AI in advertising has largely focused on improving targeting and optimisation, while the operational side of media buying has remained heavily manual. Agentic AI changes that, especially in CTV and DOOH, by autonomously managing the complexity of deals, delivery, and yield optimisation at a scale humans simply can’t match.
Most of the AI conversation in advertising has been about targeting.
Better audience segments, smarter bidding, faster creative testing. Useful improvements. But none of it changes the underlying architecture. A human still manages the deals, monitors delivery, and makes the calls.
Agentic AI is different and CTV is likely where it lands first.
Why CTV is the right starting point
CTV deal management is genuinely complicated.
Custom pricing floors, guaranteed delivery commitments, audience specifications, daypart constraints, brand safety rules, frequency caps across publishers. Every campaign involves a web of conditions that a human account manager currently has to encode, watch, and adjust manually. It is slow, it does not scale, and small errors compound across a full book of business.
Agentic AI changes that structure.
Instead of a person interpreting a deal and executing it, you have a system that holds the publisher's rules in full and executes continuously, adjusting pacing in real time, flagging exceptions, optimising yield across the entire inventory set, without waiting for a human handoff. The work that used to require constant attention happens automatically, in the background, around the clock.
The same logic applies to DOOH
DOOH has its own version of this problem.
Venue rules, audience dwell time, creative scheduling, weather triggers - all of it is currently managed manually or through rigid DSP logic that has to be set up in advance. Agentic systems can handle it dynamically, adjusting to real-world conditions rather than pre-configured instructions.
The principle is the same across both channels: wherever there are complex, layered rules governing delivery, agentic systems can execute them faster, more accurately, and without the operational ceiling that limits human-managed workflows.
What this means commercially
The publishers who move early will not just save time. They will build a compounding yield advantage that is difficult to close later.
That is worth being precise about. This is not an efficiency story in the sense of doing the same thing faster. It is a different business model. Yield optimisation that runs continuously, never misses a pacing signal, and manages complexity at a scale no operations team could match, that is a structural advantage, not a marginal one.
For CTV and DOOH specifically, the complexity that has historically made programmatic harder to scale becomes a source of value once agentic systems can manage it well. The nuance of the category, which was a friction point, becomes a moat.
The question for publishers is not whether to think about this. It is how far behind they can afford to be.