The Trade Desk Launches Kokai Zuma AI Tools; Stock Remains Down 65%

The Trade Desk (the largest independent demand-side platform) shipped Kokai Zuma this week: a conversational assistant called Ask Koa, 48-hour lift studies replacing the previous week-long turnaround, and an opening of its AI layer to outside tools. The company is claiming 32% average cost-per-acquisition improvements from early tests. The stock is still down about 65% for the year.

The feature list is less interesting than what disappeared from TTD's disclosures. The company used to recite platform adoption figures on every earnings call as the core of its bull case. That number is now gone. Companies drop a metric when it embarrasses them. The silence is the story.

The 32% CPA gain deserves skepticism. These are campaigns TTD selected, measured with attribution TTD controls. Averages bury the losers. Every major DSP has shipped a chat interface since ChatGPT. None has moved real spend. Ask Koa is a chat skin on the same bid logic underneath, so the workflow change for traders is real but thin this quarter.

The 48-hour lift study is the piece with teeth. If it holds at scale, buyers can retire some external measurement vendor contracts. That's a budget line you can actually feel. But faster in-house measurement from the platform being measured raises an obvious problem: if the referee owns the scoreboard, serious buyers will pay someone else to keep time. That's good for DoubleVerify and Integral Ad Science (independent ad verification firms), and for VideoAmp and iSpot (cross-platform measurement firms), regardless of which way the API story goes.

The buried decision is opening the AI layer to third parties. If identity, measurement, and creative vendors build on Kokai, TTD becomes the thing everyone else plugs into. If vendors route around it, this is feature parity against Google's DV360 and Amazon's DSP, both of which sit on deeper first-party data than TTD will have. The third-party integrations are what settle that question. Watch whether DoubleVerify, IAS, VideoAmp, or iSpot announce Kokai integrations by mid-2026.

Our call: TTD will not restore its Kokai platform-adoption metric with a positive figure on either the Q3 or Q4 2026 earnings call, and the stock will still be down more than 40% from its January 2026 level at the Q4 print. A single feature launch, measured by the vendor on campaigns it selected, does not reverse what the missing metric already signals. The stock moves on the adoption number returning, and the less likely reason a company buries a metric is that it quietly got better.

The Trade Desk shipped Kokai Zuma this week: a conversational assistant called Ask Koa, one-click lift studies that return in 48 hours instead of a week, and an opening of its AI layer to outside tools. The company waved a 32% average cut in cost-per-acquisition from early tests. The stock is still down about 65% for the year. And somewhere along the way, the platform adoption number TTD used to recite on every earnings call stopped showing up in its disclosures.

That last part matters more than the feature list. So let's frame what's actually on the table for operators.

This is a briefing, not a decision you have to make. The question for a publisher, agency, DSP, SSP, or measurement leader is simple: does Kokai Zuma change what you build toward, who you renew with, or how you pitch? Nothing here is hard to undo. You're not signing anything today. But the thing being decided quietly, across the whole ecosystem, is whether "AI tooling" is now table stakes in a DSP pitch or a real moat. What sets the clock is the next earnings call, when either the adoption metric comes back with a number or it stays gone.

The Skeptic. A vendor grading its own homework during a 65% stock collapse gets read as investor relations first, product truth second. The 32% CPA gain came from campaigns TTD picked, measured with attribution TTD controls. Averages bury the losers. And every major DSP has announced a chat interface since ChatGPT landed. None has moved real spend. The missing adoption number is the part that should bother you. TTD built its bull case on Kokai penetration, and now that figure isn't in the filings. In plain terms: when a company stops reciting the number it used to brag about, the silence is telling you the number got worse.

The Market Analyst. At down 65%, the market has already written off a lot of execution. So Kokai Zuma is a test of whether anything can turn sentiment. It won't, on its own. Announcements don't re-rate a stock this beaten up. A real number does. Who actually gains here: DoubleVerify and Integral Ad Science, if TTD's open API pulls their measurement into the workflow. VideoAmp and iSpot, if 48-hour lift studies become a commodity and buyers demand the same read across platforms. Who gets squeezed: any smaller DSP without an AI story, because TTD just raised the price of admission on what a DSP deck has to contain. For a general reader: the news moves TTD's neighbors more than it moves TTD.

The Strategist. The buried decision is opening the AI layer to third parties. That's the one that could matter in three years. If identity, measurement, and creative vendors build on Kokai, TTD stops being just a buying tool and becomes the thing everyone else plugs into. The Reuters terminal of programmatic. If they build around it instead, this is feature parity against Google's DV360 and Amazon's DSP, both of which sit on deeper first-party data than TTD will ever have. Plainly: TTD is trying to become the plumbing, and the proof will be whether outside vendors treat it as a foundation or a competitor.

The Operator. Tuesday morning, if you run a TTD desk, you test the 32% against your own verticals before you repeat it to a client. Averages hide desks that got worse. Ask Koa is a chat skin on the same bid logic underneath, so the workflow change for traders is real but thin this quarter. The 48-hour lift study is the piece with teeth. If it holds at scale, your measurement ops team can retire vendor contracts, which is a budget line you can actually feel. Watch for API partners shipping integrations by Q2 2026. That's when the platform story gets proof or gets quietly shelved. For a non-specialist: the flashy feature is the chatbot, the useful one is faster measurement.

The Customer / End User. Advertisers didn't ask for a chatbot. They asked for cheaper outcomes and faster answers on whether the money worked. The 48-hour lift study speaks to the second want. Ask Koa speaks to a want nobody expressed. And here's the friction: the more TTD controls the measurement of TTD's own performance, the more a serious buyer wants an independent read next to it. Faster in-house numbers raise the value of cross-platform verification. Plainly: if the referee also owns the scoreboard, buyers will pay someone else to keep time.

Where the council splits. Two real disagreements. The Strategist sees a platform being born; the Skeptic sees a wrapper on 2022's moat, now unwinding. Both can't be right, and the third-party integrations are what settle it. Second, the Operator thinks the lift study genuinely lets buyers fire measurement vendors, while the Analyst and the Customer think faster in-house numbers make buyers want an outside check more urgently. That one splits on trust: do buyers accept a seller's grade, or demand a second opinion?

What it hinges on. Three things. Does the 32% survive contact with real, varied campaigns. Do outside vendors build on Kokai or route around it. And does the adoption number come back. The council leans skeptical on the stock and the moat, and genuinely interested in the open API as the one move that could still matter. The near-term winners are the independent measurement names, because either the API pulls them in or buyers demand them as a check on TTD's own scoring.

What to verify before you act on any of this: pressure-test the CPA claim in your own accounts, and watch whether DoubleVerify, IAS, VideoAmp, or iSpot announce Kokai integrations by mid-2026. That's the corroboration or the collapse.

Prediction: The Trade Desk will not restore its Kokai platform-adoption metric with a positive figure on either its Q3 2026 or Q4 2026 earnings call, and the stock will still be down more than 40% from its January 2026 level at the Q4 2026 print.

Confidence: Medium — a company that drops its headline number mid-rollout rarely brings it back up.

Why: TTD recited platform adoption every earnings call as the core of its bull case, then stopped disclosing it exactly as the stock fell 65%. Companies remove a metric when the trend embarrasses them, not when it flatters them, so the removal itself signals the number turned against the story. A single feature launch, measured by the vendor on campaigns it selected, does not reverse that, and a stock this beaten up moves on a real adoption number returning, not on a chatbot. The opposite outcome, adoption returning strong and the stock recovering fast, would require the very metric they buried to have quietly gotten better, which is the less likely reason a company hides it.

Revisit by 2027-03-01: We're right if, through the Q4 2026 earnings call, TTD either omits the platform adoption metric or reports it flat-to-down, and the stock remains more than 40% below its January 2026 level. We're wrong if TTD reintroduces the adoption metric with year-over-year growth and the stock recovers to within 40% of its January 2026 level.