While AI took center stage at NRF 2025, it was the mention of AI shopping assistants that captured the most attention among industry leaders.
On an enterprise level, technology giant Nvidia made a big splash in the space with the release of its blueprint for AI retail shopping assistants, a generative AI reference workflow that retailers can use to develop their own AI shopping assistants. Similarly, last month, Amazon launched Nova, a framework of massive data sets which can be used to build AI-powered, sophisticated shopping agents.
By definition, these agents are built on a more sophisticated AI system referred to as agentic AI which is trained to reason and act on behalf of shoppers. Unlike traditional AI that simply answers questions, agentic AI can build agents that can multitask. Nvidia, for instance, claimed its blueprint can understand text- and image-based prompts, search for multiple items simultaneously, and complete complex and contextual tasks. While these features, to some extent, explain the gold rush for these agents, it is still early days as many retailers are still struggling to gain access to clean data.
Unpacking fancy AI agents: The way Miya Knights, director and publisher of Retail Technology, sees it, “the best way to put it is [these are] shopping assistants that shoppers can talk to in a natural way, like they do with ChatGPT right now.”
“What we think we're going to see for consumers is a shopping agent, a shopping assistant, but it’s based on agentic AI principles, which essentially means it’s an AI that acts as an agent on your behalf,” Knights told Retail Brew at NRF.
Usage of ChatGPT continues to rise in the US. This matters, Knights pointed out, because retailers are seeing more natural language search queries pop up—things like “help find me a sofa that won’t attract dog hair.”
“When the consumer is starting to think in much more natural language ways and interacting with e-commerce systems and digital systems in more natural ways, they need to upgrade what they’re doing in the backend to be able to manage that,” Knights explained.
On the sidelines of NRF, Nvidia’s Cynthia Countouris, director of retail for AI and CPG, told Retail Brew its model is designed to enable retailers to understand how to use generative AI to provide a “very natural” experience for customers.
Retail news that keeps industry pros in the know
Retail Brew delivers the latest retail industry news and insights surrounding marketing, DTC, and e-commerce to keep leaders and decision-makers up to date.
Countouris compared it to putting a retailer’s best store associate or stylist or interior designer in the hands of the consumers 24/7: “That’s really where we’re headed is these AI agents, in being agentic, can do multiple things all at one time.”
Countouris said retailers will be able to ingest their catalogs of information into Nvidia’s framework and then start to play with it. SoftServe, an IT vendor in the retail space that works with Inditex, the owner of fast-fashion brand Zara, took Nvidia’s blueprint and turned it into an actual turnkey product, and one of the first features it is going to offer its customers is virtual try-on, Countouris added.
Meanwhile, Justin Honaman, head of worldwide retail and consumer goods industry strategy and business development at Amazon, told Retail Brew that “agentic AI” would be his pick for the buzzword of NRF 2025: “And the whole idea is how do you progress AI so that it performs a task for you without having to do so much manual work?”
Sticky trap: As retailers explore ways to integrate these AI assistants into their businesses, they need to know that it is simply not enough to fawn over a fancy AI model. Bohdan Khomych, associate director of research and development at SoftServe, said that “data privacy and bias in data should be at the top of the vetting list for retailers exploring agentic AI models.”
“Proper governance is essential to decrease risks of damaging a brand’s reputation and credibility with consumers,” Khomych wrote in an emailed response to Retail Brew’s query on challenges retailers should be mindful of when coding AI agents into their businesses. “It’d take just seconds to ostracize entire groups of consumers who notice an AI assistant consistently misrepresenting sizes for certain body types or pushing products that subtly exclude certain demographics.”
Amazon’s Honaman also cautioned that while it’s exciting to talk about agentic AI, it is not yet part of organic chatter among retailers, which typically tend to move at a glacial pace when it comes to adoption of new technology solutions.
“The excitement around AI is great, but when the customers sit down with us, they want us to cut through all the buzzwords and say what’s real,” he said.