The Trade Desk's Kokai Zuma Makes Agentic Programmatic Execution a Standard Feature
Executive Summary
The Trade Desk's Kokai Zuma release embeds a conversational AI agent — Koa Assistant — into the full programmatic workflow, delivering a 32% average CPA improvement and automating campaign creation, audience discovery, and optimisation execution at scale. The tactical expertise that agency trading desks have monetised for a decade is now a software feature. CMOs with programmatic spend above $10M should run a contained Zuma pilot on one always-on campaign before October 2026 — without waiting for agency sign-off.
The Signal
The Trade Desk launched Kokai Zuma on August 28, 2026 — a major platform release that embeds agentic AI directly into its programmatic buying workflow. The centrepiece is Koa Assistant, a conversational AI backed by specialised Koa Agents capable of executing campaign creation, audience building, troubleshooting, and performance analysis autonomously. Koa already processes over 20 million ad impressions per second; Zuma upgrades the forecasting engine sitting beneath it, delivering an average 32% improvement in CPA performance in initial results. New capabilities include AI-driven frequency optimisation, automated audience discovery across 70% of global data providers in The Trade Desk's Data Marketplace, streamlined lift study creation, and bulk editing with pre-apply impact previews. Zuma is rolling out globally to Kokai clients immediately.
What Changed
Media buyers on The Trade Desk can now instruct an AI agent in plain language to build campaigns, identify audiences, diagnose underperformance, and execute optimisation changes — without navigating manual workflows. Tasks that previously required trained traders moving across multiple platform surfaces can now be initiated conversationally and automated end-to-end. The forecasting engine upgrades mean that autonomous decisions are made on materially better inventory and outcome models than were available before August 28, 2026.
Why It Matters
The sharpest implication here is structural: The Trade Desk has just commoditised the execution layer of programmatic trading. When a conversational agent can build campaigns, discover audiences, diagnose performance issues, and push optimisation changes without a trained trader touching a workflow, the value of tactical media expertise collapses. What was billable hours becomes automated overhead. For brand-side marketing teams, this is unambiguously positive. Leaner teams can run more sophisticated programmatic campaigns across more markets, more inventory types, and more audience segments simultaneously. The 32% CPA improvement is not a marginal gain — at scale, that arithmetic changes budget allocation decisions materially. A brand spending $50M annually in programmatic that captures even half that improvement is looking at a different unit economics conversation with their CFO. The pressure it creates for agency trading desks is more uncomfortable. Mid-tier programmatic specialists — the traders whose expertise was their value proposition — now face direct substitution risk. The capability gap between a junior operator using Koa Assistant and a seasoned trader running manual workflows narrows fast. Agencies that have built margin on the complexity of programmatic execution are watching that complexity dissolve. The deeper strategic logic is The Trade Desk's positioning against the walled gardens. Google and Meta have long had closed-loop AI optimisation that brands couldn't audit or interrogate. Zuma pushes autonomous, AI-driven performance optimisation into the open internet with a transparent, conversational interface — giving sophisticated buyers a reason to shift budget away from Google Display Network and Meta Audience Network and toward premium publisher inventory they can actually reason about. That is the real competitive move buried inside the product release.
Marketing Impact
media
Programmatic campaign creation, audience discovery, and optimisation are now conversationally executable. Traders who previously managed these workflows manually are partially substituted. Media teams can run more campaigns across more markets with fewer specialists, while the 32% CPA improvement materially shifts budget efficiency conversations at the CFO level.
marketing ops
Campaign troubleshooting and performance analysis — historically bottlenecks requiring trained platform operators — become agent-executable. Marketing ops teams can redirect headcount from workflow management toward strategic oversight, and bulk editing with pre-apply impact previews substantially reduces the risk cost of large-scale campaign changes.
martech
The Trade Desk's agentic layer reduces dependency on third-party optimisation tools and managed service overlays that existed to compensate for platform complexity. Martech stacks built around programmatic workflow tooling face displacement as Koa Assistant absorbs functions that previously justified separate vendor relationships.
The Exploit
Opportunity
Brand-side programmatic teams can now run Koa Assistant against live campaigns immediately, capturing the 32% CPA improvement on current spend while cutting trader headcount or redeploying it to strategy. A $20M+ programmatic budget reallocating even half the efficiency gain unlocks seven-figure savings or equivalent reinvestment capacity within a single quarter.
Risk
Agencies managing programmatic on behalf of these brands will resist adoption to protect billable complexity. Rushing Zuma deployment without updated campaign taxonomies and measurement frameworks risks polluting the forecasting engine with legacy structure.
The Move
CMOs with programmatic spend above $10M should instruct their media leads, not their agencies, to run a contained Zuma pilot on one always-on campaign before October 2026, with CPA delta versus control as the sole checkpoint. Agency sign-off should not be required or requested.
First-Mover Advantage
Gains
Brands that rebuild audience and campaign architecture around Koa's automated discovery will compound performance gains across Q4 2026 — the highest-spend quarter — while competitors are still evaluating the platform.
Risks
Agencies managing programmatic on behalf of these brands will resist adoption to protect billable complexity. Rushing Zuma deployment without updated campaign taxonomies and measurement frameworks risks polluting the forecasting engine with legacy structure.
Window
The advantage window is approximately two to three quarters — until mid-2027, when competitors have completed migration and the CPA delta normalises. The closing signal is when agency holding companies productise Zuma-based services at scale.
Winners & Losers
Winners↑
In-house programmatic teams at mid-to-large brands
Koa Assistant removes the skills barrier that historically made sophisticated programmatic execution agency-dependent. Brand-side teams that previously lacked the trader headcount to run multi-market, multi-format campaigns simultaneously can now do so through conversational instruction — with a 32% CPA improvement baked into the forecasting engine underneath. The immediate move is to audit which campaigns are still being routed through agency desks for execution complexity reasons rather than genuine strategic value, and begin a structured insourcing assessment.
Premium open-web publishers with direct Trade Desk supply relationships
Zuma's agentic audience discovery — spanning over 70% of global data providers — runs on The Trade Desk's open internet inventory stack, not Google Display Network or Meta Audience Network. As autonomous optimisation improves performance on premium publisher supply, budget gravity shifts toward addressable open-web inventory that buyers can actually interrogate and contextualise. Publishers with strong Trade Desk integrations and clean first-party data signals are structurally better positioned to capture incremental programmatic spend as brand-side teams reallocate from walled garden display.
Performance marketing teams managing high-volume, multi-market campaigns
The combination of autonomous frequency optimisation, bulk editing with pre-apply impact previews, and AI-driven audience expansion compresses the operational overhead of running complex campaigns at scale. Teams managing dozens of concurrent campaigns across markets gain execution bandwidth without proportional headcount growth. The exploit is straightforward: redirect trader time from workflow management to strategy and creative testing, where human judgment still adds differentiated value.
Agencies with agentic programmatic capability built around The Trade Desk
Not all agencies lose here — those that have already reorganised around strategic media planning, measurement design, and AI-supervised execution rather than manual trading workflows are structurally advantaged. Zuma accelerates the bifurcation between agencies that sell expertise and those that sold labour, and the former now have a more powerful execution infrastructure to offer clients. The opportunity is to repackage The Trade Desk relationships as a strategic differentiator rather than a commoditised execution service.
Losers↓
Agency trading desks built on programmatic execution complexity
The billable-hours model for programmatic trading was already under pressure; Koa Assistant effectively automates the task stack that justified specialist trader headcount — campaign creation, audience building, performance diagnosis, and optimisation execution. When a conversational agent can replicate that workflow and deliver materially better CPA outcomes, the expertise premium collapses. The defensive play is rapid repositioning toward measurement strategy, media mix modelling, and cross-channel planning — capabilities where automation is a tool rather than a replacement — but the window to make that shift without client attrition is narrow.
Mid-tier programmatic specialists and independent trading consultants
The specific skill set most threatened by Zuma is the tactical trader who built their value proposition on platform proficiency — knowing where to find the right audiences, how to structure campaign hierarchies, and when to intervene on underperforming line items. Koa Assistant now performs those functions conversationally and with better forecasting models than most traders have access to. Individuals in this category face direct substitution pressure and should be retraining toward data strategy, identity architecture, and measurement design rather than execution optimisation.
Strategic Outlook
The programmatic trading desk as a staffing model is entering structural contraction. Kokai Zuma is not the endpoint — it is the first fully productised agentic release from a major DSP, and Google's Display & Video 360 and Xandr will respond within two to three quarters with equivalent or competing capability. The race is now about which platform's AI model produces better outcome predictions at scale, not which has the better UI. The Trade Desk's 20-million-impression-per-second inference infrastructure gives it a genuine data moat, but that moat is only durable if it retains inventory breadth. Expect accelerated consolidation among mid-tier trading desks through Q4 2026 into 2027, as the complexity premium that justified agency margin evaporates. Brands with the contractual flexibility to shift programmatic execution in-house will begin doing so; the agencies that survive will be those repositioning around strategy, creative, and measurement interpretation rather than execution.