Agentic Commerce Moves from Demo to Production
Shoptalk 2026 in Las Vegas drew more than 10,000 retail and commerce leaders for sessions featuring Coach, OpenAI, Wayfair, Meta, Sephora, and others. The through-line was less about agentic commerce’s future potential and more about what is already running in production: conversational assistants reshaping how shoppers discover products, in-ad purchase flows collapsing the path to conversion, on-site agents driving brand engagement, and third-party surfaces giving brands fresh off-site channels to reach customers.
Despite the momentum, a clear gap persists between what retailers observe today and what they can confidently plan for next. Search and discovery have clearly shifted, but the market’s trajectory remains uncertain. Three themes dominated conversations throughout the week.
Retailers Need a Framework Before Scaling Agentic Efforts
Agentic commerce remains early-stage, leaving brands to figure out entry points, partnerships, product data structure, and syndication across AI surfaces without an established playbook. The pace of change compounds the challenge. In the three months since NRF, loyalty data moved from theoretical advantage to live product. Sephora’s global chief digital officer described how the brand now uses loyalty data inside its ChatGPT app to personalize recommendations and surface perks like samples and free shipping.
Discovery is where the urgency is most acute. In the Women in AI session with PwC, Klarna, Novi, and Stripe, Novi’s CEO argued that AI agents constitute a new storefront—brands that are not discoverable there risk invisibility. OpenAI’s product partnerships lead for search and commerce reported in a keynote that more than half of platform searches are discovery-based, and 70% of those include constraints. Shoppers are not typing simple keywords; they are submitting context-rich prompts about planning a trip to Greece, organizing a Super Bowl party, or comparing appliances.
For businesses, this raises the question of what a product must look like to be found by agents. In Tapestry’s session with Stripe, the company’s SVP of global digital product and omnichannel innovation stressed direct product feeds, which provide agents with more structured, current product data than web crawling can offer. Stripe’s Agentic Commerce Suite helps retailers syndicate catalogs across supported agents without maintaining separate integrations. Many retailers described taking a similar test-and-learn approach, running experiments to see how products are found, recommended, and purchased across different AI surfaces.
Commerce Is Moving Beyond the Chat Window
Agentic primitives are appearing across an expanding set of surfaces, each with its own maturity level and path from discovery to conversion. The shift extends into native checkout, customer service, and the backend systems determining whether a customer can find, buy, and receive the correct product smoothly.
Meta demonstrated one instance with a new Facebook checkout flow built on the Agentic Commerce Protocol (ACP). It takes shoppers from an ad click through product details and AI-generated review summaries to purchase without leaving the app. Not every embedded checkout is agentic, but the flow reflects broader momentum toward more embedded commerce.
In a session on the future of retail, Shoptalk’s head of content and insights predicted a new wave of consumer brands will be built natively on agentic infrastructure, aided by lower customer acquisition costs and less dependence on their own websites as the primary entry point. AI startups exhibiting at Shoptalk spanned every layer of the commerce stack—from catalog enrichment and discovery to post-purchase. Agentic commerce will not be defined by a single LLM or channel. As category-specific apps appear across fashion, beauty, home decor, and other verticals, retailers face the choice of where to invest scarce time and resources: first-party or third-party agentic experiences.
Brand Differentiation Gains Importance in an AI World
When AI makes discovery and comparison easier, trust and brand affinity must carry more weight in persuading shoppers to choose one brand over another. That message recurred throughout Shoptalk sessions.
New Balance’s president and CEO framed customer preference as a product of consistency rather than constant promotions. The company invested over $25 million in store updates in 2025 and is training associates to explain technical products in more depth to build credibility. Tapestry pointed to Coach’s ongoing Gen Z research as part of the brand’s sustained relevance. Victoria’s Secret’s CEO described growing customer preference for store experiences that feel comforting and confidence-building—what she called the “soothing economy.” Stitch Fix’s CEO introduced Stitch Fix Vision, an AI-powered visualization tool for personalized outfit discovery built on the company’s first-party client data.
As agentic commerce accelerates, brand infrastructure must work across a growing number of surfaces. Retailers need unified customer data and systems capable of carrying identity and context across channels.
Checkout Fundamentals Still Underpin Agentic Flows
In agent-driven journeys, customers often reach checkout ready to buy and less tolerant of friction, making payment performance critical. Stripe’s Optimized Checkout Suite dynamically presents the payment methods most likely to convert, drawing on more than 100 signals. Businesses see a 2%–3% average conversion increase after adopting Stripe’s optimized payment surfaces, showing how payment method presentation, fraud decisions, and checkout flow affect revenue.
Agentic flows depend on the same fundamentals driving retail performance today: fast, branded checkout; appropriate payment methods; robust fraud controls; and unified commerce data. Connecting these systems across online, in-store, and in-app channels simplifies delivering personalized, consistent experiences wherever customers interact with a brand. With the Agentic Commerce Suite, businesses can connect catalog and commerce infrastructure once, then extend into compatible agents and surfaces as they emerge.
