Agentic might be changing the game, but data still determines who wins
Opinion
AI agents have already infiltrated retail, but how they inform advertising agents could take retail media to a whole new level.
Shoppers are beginning to enjoy more immersive, agent-powered commerce journeys.
Instacart’s agent is a clear signal of where things are heading. Customers can type in natural language prompts such as what’s a good dessert for summer barbecues? or show me high-protein breakfasts under 500 calories. The system responds with curated recipes and product suggestions pulled directly from local grocers’ shelves, turning vague intent into ready-to-purchase baskets.
Walmart is also moving at pace. Its Sparky assistant already helps customers with product recommendations, budget planning and even weather checks, and soon it will go further, automatically reordering everyday essentials so shoppers never run out.
Mastercard’s Agent Pay initiative and Visa’s Intelligent Commerce program illustrate how quickly this transformation is taking shape. The rails for agent-led transactions are already being laid, signalling that this is more than a concept for the future.
What happens at the front end, facing the customer, is increasingly shaping everything that happens at the back end.
Walmart has launched a unified framework for AI agents, consolidating multiple retail roles into four domain-specific “super agents” for shoppers, employees, suppliers and developers. Together, they personalise the shopping experience, provide timely recommendations and streamline operations across the retailer’s vast ecosystem.
Brand campaigns today and tomorrow
The reality today is that brands are in contention for the first conversation customers have with these shopping agents.
Suppose a shopper asks about a skincare routine for sensitive skin or noise-cancelling headphones for travel, the products that surface are those with well-structured data and clear, differentiated attributes.
Provenance, sustainability scores, and performance claims are not just nice-to-haves; they are the features that ensure the agent can identify and recommend a product in response to a specific query.
The next step, however, points to something more transformative.
Imagine if the information captured by retailer-owned shopping agents could also inform advertising agents.
The intent signals from a shopper’s interaction could feed directly into campaigns, allowing brand agents to trigger hyper-personalised messages across different channels.
Picture a headphone manufacturer whose advertising agent detects that a customer has been researching product specs: it could instantly serve a podcast ad demonstrating the benefits of noise cancellation in context.
In beauty retail, an agent could run interactive, mobile CTV ads that draw on a customer’s previous skincare chats with a brand’s bot, dynamically tailoring product demos, advice, or even virtual try-on overlays.
This is where the promise of agent-to-agent collaboration becomes visible. Retail agents capture the signals, and advertising agents turn those signals into intelligent, real-time activations — effectively a more dynamic, agentic retail media.
Agents as autonomous dependents
The industry is already moving towards greater trust in AI autonomy.
Just 6% of senior European marketers believe their organisations will never allow AI to make independent decisions. Nearly half of businesses developing in-house AI models already grant some decision-making authority, while fewer than one in five insist on complete human oversight. The clear lesson is that, to scale effectively, AI must be permitted to act, not simply observe and analyse.
That said, balance is crucial. Human creativity and control still shape how a brand looks, feels and evolves.
AI might execute in real time, but marketers remain responsible for ensuring campaigns reflect brand values and resonate with customers on a human level.
Dependency is a critical factor in this discussion. The rise of generative AI and zero-click search has often been portrayed as a threat, sparking fears of declining organic traffic. But the next wave, led by AI agents, could instead create a more symbiotic relationship between brands and retailers. Rather than sidelining retailers, agents position them much closer to the centre of how AI commerce operates.
The reason is simple: agents like Sparky need enormous amounts of high-quality data to collapse thousands of options into a handful of meaningful choices.
Identity becomes a vital enabler, ensuring campaigns reach real individuals in measurable, consistent ways. Signals from loyalty programmes, purchase histories and customer interactions are the assets that will determine whether the agent can deliver not just any option, but the right one — rooted in an understanding of each individual’s context, habits and preferences.
This is why retailers are evolving into the nervous system of AI-to-AI trade. They provide the connective layer through which data flows, campaigns are activated, and signals move seamlessly between retailers, brands and agents.
Journeys still span many channels — from email to stores, apps and TV — but retail media is increasingly the coordination point that ensures these touchpoints connect intelligently.
The value of commerce data
Reflecting this shift, some retailers are prioritising interoperability, anticipating that shoppers will expect agent-powered experiences to coordinate across multiple brands and networks.
At the same time, major players are escalating efforts to block third-party AI shopping agents and bots from scraping their sites, underscoring the value of proprietary data.
We are standing at the threshold of significant change. Ultimately, the success or failure of AI agents depends on the quality of the retailer data.
Those who treat it not as a commodity to be traded, but as the engine of smarter, two-way relationships with customers, will be the ones who thrive in the agentic age.
Esme Robinson is director, platform solutions at Epsilon