The Maze: An AI shopping agent can find and compare products for a shopper. That creates a difficult bargain for retailers: make the catalog easy for an outsider to read while keeping enough customer knowledge to earn the next visit. A BCG survey of 50 major global retailers draws that boundary clearly. Product facts are widely shared or earmarked for sharing. Customer histories remain guarded. The harder question is whether retailers can turn the information they keep into a better shopping experience.
A public price does not require a public pricing strategy. Among retailer respondents, 86% share product attributes and descriptions with external agents or plan to within two years. The shares are 73% for prices and promotions and 68% for ratings and reviews. These facts help an assistant understand an offer and compare it with alternatives. Separately, 87% say they would never share customer or customer-behavior data, 62% would never share margins and costs, and 56% would never share pricing logic. The distinction matters: a shopper needs the selling price, not the merchant's minimum acceptable margin. These are different questions, so the percentages are not opposing halves of one total.
The preparation starts with accurate product facts. On two-year marketplace roadmaps, 67% plan richer product data, 58% real-time availability and 45% better seller data. Conversational search is planned by 39%. Better product information therefore leads a chat interface by 28 percentage points in this sample. A marketplace lets outside sellers offer products through the retailer's platform; their inaccurate listings can weaken the retailer's whole offer. Clean descriptions help an assistant choose an item. Current stock tells it whether that choice is useful. Meanwhile, 95% of marketplaces plan at least three of the listed actions. That describes ambition, not how many have completed the work.
Keeping customer data only helps if the retailer can act on it. Purchase frequency, basket contents, brand preferences and responses to offers can help a merchant make more relevant suggestions. Within two years, 67% plan personalized product or basket recommendations, 63% tailored offers and 54% challenges or game-like experiences. Fewer plan the supporting capabilities: 42% coordination across shopping channels, 35% immediate capture of what a customer wants and 33% real-time next-best action. The last means choosing the useful offer or step while a shopper is still there. These separate plans are not a conversion funnel. They suggest a practical execution question: can the retailer use its customer knowledge quickly enough to make returning worthwhile?
Advertising needs proof as well as a new place to appear. Retail media is advertising a retailer sells around its shopping experience or customer data. The readiness pattern puts sponsored visibility on external AI platforms and recommendations inside retailers' own assistants largely in future plans. Core targeting and measurement are further along. Among retailers that launched enhancements, 65% report better campaign performance, 55% more demand and spending from outside marketplace sellers, and 40% higher advertising prices. Those are shares reporting benefits, not the size of a sales lift. The launched subgroup's size is not disclosed. Incrementality measurement asks whether advertising caused extra purchases that would otherwise not have happened. That is a stronger test for an advertiser's budget than simply counting sales after an ad was seen.
Why it matters: Our read is that retailers need one operating plan for product accuracy, customer relationships and advertising. Fixing the catalog can help win a recommendation. Using consented customer knowledge well can give the shopper a reason to return. Demonstrating extra sales can help earn the advertiser's next payment. A chat interface depends on those foundations. This small sample of major retailers describes intentions and self-reported benefits; it does not establish how the whole retail market will perform.
Sources: BCG, How Retailers Can Reinvent the Digital Flywheel for Agentic Commerce, October 21, 2026. Supplied Exhibits 3–6; retailer survey n = 50, brand/seller survey n = 100.
