The Trade Desk announced Kokai Zuma on August 27, 2026, and the feature list reads like a wish list from every programmatic buyer who's complained about the platform's UI since the Solimar days. Agentic AI assistants. Automated audience building. One-click brand lift studies. Bulk-change modeling. The periodic table interface? No longer the default view.

The question for ops teams isn't whether these features sound good. It's whether they actually reduce cycle time, cut errors, and produce measurable lift in a stack where 94% of enterprise GTM teams report their infrastructure isn't AI-ready.

What Zuma Actually Changes

Three buckets matter. First: the Koa Assistant, a conversational AI tool now in open beta that lets buyers prompt specialized agents for audience creation, frequency optimization, troubleshooting, and mid-flight performance pulls. Buyers can also connect third-party agents via TTD's Open Agentic Kit. "Our preferred use case is using the in-platform AI assistant," said Samantha Ross, senior director of product marketing at The Trade Desk.

Second: a UI overhaul. Key metrics and data visualizations now sit at the top of every campaign landing page. Bulk changes across dozens or hundreds of ad groups happen simultaneously, and a new modeling feature shows projected impact before changes go live. Settings that block ad delivery get flagged in red; agentic recommendations show up in yellow.

Third, and probably the most consequential for B2B teams trying to prove incrementality: measurement shortcuts. Launching a Lucid brand lift study used to require about a week of coordination. Now it's a single click inside Kokai, because TTD automated the feasibility and setup checks on its backend. Nielsen IQ lift studies got the same treatment. TTD also added geo-level conversion lift measurement and ecommerce reporting that breaks down household penetration by retailer. "For too long, the industry has relied on proxy metrics like CTR, CPA and video completion rate," Ross said. The new tools push toward "business growth, sales and incrementality."

The Context That Complicates Things

Agentic workflows in DSPs aren't a TTD invention. Google DV360, Amazon DSP, and Yahoo DSP have all shipped similar capabilities. Google is going further: starting September 1, 2026, eligible Search campaigns begin automatic migration to AI Max, making AI-driven campaign management the default. Zuma doesn't give TTD a monopoly on AI-assisted buying. It brings the platform closer to parity with competitors who've been shipping these tools for months.

There's also the credibility question. CEO Jeff Green acknowledged Kokai hadn't delivered the expected advertiser efficiency improvements. Honest admission, but it means ops teams should treat Zuma's promises as hypotheses, not guarantees. The Audience Creation agent sounds useful, especially for buyers paying TTD's 4.4% impression fee for Audience Unlimited access. Whether it actually reduces audience-building time or just reshuffles where the work happens is something you'd need to measure in your own account.

What to Measure (and What Not to Over-Interpret)

Cycle time. Track how long it takes to go from media plan to live campaign before and after enabling the Koa Assistant. Baseline the current process in hours. A 30%+ reduction is meaningful; anything under 15% might just be novelty effect.

Error rate. Count delivery-blocking settings flagged by the new Applied Settings view in your first two weeks. If the red flags catch issues you'd previously missed until mid-flight, that's real operational value.

Lift study velocity. The one-click Lucid integration should compress a week-long setup into minutes. Measure whether your team actually runs more lift studies as a result. Running them faster only matters if you run more and use the results to reallocate spend.

Guardrail: data quality. With 70% of enterprise GTM teams reporting poor data quality, any AI-generated audience segment needs human review before activation. Don't skip QA because the agent made it feel easy.

The Structural Question Nobody's Asking

TTD operates on the open web. Google, Meta, and Amazon own walled gardens with broader user data and more tightly integrated AI ad systems. "Easy buttons" reduce friction inside TTD's platform, but they don't change the competitive data advantage walled gardens hold. For B2B buyers, the open web's targeting signal has always been thinner, and layering AI on top of thin signal doesn't automatically make it thicker.

Jaime Nash, senior director of product marketing at TTD, framed the Audience Creation agent as making "purchase-level data more accessible and discoverable for advertisers." That's the right direction. But accessible data and high-quality data aren't the same thing, and the distinction will determine whether Zuma's agentic features produce real lift or just faster execution of mediocre targeting.

Adoption numbers tell a similar story. According to Salesforce's 2026 marketing benchmark, AI use in marketing rose to 24.2% of marketing activities, up from 13.1% in 2024. Yet only 22% of B2B organizations report fully implemented gen AI capabilities. The gap between "we use AI somewhere" and "AI reliably improves our outcomes" is where most teams live right now.

Zuma's measurement improvements are probably its most durable contribution. One-click lift studies and geo-level incrementality testing lower the barrier to proving (or disproving) whether programmatic spend actually drives business outcomes. That's the part worth adopting immediately, even if you're skeptical about the agentic features. Proof compounds. Features come and go.