In this week’s column, Charlotte McEleny takes a look at what the retail media ecosystem needs to do to embrace AI…

As AI reshapes how retail media is planned, sold, and bought, WPP Media's Alin Popescu, Starcom India's Rathi Gangappa and independent advisor Isabella Spragg outline what has to change on both sides of the table before the technology can do anything useful with what the industry holds.

Retail media has spent the last five years monetising attention, according to Alin Popescu, head of commerce strategy at WPP Media. The next five, he argues, will be about something else entirely. "AI is not the story," Popescu says. "Data collaboration is the story. AI is simply exposing who has it and who doesn’t." 

His argument is that the industry’s current AI conversation (automation, creative generation, and campaign optimisation) is solving for productivity, not advantage. "While important, these innovations are unlikely to create a lasting competitive advantage," he adds. "The real AI battleground is data collaboration."

That puts a different set of questions on the sell-side agenda: not how fast we can automate, but who we can usefully connect to – and whether the foundations exist for that connection to mean anything.

Sell-side: collaborate or get the foundations right first

Most retailers, in Popescu’s view, already hold valuable first-party data, but the problem is that it’s scattered. "Customer intelligence remains fragmented across loyalty programmes, ecommerce platforms, physical stores, media exposure logs, CRM systems, marketplaces, and brand-owned ecosystems," he says. "AI cannot create intelligence from data it cannot access."

The fragmentation problem is more acute in APAC than in more consolidated markets, where "brands and retailers operate across dozens of banners, channels, loyalty programmes, and varying levels of retail media maturity." 

He argues this changes where the value sits, "The value of AI will not come from isolated datasets, but from the ability to connect intelligence across an increasingly complex commerce ecosystem."

The APAC context gives that argument particular weight. According to Dentsu’s APAC Retail Navigator 2025, nearly 70% of APAC consumers now shop on fast commerce platforms at least once a month, representing a pace of adoption that cuts across markets at very different stages of retail media maturity. 

India is the clearest case of where that pace meets AI pressure. Rathi Gangappa, chief executive officer at Starcom India, says quick commerce is the most AI-forward retail segment in the market, precisely because its operating conditions demand it. 

"Real-time inventory and demand forecasting demand it," she says, but she is clear that the ad products have not kept pace. "Most are selling basic visibility slots rather than AI-optimised, inventory-aware placements." In her diagnosis, the gap is one of enablement and not infrastructure.

Isabella Spragg, an independent advisor on data, retail media, and AI strategy, shares a similar readiness concern at the network level. "Most retail media networks are still focused on building scale, proving measurement, and expanding advertiser adoption," she says. "AI is certainly part of the conversation, but for many retailers it remains an overlay rather than something fundamentally embedded into how the network operates." 

Where Popescu sees fragmentation as an argument for collaboration across organisations, Spragg sees it as an argument for getting the basics right within one first: "Many retailers have valuable customer data assets but struggle with fragmented identity, inconsistent data quality or limited interoperability across platforms."

"RMNs should be investing in the assets that become more valuable in an AI-driven ecosystem: proprietary data, trusted measurement, interoperable infrastructure and high-quality inventory," she adds. 

AI capability itself, in her view, will not be the differentiator for long. "AI capabilities will become increasingly commoditised, but differentiated data and demonstrable outcomes remain difficult to replicate."

Networks that can go beyond media to use customer insights to support planning, forecasting, and merchandising are, she argues, the ones most likely to build a durable advantage. Gangappa makes a related point on sell-side readiness: networks need to make product catalogues AI and agent-readable so products can surface through emerging AI assistants, and build native advertising experiences within conversational interfaces rather than mapping traditional ad formats onto new environments.

Privacy-safe access is where Popescu’s collaboration thesis gets technical. He points to clean rooms and federated data models as the infrastructure that makes cross-party intelligence sharing viable without retailers or brands having to hand over raw customer data. 

"Privacy-enhancing technologies, clean rooms, and federated data collaboration will become strategic growth enablers rather than compliance tools," he says. He also makes a specific case for physical retail in this shift. "Physical retail becomes more valuable, not less, in an AI-driven world," he says. "In APAC, where a significant amount of commerce still happens in-store, physical retail remains the richest source of shopper intent, transaction, and behavioural signals."

The networks best placed to compete, he argues, will be the ones that connect digital and physical touchpoints into what he calls "a unified intelligence layer".

Buy-side: transparency gets harder, not easier

If the sell-side question is what to build, the buy-side question is what to trust. Spragg flags transparency as the biggest risk facing advertisers as more AI-driven tools enter the market. "While automation can improve efficiency, there is a risk that decision-making becomes increasingly opaque," she says. 

Advertisers, in her view, will need clarity "in how audiences are being selected, the data fuelling or underpinning the tools, how budgets are being allocated and how outcomes are being measured". This clarity gets harder to maintain as more parties layer AI tooling into the chain.

Gangappa takes that transparency concern further, into a specific measurement problem. On the buy-side, she argues, the biggest challenge is proving true business impact "through robust measurement and incrementality, rather than relying on platform-reported performance alone". As AI optimises within walled environments, the gap between what platforms report and what advertisers can independently verify is likely to widen, making incrementality testing a baseline requirement for any advertiser wanting to understand what AI-driven buying is actually delivering.

Spragg adds a second pressure: fragmentation on the buy-side itself. As retailers, platforms, agencies and technology providers each build their own AI capabilities, advertisers are left reconciling outputs from systems that do not share a common standard. Her advice: stay anchored to "data quality, measurement discipline, and clear business objectives", rather than chasing every new AI capability as it emerges.

Gangappa also points to a longer-term change in how products get discovered at all. As AI assistants gain influence over purchase decisions, brands need to ensure their product data is AI-ready to remain visible in agent-led discovery environments. That raises a related challenge around direct consumer relationships, which are built on years of first-party data investment and are at risk if discovery shifts to environments brands do not control.

Spragg names personalisation, measurement, and operational efficiency as the areas where AI can act directly today. "AI has the potential to significantly improve audience selection, creative optimisation, and customer experiences across both owned and paid channels," she says. 

Further out, she expects a more fundamental change in how campaigns get managed: agentic AI handling planning and optimisation autonomously across channels. "As autonomous systems become capable of planning, executing, and optimising campaigns across multiple channels, we may see a shift away from manual campaign management toward outcome-based orchestration," she says, noting a change with implications for how retail media inventory gets packaged and bought, not just how it gets managed day to day.

The move to monetising intelligence

Whether the next competitive edge comes from connecting data across organisations, from getting a single organisation’s data and measurement right first, or from making sure the product catalogue is readable by an AI agent before any of the above, all three arguments point away from where the industry has spent the last five years competing. 

Where reach and scale built the category, what happens next depends on whether retail media networks treat their data as something to protect, something to connect, or both. And on the buy-side, it depends on whether transparency and incrementality discipline can be built into AI-driven buying as fast as the tooling itself is being adopted. As Popescu suggests, if the industry has spent the last five years monetising attention, the next five will be about monetising intelligence and, ultimately, decisions.