AI agents have seized their place at the retailer table, based on what CIOs and CTOs from prominent brands said at this year’s NRF Retail Big Show in New York City.
Straight out of the gate at the industry’s marquee conference, AI emerged as a definitive focus of retail giants like The Home Depot, Wayfair and URBN, whose brands include Urban Outfitters and Anthropologie. While their CIOs and CTOs talked up AI agents as an inevitable evolution of retail, they also acknowledged a difficult truth: that they had spent the better part of 10 years seeing bots as a digital plague and blocking them from their sites.
The issue surfaced during a panel moderated by Jason Del Rey, founder of The Aisle, who pointed to the public’s negative reaction to AI agents, citing his own instinct to close chatbot boxes as soon as they appear.
Angie Brown, CIO at hardware giant Home Depot — wryly noting that Del Rey’s trigger response to bots was “exactly what we’re all trying to avoid” — sought to make the case for what today’s sophisticated AI agents can offer. They’re designed to meet customers “where they are,” to help with purchase decisions, she said. Home Depot first explored AI in its search bar and search results and now aims to further interact with customers via AI agents about the projects they are working on.
“We’re trying a few different ways … to be there in the moment to help the customer through their experience,” she said.
AI agents meet real-world buyers
The goal of Home Depot’s AI use, Brown said, is to focus on solving customer problems and projects by providing the company’s expertise through the AI agents — with the intent of driving purchases. “If we can remove friction from the buying experience, and if we can help our customers and our associates with know-how, then that seems like a sweet spot for us to focus.”
The use of shopping agents remains an ongoing learning process, said Fiona Tan, CTO at home furnishings retailer Wayfair. Even instances when a customer abandons a shopping cart can be a lesson for AI and build goodwill for the brand.
“Sometimes the fact that they didn’t make a purchase is actually a win, because they found out that actually it’s not going to fit — the sofa’s too big, [and ]God forbid, you have to return a sofa,” Tan said, explaining that the information in the algorithm saved the customer from making a mistake.
Making the digital nuts and bolts of AI agents work
In a separate fireside chat, C-suite leaders from Stripe and URBN riffed about how they work with AI within their organizations and what they need to use it collaboratively.
Maia Josebachvili, chief revenue officer of AI at financial services company Stripe, asked about making life easier for URBN, and ostensibly other brands, with the handling of data in this AI-empowered world. “How do we help ensure that your products are legible by the AI agents?” she asked.
There is some work to be done on that front, according to URBN CIO Rob Frieman. The structured product data that retailers were already familiar with was readily understood by older infrastructure and inventory systems. There was little confusion about categorizing a pair of jeans. But the new infrastructure and landscape with AI agents in the mix are changing what may happen with the data.
“In a nondeterministic environment like LLM, how is it going to surface that in a way that we think is robust … each time they see a product from those cases?” he asked.
Maia Josebachvili, chief revenue officer at Stripe. [photo by Joao-Pierre S. Ruth for InformationWeek]
There can be confusion in understanding what customers want versus how data and metadata for products are handled by AI agents, Frieman said. For example, when customers interact with an LLM to find a pair of jeans, the terms they use might not mesh with an AI agent’s understanding of a product’s data. “Does your product actually say jeans?” he asked.
The product data might use the term denim instead, which could spawn digital confusion. LLM partners are learning to navigate this landscape, Frieman said.
URBN’s brands, such as Urban Outfitters, have launched an agentic shopping experience with Microsoft’s Copilot, which he said may be the future of agentic commerce. Frieman compared the current AI-driven transformation to the early days of e-commerce, which had its bumps and growing pains before it became an industry staple. “In a lot of ways, we feel like this is the same territory,” he said.
Frieman echoed some of Brown’s notions about the use of AI agents to connect with consumers as they explore potential purchases. “We really want to meet our customers where and how they want to shop, how they want to discover products,” he said.
There seemed to be a collective, determined inevitability of AI agents in retail at the conference. Frieman talked up URBN working with Stripe and Microsoft, so its customers can use Copilot to find an outfit for a specific need or occasion. He demonstrated how a customer could tell the AI the parameters of what they were looking for. Guidance from AI agents could then provide options for quicker purchase decisions. “It’s really critical,” he said. “That’s really at the crux of a smooth, frictionless customer experience, but also one that still retains that brand connection.”
Parting bits of friction to address
The rush to populate AI agents across the retail sector does come with its share of hurdles. “Product data is not easy. You want to make sure that everyone’s got initiatives for AEO [answer engine optimization]which is really about how you position the product data internally,” Frieman said.
The campaign to see bots accepted by consumers is also a significant change from past behaviors. “We spent the last decade saying, ‘No bots on our website,’” Josebachvili said. “Now we’re saying the opposite.”
Further, Frieman said even with innovations AI can offer, security must be part of the development mix to ensure each bot is a trusted partner and each customer is real. “Even if both of those boxes are ticked, you still want to make sure,” he said.
