Retail media growth has been built on the bedrock premise of influencing the customer before the checkout button is pressed.
Agentic commerce raises some questions for that premise. If software increasingly chooses products, compares alternatives and completes purchases, retailers may need new ways to shape outcomes and monetize attention.
An April PYMNTS Intelligence report, “The Intelligent Spend Shift: How Card Platforms Can Prepare for Agentic Commerce,” found that commerce is moving beyond recommendation engines toward systems capable of initiating transactions on behalf of consumers. Payments infrastructure, authorization and commerce experiences are increasingly designed to support machine-driven transactions at scale.
Consumers are already showing interest in these experiences. According to the report, 48% of consumers said they are at least somewhat interested in using AI agents to buy groceries or plan meals. Consumers also expressed interest in subscription management and gift purchasing.
When Shelf Placement Becomes Recommendation Logic
Retail media networks have historically depended on human attention. Sponsored listings, search placement and promotions are designed to persuade and encourage discovery.
Chris Selland, founder of Differential Factor and lecturer at Northeastern University’s D’Amore-McKim School of Business, told PYMNTS that agentic commerce changes where that competition occurs.
Advertisement: Scroll to Continue
“AEO [AI engine optimization] is the new battleground for digital shelf space,” Selland said, describing a future in which product selection increasingly happens through API calls rather than browsing behavior. In that environment, structured product data, fulfillment information, trust signals and availability become inputs that determine whether an AI agent includes a retailer or product in its recommendation process.
That logic carries implications for measurement. Retail media networks today emphasize impressions, clicks and conversion rates tied to user engagement. But if agents reduce time spent browsing, those indicators may become less useful proxies for influence.
Selland said retailers will increasingly need to understand whether incentives affect agent decisions rather than human ones and track how often automated purchasing systems alter outcomes based on pricing, inventory or merchant preferences.
Promotions themselves may also change.
Traditional retail promotions often blend urgency, placement and visual presentation. Agent-driven commerce may reward different signals. Software systems evaluating transactions in real time may prioritize price, delivery performance and explicit user instructions more heavily than visual merchandising.
Promotions will not disappear but could become increasingly programmatic, serving as mechanisms that influence preference settings and transaction logic rather than shopper attention in the moment, Selland said.
“The ability to rank at the top of an AI agent’s consideration set is everything,” he said. “Measuring how often media networks’ incentives convince an agent to switch is critical.”
Examining Infrastructure
As purchasing becomes increasingly automated, infrastructure starts to matter as much as merchandising.
Existing payments systems were largely designed for human-initiated transactions and identified APIs, cloud-based architectures, tokenization and programmable controls as foundational capabilities for agentic commerce.
Retailers hoping to keep AI transactions inside their own ecosystems may therefore need to make execution easier, not just discovery.
Selland pointed to secure wallet authentication, native transaction execution through APIs and loyalty environments that allow consumers’ agents to exchange preference data for benefits and incentives.
At the same time, governance questions are emerging.
Alpesh Patel, strategic partnership director at Cartex, told PYMNTS that AI commerce introduces the need for know your agent approaches that go beyond traditional identity verification.
Retailers increasingly need to establish who authorized the agent, what authority was granted and who carries responsibility if automated decisions create disputes, Patel said.
“Chargeback infrastructure also needs re-engineering,” he said. “Agentic commerce introduces a new layer to the equation. Was the agent acting within its delegated scope or not? Current scheme rules, chargeback reason codes and merchant agreements are not equipped to handle this distinction.
“Until liability within this new landscape is legally defined, merchants are left facing asymmetric risk,” he added. “They bear the operational costs of failed agent transactions and the chargeback exposure from disputed ones, with no clear way to defend their position.”
Retail media networks are unlikely to disappear. But if purchasing becomes increasingly delegated, influence would increasingly belong to the retailers and platforms that can become machine-readable, machine-trusted and transaction-ready before the consumer ever arrives.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
