If data is indeed the oil of the 21st century — the one commodity everyone needs — then the growth of AI-powered search is about to expose a retail industry energy crisis. The data requirements for being found during an agentic commerce search go far beyond having the right basic product data, findable with a keyword, available on a retailer’s site.
Now, with an increasing number of conversational searches that are typically 2X to 3X longer than would typically be seen in a search bar, AI-powered searches provide many more contextual clues for what the shopper is seeking (“I’m attending a barbecue but going to a play later that day, what can I wear that will be appropriate for both?”). If a product fits into this particular situational query, the AI will surface it — but only if its data is complete, clean and ready to be accessed.
“In traditional search, having bad data hurt the rankings of individual products,” said Stuart Klein, Partner in the Digital and Analytics practice of Kearney in an interview with Retail TouchPoints. “An agentic search doesn’t return a page of rankings; it gives a recommendation based on what it knows about you, what your interactions have been, etc. That’s why people like it, because it’s very prescriptive with its recommendations, and it’s returning them [to the shopper] in a completely new format that allows them to visualize the product and interact with the recommendations.”
The challenge comes because, with AI-powered search, results are an all-or-nothing proposition: “In this case, having bad data completely eliminates you from the consideration set,” Klein said.
And even though agentic AI “is not commanding a large percentage of purchase volume right now, it is taking a lot of research volume,” Klein noted. This makes the impact of being excluded from consideration an even bigger challenge, since these AI searches are becoming the beginning of more and more shopper journeys.
An April 2026 McKinsey survey revealed that 68% of consumers have used an AI-powered tool over the past three months, with usage concentrated in the early phases of the shopper journey: 62% are using these tools to compare brands, models, prices and reviews, the most popular application for them.
Internal Data Connections are a Key First Step
It’s not just the cleanliness and quality of the data itself; it’s also about how it’s connected internally (if indeed it is) and whether it has been set up to allow for analysis and action.
“We talk about meeting the customer where they are, but retailers also are having to meet agentic search where it is,” said Nick Kramer, Principal, AI and Applied Solutions at SSA & Co. in an interview with Retail TouchPoints. That requires a bit of corporate soul-searching, he added.
“Retailers need to ask themselves key questions,” said Kramer “Is all of your data connected? Does all of your data tell one story? You might have 30 to 40 product attributes in a PIM [product information management] solution and 100 attributes in a clienteling or CRM [customer relationship management] solution — are they connected? And does the data ‘know’ where the product can actually be purchased — a store, online or on Amazon? And is this data connected to the product and customer segmentation attributes? From what we see, the answer too often is ‘no.’”
Kramer added that this is, unfortunately, not a new problem, although it has become more urgent in recent years. He discussed a master data project that SSA did for a major clothing retailer that occurred over a decade ago. “The problem wasn’t that they had data overlapping and contradicting each other,” he said. “The problem was that they had too little data, and it was stuck in pockets and not communicating with each other, and that’s still true of retail today.”
While the challenges haven’t changed, the tools for overcoming them have, he said, noting that AI has lowered the barrier for entry for getting data, organizing it and analyzing it over the past three to five years. “Then the questions become: Do I know what I’m trying to do, i.e., can I articulate how to solve the problem? Do I have the people and the tools to accomplish it?” he said.
Kramer warned that using these tools also requires a significant commitment of resources. “You will have to dramatically scale your data engineering capabilities,” he said. “I can walk into any client and they’ll say the demands on their data are much higher, as is the frequency of change and the frequency of iteration. Every data engineering team I see is swamped, and because retail has traditionally underinvested in this type of infrastructure, this problem will really separate the winners from the losers.”
Managing and Influencing External Data Sources
The rise of AI also has renewed attention on external data sources, particularly since AI-powered search takes into account a wide range of inputs — from social media text and images to influencer activity, media coverage and user reviews.
“Brands have three layers of data to influence,” said Klein. “There’s structured product data — the facts. There’s behavioral data, such as what’s happening on Reddit or TikTok, reflecting what people are doing [in relation to a brand or product]. Then there are trust and contextual signals, and those come from how others, for instance influencers and the news” interact with and comment on the brand.
Brand Messaging Becomes Harder to Control
Unfortunately, there’s only so much any individual retailer or brand can do to manage all these data sources, and that’s part of what these analysts see as a power shift from brands to consumers.
In general, “agentic commerce is undermining the brand’s ability to control the message,” said Kramer. Brands have traditionally preferred tools such as email marketing, which has been a go-to for retail for decades. “It was the [marketing channel] brands had the most control over, but they were dealing with weak or basic signals” for both demand and customer actions, he said.
“Now, AI has empowered the consumer to search on their terms; they’re using their language to find stuff, not waiting to receive an email blast or responding because there’s a picture of a puppy in the email,” Kramer added. “Now they’re saying ‘I want this, this is what I’m looking for.’ So now it’s not enough just to describe a product; you need to fit the product with the customer’s experience.”
It’s a heavy-duty data challenge, but there is an upside. The contextual information that consumers provide with their AI-powered searches also are an invaluable source of data.
And while it’s virtually impossible for a retailer to think of all the potential ways a product could be used and what problems it might solve — much less every lifestyle vibe it aligns with — “a more sophisticated approach starts with rethinking your customer segmentation and profiles, as well as all those customer ‘interest’ signals and what goes with what,” said Kramer.
“You also need to create the opportunity for the AI agents to make connections that you haven’t explicitly made,” he added, although he warned: “And that’s a much harder problem.”
