AI Shopping Is Going Mainstream but Retail Infrastructure Is Still Playing Catch-Up

AI-powered shopping was supposed to change retail overnight, but the future still feels stuck in beta. Retailers and brands are piling into shopping tools on AI platforms like ChatGPT and Claude, though adoption is still moving more slowly than the hype suggested. Even so, firms like McKinsey believe AI-driven commerce could unlock a $5T global retail opportunity by 2030.

Retail vs. algorithms: The fundamental problem is that large language models were never designed for commerce. They can predict the next word in a sentence, but managing real-time inventory, shipping costs, loyalty rewards, and checkout systems is a far messier challenge. Even ChatGPT executives have admitted the industry underestimated just how complicated retail infrastructure really is, especially once you add things like basket promos, in-store pickup, and region-specific systems. That realization is now pushing the industry into a new race to rebuild commerce infrastructure for the AI era.

• OpenAI and Stripe built the Agentic Commerce Protocol, a system designed to let AI assistants browse, compare, and complete purchases for users.
• Similarly, GoogleGOOGL
  has partnered with companies on the Universal Commerce Protocol, the more robust option that handles scheduling and returns.

AI Shopping’s Experiment Phase

Google may already have the edge, combining decades of shopping data with years of Gmail, Photos, and Drive activity to power more personalized recommendations for Gemini users. But while companies race to launch AI shopping tools, the ecosystem itself is still highly fragmented. Despite nearly 900 ChatGPT apps and 353 Claude connectors now live, users often still need to manually find, install, and activate retailer apps before getting recommendations. Even so, brands are already experimenting with more creative AI shopping experiences:

• TargetTGT
  and WalmartWMT
  have launched shopping apps inside ChatGPT as retailers race to test AI-powered commerce.
• Sephora’s app lets users redeem loyalty rewards, while Starbucks’SBUX
  beta version suggests drinks based on outfits and personal style.

Retail’s premature rush: Consumer appetite is clearly there, with ChatGPT already handling roughly 50M shopping-related queries daily despite clunky checkout systems. Yet retailer engagement remains “pretty low,” according to Alpic’s Dimitri Ewald, partly because many shoppers still do not even know these tools exist. That gap is reflected across the industry, with a Retail Economics survey finding AI adoption is moving faster in back-end functions like HR and finance than on the shop floor. As Forrester analyst Emily Pfeiffer put it, “Nobody has figured it out, but everyone has FOMO. Everyone is prematurely rushing to market.”