At NRF 2026, the message was sober: agentic AI isn’t a demo, it’s infrastructure. SAP, Dunnhumby and Mirakl explain why data, stores and discovery now determine whether retailers remain visible, trusted and chosen.
(Adobe Stock)
Agentic AI isn’t a shiny add-on. It’s becoming existential for retail.
For years, retailers were told the future was digital, frictionless and screen-first. Stores were liabilities. Loyalty was a program. Data was something marketing worried about later. That story has quietly collapsed.
At NRF 2026, the language may have been louder – agentic, autonomous, AI everywhere – but the underlying message from retailers, platforms and data specialists was far more sober. Agentic AI isn’t about clever demos or experimental pilots. It’s about whether a retailer can still be found, trusted and chosen in a world where machines increasingly mediate the buying journey. And if they can’t, they won’t remain viable.
We caught up with leaders from SAP, Dunnhumby and Mirakl to understand what’s really forming – and why this moment feels different from every other ‘AI wave’ retail has lived through.
“Agentic is the word of the day,” said Sarah Richter, CMO at SAP. “It’s on every booth. It’s on everyone’s lips.”
But what interested her wasn’t the hype – it was what retailers were actually doing with it.
“When you think about AI agents, it’s all about data,” she said. “It’s about how you activate the oceans of data you already have, often stuck in silos, and use AI to activate it in real time, not days or weeks later.”
That shift in timing matters. Retail has spent the last decade optimising analysis after the fact. Agentic AI pushes decision-making into the moment of intent – when a customer is browsing, hesitating, asking, comparing or walking into a store.
And crucially, that moment still happens offline.
Advertisement
The ‘death of the store’ narrative hasn’t aged well. “I don’t believe the in-store experience is disappearing,” Richter said. “In-store is still just as much a channel and interaction point as any digital one.”
What’s changed is that digital intelligence is now flowing into physical space.
She pointed to Molton Brown, where in-store fragrance finders draw on customer data from past digital interactions to shape recommendations in real time. “That’s immediately bridging the gap between digital and in-person.”
From Dunnhumby, Erich Kahner, director of customer strategy and insights for the Americas, made a similar point from the data side.
“What we learned very quickly,” he said, “is that retailers with an installed store base who can blend digital with physical have an edge. Digital isn’t the tail wagging the dog. It complements what really matters – having a great place for people to shop.”
Legacy retailers didn’t survive by accident. They survived because physical scale, combined with data and experimentation, turned out to be an advantage – not a drag.
Where agentic AI starts to feel existential is loyalty – or rather, how fragile it really is.
“You don’t get loyalty when you rely on siloed data,” Richter said. “You get the illusion of loyalty.”
She cited research showing that while over 70% of consumers say they’re loyal to a brand, more than half will switch after a single bad experience – often without complaining.
“When data is siloed, you’re not actually understanding your customer. You’re not personalizing. And you don’t really understand your business.”
That problem gets worse in an agent-mediated world. If an AI assistant is choosing which product to surface, which retailer to trust, or which deal to recommend, weak data doesn’t just limit growth – it makes you invisible.
For all the urgency, no one pretended this was finished technology.
“I don’t think any retailer is getting it right yet,” Kahner said. “They’re all in experimentation.”
Even the biggest players are piloting cautiously. Walmart’s agentic shopping tools are still market-limited. Failures are common – and expected.
“Most AI projects should fail,” Kahner said. “That’s normal for innovation.”
What matters is how retailers fail. Blind enthusiasm, he warned, is as dangerous as inertia. “There’s a risk of getting too enamored with AI agents. You lose human judgment, domain knowledge and the creativity that category managers and CPG partners bring.”
Agentic systems, he noted, are still “like toddlers – learning to walk.” Powerful, but fallible. Autonomous retail remains a long way off.
Advertisement
From Mirakl, the focus was less on automation and more on discoverability. “We’re still knee-deep in the experimental phase,” said Amelia Van Camp, head of agentic commerce. “But retailers are asking themselves, ‘How do I show up? How do I make sure my products are purchasable in these environments?’”
That question now extends beyond a retailer’s own site.
“Data is multi-sourced now,” she said. “It’s not just your website or product listings. AI agents are pulling information from Wikipedia, Reddit, third-party sources.”
This means that brand trust, accuracy and data hygiene now sit outside traditional retail control. “If you don’t solve discoverability, ranking and transaction,” she warned, “the real risk is that you disappear.”
Advertisement
Retail has weathered plenty of “existential” technology shifts before. Most weren’t. This one might be.
Agentic AI doesn’t just change how retailers market. It changes how customers arrive, how choices are filtered, and how relevance is decided – often without a screen ever loading.
As Kahner put it: “Just like online search became table stakes in the 2000s, agentic commerce will become table stakes.”
The difference is that table stakes now determine whether you’re even invited to play.
Asked what retailers should do now, the answers converged. “Loyalty can’t just be a program,” Richter said. “It has to be a strategic compass.”
Van Camp echoed it: “Look at your data holistically. Understand how it supports your omnichannel strategy – and how it shows up in agentic environments.”
And from Dunnhumby: don’t panic – but don’t look away. “Keep experimenting,” Kahner said. “Use the right data. Remember the retail basics. And don’t take your eye off prices, promotions and products.”
Because agentic AI isn’t replacing retail fundamentals. It’s judging who actually understands them.
Commerce media is moving fast. The Drum Awards for Commerce Media are here to recognize the work keeping pace – and setting the standard. Join us in Miami in April as part of The Drum’s year-round Commerce Media program. Learn more and enter.