The Trade Desk’s revenge?

John Ebbert

The Trade Desk and Zuma

“New AI-driven optimizations that drive a 32 percent average improvement in CPAs, at scale.” – Eric Bosco, VP Product, The Trade Desk on LinkedIn

The Trade Desk launched Kokai Zuma today, overhauling its buying platform with a redesigned interface, expanded AI capabilities and easier access to measurement.

Ad Age senior editor Garett Sloane puts the release in context:

“The announcement is basically a redesign, software update, but what The Trade Desk focused on shows where it’s trying to be responsive to some of the ad buyers who had issues with the old system.

Zuma has an applied settings page that gives an overview of what toggles are turned on and off in campaigns, and a way to easily control them. If you recall, some ad buyers complained about hard to find settings that allegedly switched themselves on and off.

The new platform has a chat interface, making Koa more of a personal assistant. There is also an agentic toolkit that will make it interoperable with buyers’ AI.”

Read more on LinkedIn. (August 27)

Two things worth highlighting:

  • Workflow: As tipsheet covered yesterday, Koa Assistant routes buyers to specialized agents for campaign creation, insights, troubleshooting and other tasks. And the direction points beyond the DSP interface: The Trade Desk’s Open Agentic Kit, launched in April, is designed to let buyers access those capabilities through outside AI tools.
  • Measurement: Zuma makes Conversion Lift experiments easier to incorporate into campaigns, bringing incrementality closer to the normal buying workflow.

Read:

  • Story: The Trade Desk is overhauling its ad-buying platform—here’s what marketers need to know (August 27) – Ad Age (subscription)
  • Product: The Trade Desk Introduces Kokai Zuma, the Latest Release of Kokai – The Trade Desk
  • On the launch: “TTD’s role has always been to be a challenger and we’re at our best when we’re taking steps forward for the open internet…” (August 27) – Jeff Green, CEO, The Trade Desk on LinkedIn

From tipsheet: Just in time for Q4, The Trade Desk is giving traders more visibility into Kokai just as Koa starts doing more of the work itself. Can that get growth moving again at TTD? We’ll see.

CONNECTED TV

Nielsen, The Trade Desk partner for decisioning

  • “Gracenote’s first DSP integration brings show-level content intelligence to The Trade Desk” (August 27) – Nielsen’s Gracenote

tl;dr: Gracenote’s intelligence is becoming portable across the ad tech stack.

From tipsheet: Gracenote’s show-level signals already enrich publisher and SSP inventory. Now they can inform decisions directly inside The Trade Desk. At the same time, containerization is moving buyer models in the opposite direction, closer to supply.

The boundaries between sell-side and buy-side decisioning are starting to blur.

COMMERCE MEDIA

Amazon meets auto-tagging on YouTube

Amazon has joined the YouTube Shopping Affiliate Program, letting eligible U.S. creators tag Amazon products directly in Shorts, long-form videos and livestreams and earn commissions when viewers buy.

In a YouTube Short, YouTube creator liaison Rene Ritchie explains:

  • “If you have auto tagging enabled or choose to turn it on, YouTube systems may automatically review your recent uploads to identify and tag eligible Amazon products.”

YouTube says a July experiment found videos using its Shopping product tags generated more than 110% more product clicks than description links alone.

Visit: YouTube Shopping Creators can now tag Amazon products (August 27) – YouTube Official Blog

More: Tag Amazon products with YouTube Shopping (August 27) – YouTube Help

From tipsheet: YouTube is marrying its creator and discovery engine to Amazon’s catalog, transaction and purchase signals. Add auto-tagging, and it all starts to look like a revenue machine.

Related: YouTube CEO Neal Mohan speaks to Stratechery’s Ben Thompson about the auto-tagging opportunity (September 2025) – tipsheet

AGENTS

Opinion: The prompt will disappear

Yobi CEO Max Snow believes that marketers are mistaking today’s prompt-based AI agents for the end state.

Writing in The Trade Desk’s The Current, Snow says that instead of describing a campaign to an AI agent, marketers will eventually point the system toward the outcome that matters, such as a purchase, new account or upgrade, and let it autonomously plan, buy, measure and optimize toward that goal.

What replaces the prompt? Snow points to AI trained on behavioral data rather than text:

“The same way language models learned to write by reading, behavioral models learn to predict what consumers will do by studying what consumers have actually done. When you designate a purchase event as your goal, this kind of model already understands the patterns that lead to a purchase like yours.”

Snow also argues the intelligence generated by that outcome data should live with the brand and follow its spend across the open web and CTV.

Read: The future of agentic AI will not be prompt-based (August 27) – Max Snow on The Current

From tipsheet: Goodbye prompt engineer, we hardly knew ye: the valuable input isn’t the prompt but the proprietary signal that tells the machine whether it succeeded.

MARKETING

Gradial builds ‘enterprise marketing harness’

AI marketing startup Gradial launched new capabilities Wednesday that let its agents create on-brand marketing assets at scale.

The news comes after the startup raised $65 million in Series C financing (at a $650 million valuation) in June with the goal of building “the First System of Work for Enterprise Marketing.” Read the release.

In Wednesday’s AdExchanger article, Gradial CEO and co-founder Doug Tallmadge gives that ambition another name: an “enterprise marketing harness.”

Joanna Gerber reports:

“Enterprise software systems have historically demanded a lot of cross-team communication, said Tallmadge. The marketing team might oversee brief writing and incorporating revenue goals, while the product team is building the actual product and designers are revising ad creative based on feedback from tests run by performance marketers.

Gradial was designed to bring together all the moving parts.”

Read: Gradial Is Ready To Unify The Marketing Workflow With Its AI Harness (August 26) – AdExchanger

From tipsheet: Gradial is another example of an agentic layer emerging above the existing marketing stack, pulling together context, executing work and optimizing against results. Salesforce is building toward a similar destination from inside its own stack. Gradial is trying to sit above everyone else’s.

Buy me, Marc B.: Gradial is promoting its attendance at Salesforce’s Dreamforce 2026 next month.

AGENTS

Google Ads API gets an AI update

From tipsheet: Google is making its advertising infrastructure legible to outside AI agents. Today that’s reporting, diagnostics and development and tomorrow the architecture points toward a world where the agent becomes the interface and Google Ads becomes infrastructure underneath it.

LLMs & CHATBOTS

Developments

  • Gemini Omni 1.1 Flash lets you build (video) with more control (August 27) – Google
  • The Hugging Face incident and the road ahead (August 26) – OpenAI
  • Nvidia Agrees to Buy Hugging Face For $12.9 Billion (August 27) – The Information (subscription)

LLMs & CHATBOTS

Opinion: The lesson of Hugging Face for ads

Noting Nvidia’s reported agreement to acquire Hugging Face, Frank Sinton, CEO of Adgentek, wrote on LinkedIn:

“Hugging Face is where models get found, versioned, benchmarked, and trusted. That layer turned out to be worth more than most of the models sitting on it. Discovery and provenance are what make an ecosystem usable.

Advertising is about to learn the same lesson.

As agents take on more of the planning and buying, the scarce thing is not another model or another optimization algorithm. It is the layer where agents discover inventory, evaluate it, and transact against it without a custom integration for every counterparty.”

Read more. (August 27)

More opinions:

  • “Creator Effectiveness & Digital Advertising’s Original Sin” (August 27) – Andrew Lipsman on Media, Ads + Commerce Substack (subscription)
  • “Independent Ad Measurement (Outside Of TV) Is Dead” – Dave Morgan, CEO, Simulmedia on MediaPost

PEOPLE MOVES

HR machinations

Coming

  • “I couldn’t be more excited to partner with Garland Hill and Mark Douglas as we build out our agency ecosystem and make Performance TV as accessible as search and social.” (August 27) – Former Snap executive Frank Lee joins MNTN as SVP of Agency Partnerships. Read on LinkedIn
  • “Very happy to (finally) announce that I’ve joined Shopsense AI as their Chief Product Officer.” (August 27) – Former Amazon Marketing Cloud executive Alan Lewis on LinkedIn

Going

  • “Today marks the close of an incredible six-year chapter at Criteo… There is tremendous opportunity still ahead, and I’ll remain an advisor through the end of September to support a smooth transition… I’ll share more about my next chapter soon…” (August 27) – Todd Parsons, President and Chief Product Officer, Criteo on LinkedIn

MORE

  • India gets ChatGPT ads: “An exciting new chapter for ChatGPT and for businesses in India. Today, we are bringing ChatGPT Ads to India, creating a new way for businesses to be discovered when people are actively exploring, comparing and making decisions.” (August 27) – Nitin Bawankule, Head of Enterprise, OpenAI on LinkedIn
  • AI licensing: USA Today Co. is reformatting content to attract more AI licensing deals (August 26) – Digiday
  • RMNs: The Largest Shopping Mall Operator Has Its Own Retail Media Network (August 27) – AdExchanger
  • TikTok expands into DOOH: TikTok is expanding its “Out of Phone” program across Europe, bringing TikTok content and ads to retail, cinema and other out-of-home screens (August 27) – MediaPost