The Trade Desk and Dstillery link AI agents to mid-campaign audience changes

2 hours ago

6 min read

The Trade Desk and Dstillery have connected their AI systems so advertisers can build and activate custom audience segments during a live campaign, rather than waiting days before launch. The integration links The Trade Desk’s Ask Koa agentic platform with Dstillery’s DS-1 platform, enabling the two systems to exchange information and create new audiences based on campaign performance and the advertiser’s targeting goals.

The Trade Desk and Dstillery link AI agents to mid-campaign audience changes

The announcement, reported by Adweek on November 4, 2026, is significant because it moves audience targeting from a largely pre-launch planning task toward an ongoing, in-flight workflow. It does not mean that AI is fully replacing media buyers; rather, it automates a specific and often slow part of the programmatic process: identifying, building, and activating a revised audience after a campaign has already begun.

What the integration does

Under the integration, Ask Koa and DS-1 can communicate while a campaign is running. They use performance insights and the advertiser’s stated targeting objectives to identify potential new audiences, build custom segments, and make those segments available for activation through The Trade Desk.

In practical terms, a campaign that initially targets one audience definition may reveal that another group is responding more effectively. Previously, a media team would need to review performance, coordinate with Dstillery, request or refine a custom audience, and then activate it in The Trade Desk. The new connection is designed to reduce that manual coordination and shorten the time between identifying an opportunity and acting on it.

The companies describe the result as custom audiences being built in minutes rather than days. Dstillery’s chief product officer, Taejin “TJ” In, said that advertisers could previously buy Dstillery custom audiences through The Trade Desk, but building them required manual coordination and typically took four days.

Why mid-campaign matters

Programmatic campaigns are often planned with audience segments selected before launch. Once a campaign is live, buyers can adjust bids, budgets, creative, frequency, and other settings, but changing the underlying audience can be more complicated because it may require data analysis, audience modeling, approval, and technical activation.

That delay matters because campaign performance can change quickly. An audience that looked promising during planning may underperform once it encounters real inventory, real creative, and real market conditions. Conversely, performance data may reveal an overlooked group that is more likely to convert, engage, or meet another campaign objective.

The Trade Desk and Dstillery’s approach is aimed at that gap. Instead of treating audience construction as a one-time setup decision, the integration makes it possible for agents to propose and build revised audiences while the campaign is still running. The Trade Desk described this as agents “talk[ing] to each other in the middle of a campaign,” rather than building custom audiences only before launch.

The role of Ask Koa and DS-1

Ask Koa is The Trade Desk’s conversational agentic platform. Its role in this integration is to connect campaign context—such as performance insights and targeting goals—with Dstillery’s audience-building capabilities.

DS-1 is Dstillery’s agentic platform for audience work. Dstillery positions it as a way to streamline audience discovery, custom modeling, and activation, tasks that historically required manual work across platforms.

The division of labor is therefore relatively clear: The Trade Desk provides the buying platform and campaign context, while Dstillery contributes its audience-intelligence and custom-audience modeling capability. The integration is not simply a data feed; it is a connection between two AI systems that can collaborate during an active campaign.

What changes for advertisers

For advertisers and agencies, the most immediate change is speed. If a campaign is underperforming against a particular audience or if performance suggests a better segment, the team may be able to create and activate a new custom audience much faster than before.

That could be especially useful for campaigns with limited flight windows, such as product launches, seasonal promotions, political advertising, or time-sensitive retail campaigns. In those cases, waiting several days to rebuild an audience can mean missing much of the period in which the adjustment could improve results.

The integration may also reduce the operational burden on media teams. Rather than managing email chains, manual data requests, and multiple platform handoffs, buyers may be able to use an agent-driven workflow to translate performance observations into a new audience segment. Dstillery has previously described DS-1 as intended to remove platform-hopping, manual handoffs, and waiting times from audience workflows.

However, the announcement does not establish that every advertiser will automatically benefit. The value depends on the quality of the underlying performance data, the advertiser’s objectives, the availability of relevant signals, and the effectiveness of the audiences that the system creates.

Important limitations

The companies’ claims concern speed and workflow automation, not guaranteed performance improvements. The public reporting says the agents can build custom audiences in minutes and use performance insights and targeting goals to inform those audiences; it does not provide a verified, campaign-level result showing that this integration improves return on ad spend, conversions, or sales.

Adweek also reported that some ad buyers see the potential time savings but remain skeptical about whether the integration will lead to higher ad spend. That skepticism is important because faster audience creation is valuable only if the resulting audiences are genuinely better aligned with campaign goals.

There is also a governance issue. Automated audience creation can make campaign operations faster, but advertisers still need clear rules about what data may be used, what audience definitions are acceptable, how brand-safety and privacy requirements are handled, and who approves material changes to targeting. The public information available does not detail those controls for this specific integration.

Broader industry context

The announcement fits a wider shift in advertising technology toward “agentic” AI: software systems that can perform multi-step tasks, communicate with other systems, and execute parts of a workflow with limited human intervention. In this case, the task is audience construction and activation—not bidding, creative production, or full campaign management.

The Trade Desk has been positioning AI as a way to reduce manual work in media buying. Dstillery, meanwhile, has been developing DS-1 as an agentic interface for refining audiences more quickly. In March 2026, Dstillery announced a separate partnership with Keynes and The Trade Desk involving DS-1, custom audience building, and activation through The Trade Desk.

The new November integration extends that direction by making the agents capable of communicating during live campaigns. The distinction is meaningful: earlier agentic workflows could make audience creation faster before launch, while this connection is specifically framed around mid-campaign audience changes.

What it may mean for media buyers

For experienced programmatic traders, the technology could shift time away from repetitive coordination and toward higher-level decisions: defining success metrics, reviewing audience logic, setting guardrails, and evaluating whether an AI-generated segment actually improves outcomes.

For smaller teams, it could make sophisticated custom-audience tactics more accessible by reducing the operational friction involved in building them. For larger advertisers, it could enable more frequent testing and refinement, provided that teams have the measurement discipline to distinguish genuine performance gains from random variation.

The key question is not whether an audience can be built in minutes. It is whether the audience built in minutes is better than the one a skilled media team would have built with more time, more context, and more oversight. The announcement makes the speed advantage clear; the performance advantage remains something advertisers will need to test in their own campaigns.

Bottom line

The Trade Desk and Dstillery have created a practical example of agentic AI in programmatic advertising: two AI systems working together to revise audience targeting while a campaign is live. The integration could reduce a four-day manual process to minutes and make mid-flight audience optimization more responsive.

But it should be understood as an automation of audience workflow, not proof that AI can autonomously deliver better advertising outcomes. Its real value will depend on whether advertisers can validate that the new audiences improve campaign performance without compromising targeting quality, privacy, brand safety, or human control.