is not massive, but the 32 percent CPA gain gives buyers a tangible benchmark to evaluate whether Koa's automated approach outperforms their existing manual workflows.
What Trading Teams Should Do Next
The competitive backdrop is worth keeping in mind. The Trade Desk operates as an independent DSP, competing against demand-side platforms tied to larger walled gardens as well as retail media and CTV-specific buying tools. Demonstrating a concrete performance lift from its own AI tooling helps the company make the case that independence does not mean falling behind on automation, a narrative that matters as more advertising budget shifts toward platforms that can prove efficiency gains with real numbers rather than roadmap promises.
Why it matters: Marketing teams running programmatic campaigns through The Trade Desk should treat Koa as a tool worth testing rather than a wholesale replacement for strategic oversight. The reported CPA gains suggest real efficiency potential, but growth leaders should validate results against their own KPIs and audience definitions before shifting significant budget toward fully agentic campaign creation.
The Signal Brief · Mariano De Vitto — Head of Marketing, Barcelona