Amazon DSP puts AI agents at the centre of buying, creative and targeting: Ankur Rathi

Speaking at the e4m Real Time Programmatic Advertising Conference 2026, Ankur Rathi, Head, Amazon DSP, Amazon Ads, said AI tools aim to solve problems marketers have battled for years: wastage, efficiency, creative and measurement, now compounded by brand safety, security, and choice. His session was titled "Amazon DSP: AI-led new frontiers".

Before AI, Rathi said, came the groundwork. Amazon has unified its campaign manager console so that search, display and video campaigns sit in a single view. He said the redesign brings "visibility, speed, and smarter decisioning in every action," replacing workflows that earlier needed multiple windows and interfaces.

The centrepiece was Amazon Ads Agent, a conversational buying and campaign management capability that draws on insights and signals to make recommendations, allowing advertisers to spend more time deciding how to deploy budgets. On creative, it uses generative AI to handle ideation, design and finished display and video ads. It draws on product store insights, and for advertisers who are not Amazon sellers, on links or literature they share with the agent.

On targeting, Rathi said customer feedback showed that understanding the right audiences on Amazon DSP usually takes time. A conversational targeting agent, built on category and retail insights, is meant to shorten the time taken and counter cognitive biases. "It's going to give you a fresh perspective on what you may have missed in your earlier campaigns or you may have over-indexed on," he said.

Performance+, which he introduced at the event last year and which is now used by hundreds of brands in India, gets two new tactics. Search Remarketing captures the search intent of shoppers who looked for a product or category but did not buy, and builds real-time audiences to serve on display and video. A leading PC brand that deployed it at launch saw a 2x jump in ROAS against standard display campaigns, despite already earning strong returns from search.

Ad Exposure Remarketing creates real-time audiences from people exposed to a brand's communication during a campaign, for sequential messaging or display-led conversion. On measurement, Persona Insights shows which cohorts are outperforming others. In the example shown, luxury shoppers had a higher purchase rate, allowing advertisers to edit a campaign mid-flight or start a new line with more budget on that cohort.

Geographic Insights and Activation, or GIA, generates a heat map from a brand's ASIN data, showing where it is most and least penetrated, indexed against the category. Using penetration bands of 30, 70 and 100, advertisers can decide whether to acquire new customers, defend existing territory or enter regions where the category outpaces them. It is built on Amazon's first-party purchase and detail page view insights, and advertisers can add their own first-party and offline data through Amazon Marketing Cloud.

In a case study, Purina saw new-to-brand customers rise significantly while the cost of acquiring them fell.