Mythbuster: What AI is not about to do in advertising
Briefly

Mythbuster: What AI is not about to do in advertising
"Myth: AI is about to replace media buyers AI is not replacing buyers. It is replacing their admin work. The real deployments that are scaling are not autonomous trading agents. They are tools that eliminate hours of setup, normalization and reporting labor, turning two-hour campaign builds into ten-minute workflows with zero errors. As a result, buyers are becoming supervisors, not casualties. "I'd say 99% of the conversation that you hear at the moment is around the workflow piece - the things you can see and feel," said Michael Richardson, vp of product at Index Exchange."
"Myth: LLMs will run the bidder Live bidding is a control problem, not a reasoning one. The engines that clear auctions are built to be fast, deterministic and cheap at massive scale. LLMs are probabilistic, slow and expensive by comparison, which makes them well suited to orchestration and interfaces, but a poor fit for the bidstream itself. The future is not generative models replacing bidders. It's generative layers wrapped around them."
"Myth: The agentic future is here. What's really being built is an RTB operating system Index Exchange, Yahoo DSP and others are modernizing the pipes - containerizing auctions, standardizing APIs and lowering switching costs - so the open web can someday support advanced automation. That is infrastructure work, not sentient buying. The future is being enabled, not activated. "I really do feel like we're redefining what it means to be a DSP," said Adam Roodman, GM of Yahoo DSP on the current trajectory his business is on. "For us, that's partly because of what technology has become available to us with agentic.""
AI is automating administrative and workflow tasks for media buyers, converting hours of campaign setup, normalization and reporting into minutes with fewer errors. Deployments scale as tools that eliminate manual labor rather than autonomous trading agents. Live bidding remains a control and speed problem; probabilistic LLMs are ill-suited for the bidstream but useful for orchestration and interfaces. Programmatic infrastructure is being modernized through containerized auctions, standardized APIs and lower switching costs to enable advanced automation on the open web. Brand investment remains a defensible moat as programmatic and AI capabilities evolve.
Read at Digiday
Unable to calculate read time
[
|
]