Anthropic has launched a set of blueprints designed to help retailers build Anthropic Claude retail agents for the upcoming holiday shopping season, with the technology explicitly stopping short of completing purchases on customers’ behalf.
The agents, built on Claude, can search product catalogues, make recommendations, assemble shopping carts, and handle customer service queries. What they cannot do is finalise a transaction. The customer still checks out through the retailer’s own existing process, with payment remaining firmly in the merchant’s hands.
What the blueprint actually does
The toolkit gives retailers more than a conversational shop front. Merchants can also use it to generate suggestions around inventory management, pricing strategy, and marketing campaigns. Built-in controls tie product and pricing recommendations to catalogue data, and the agents are designed to prevent manipulative upselling. Because the tool operates within the retailer’s own website or app, the merchant keeps control of the customer experience and the underlying data.
According to Reuters, Angela Jiang, Anthropic’s head of product on its Claude platform, is behind the push. The timing is deliberate: getting the technology embedded before peak retail traffic arrives gives merchants a window to test and refine the agents before volumes climb.
The early numbers Anthropic is pointing to are striking. The company says early tests showed that retail site visits driven by AI converted at a 60% higher rate than traffic from other sources. For one partner using the technology, cart sizes increased by 30% to 35%. Those are the figures Anthropic is leading with, and retailers will be watching whether they hold up at scale.
Anthropic Claude retail agents and the question of how far AI should go
The launch does not arrive in a vacuum. Digital Commerce 360 reports that Accenture, Mastercard and Visa are among the users listed at launch, which suggests the blueprint has already attracted partners with significant reach into the retail payments ecosystem. That combination of a major card network, an acquirer-adjacent giant, and a global consultancy points to how broadly Anthropic is pitching the tool.
Don Apgar, Director of Merchant Payments at Javelin Strategy & Research, offered a pointed read on why the product is structured the way it is. ‘Consumers are increasingly using AI tools in their shopping searches to find what they are looking for, but are not ready to delegate their purchase decisions at scale,’ he said. ‘This is where merchants are building right now, and Anthropic has recognised that with this product toolkit. Many high-profile retailers are already reporting positive results in sales lifts when AI agents can discover their products, present them to consumers, and suggest additional products that would enhance the purchase.’
That framing matters. The deliberate limitation on completing purchases is not just a technical design choice: it reflects where consumer trust currently sits. Retailers have their own reasons to keep the final transaction step in-house, not least because greater reliance on third-party AI in the checkout flow could weaken direct customer relationships over time.
Apgar was blunter in a second comment. ‘The tech is there, consumers are not yet,’ he said. ‘By launching this product, Anthropic is acknowledging that the industry is way out over its skis on agentic commerce and dialing it back to what merchants and consumers are looking for in 2026.’
Which leaves an obvious question: at what point does the calculus shift? If conversion rates and cart sizes continue to outperform on AI-driven traffic, the pressure on retailers to extend agent capabilities further into the checkout process will grow. For now, though, Anthropic Claude retail agents are positioned as a discovery and assistance layer, with the human customer and the retailer’s checkout holding the final step. Whether that boundary moves will depend on whether consumers decide they are ready to hand it over.
