The Trade Desk and Dstillery Link AI Agents for Mid-Campaign Audiences

Adweek reported on 4 November that The Trade Desk and Dstillery have connected AI agents so that audience segments can be created and reshaped while a campaign is live, rather than only in the planning phase. The two companies present the integration as a way to shorten the loop between audience definition and delivery.

The report adds a counterweight: buyers interviewed see the appeal of faster audience creation, but they question whether agentic integrations will meaningfully change media buying. Their argument is that the larger value still sits in knowing which audiences are right for a brand. The announcement does not specify which agent capabilities run on either side of the integration, what data sources feed the audience construction, or how pricing and access are structured.

What actually changes in the workflow

Most buying teams already work with a static picture: segments are defined up front, launched, then reviewed on a weekly or monthly cadence. An agent layer that operates during flight compresses that cycle. In practice, that means the operational question shifts from "how do we build the segment" to "what rules govern how it changes, and who approves a change".

That is a governance problem more than a technology one. A system that edits audiences mid-campaign can quietly move spend toward whatever signal it optimises for. If the target is a cheap click or a broad reach metric, faster iteration simply produces faster drift away from the audience the brand actually wanted. Nothing in the report indicates how the partners handle approval, audit trails or rollback.

What a team can check this week

Before adding any agent-driven audience layer, the useful work is internal and cheap.

• Confirm that your audience definitions are documented well enough that an external system could reproduce them, including exclusions and frequency caps.
• Agree on one or two outcome metrics that any automated audience change must not degrade, and decide who has authority to pause it.
• Check whether your reporting can attribute results to a segment that changed mid-flight, or whether changes collapse into a single campaign line.

If the answer to the third point is no, faster audience creation will mostly generate numbers nobody can interpret.

The relevance question stays open

The buyers' scepticism in the Adweek report is worth taking seriously. Speed of audience construction is a supply-side improvement: it reduces the time between an idea and its execution. It does not by itself tell you whether the idea was good. The expensive failure mode in paid media is rarely slow setup; it is well-executed targeting of a segment that was never commercially meaningful.

Agentic tooling tends to be sold on velocity. For founders and marketing leads, the test is whether it improves decisions, not throughput. A reasonable position for now: treat mid-campaign audience automation as an experiment with a defined guardrail metric, run it alongside a stable control, and require evidence that the automated changes beat a fixed segment before expanding scope. The announcement leaves the commercial and measurement details unspecified, which is itself a reason to start small.