Artificial‑intelligence firm Anthropic announced on September 2 that it is providing retailers with design blueprints to build shopper and merchant agents on its Claude language model. The move comes as online shoppers increasingly use AI to compare prices, check product availability and receive personalized recommendations, and as retailers look for ways to boost conversion rates before the holiday shopping period.

The blueprints are a set of guidelines that help developers create conversational agents that can suggest items based on a customer’s preferences, add products to a shopping cart, and provide inventory and pricing information. According to Anthropic, the merchant blueprint can recommend inventory, prices and marketing campaigns but does not complete purchases on behalf of shoppers or merchants.

Early adopters of the blueprints include major e‑commerce and payment companies such as Shopify, Visa, Mastercard and Accenture. In pilot deployments, one partner reported that the use of a Claude‑powered shopper agent increased average cart size by 30–35% and made customers 60% more likely to complete a purchase. The company’s head of product, Angela Jiang, said the results were encouraging.

The announcement follows a surge in AI‑driven traffic to retail sites. Adobe Analytics reported that in the first quarter of 2026, AI‑generated visits to U.S. retail websites rose 393% year‑over‑year. The same report found that traffic from AI sources converted at a rate 60% higher than traffic from other channels, and that revenue per visit was 37% higher.

Anthropic’s Claude model is available in several tiers – Haiku, Sonnet, Opus and Fable – and can be accessed through an online chatbot or an API. The company has positioned Claude as a “problem‑solving” LLM that can be fine‑tuned for specific business needs. The new blueprints are part of a broader effort to enable enterprises to deploy AI agents that can reason through complex tasks and execute actions autonomously.

Retailers can integrate the blueprints with their existing inventory and order‑management systems. The agents can pull real‑time data on stock levels, apply dynamic pricing rules and suggest complementary products. Because the blueprints do not authorize the agent to finalize a transaction, merchants retain control over the checkout process.

The release comes at a time when retailers are under pressure to improve the online shopping experience. According to Adobe Analytics, AI‑driven site visits convert 60% better than non‑AI traffic, a trend that has reversed the previous year’s pattern where human‑driven traffic had a higher conversion advantage.

Anthropic’s move is also aligned with broader industry trends. Several AI startups and established tech firms are building commerce‑focused agents that can handle product discovery, price comparison and personalized recommendations. The company’s public‑benefit status and its focus on safety and alignment are intended to address concerns about the responsible use of generative AI in consumer‑facing applications.

At present, the blueprints are only a starting point. Developers will need to customize the agent logic, integrate it with back‑end systems and test it in real‑world scenarios. Anthropic has not announced a commercial pricing model for the blueprints, and it has not disclosed the number of retailers that have adopted the design.

The announcement underscores the growing importance of conversational AI in retail. As shoppers increasingly rely on AI to research products and make purchase decisions, retailers that can provide seamless, personalized interactions are likely to see higher engagement and conversion rates. Anthropic’s Claude blueprints offer a structured approach for building such agents, but the ultimate success will depend on how well retailers can tailor the agents to their specific customer base and operational workflows.

In the coming weeks, retailers that have adopted the blueprints will likely share performance data and best practices. Anthropic may also release additional tooling or support resources to help developers implement the agents more efficiently. The broader market will watch to see whether the use of Claude‑based agents can deliver sustained improvements in conversion and average order value across different retail verticals.

For now, the blueprints represent a concrete step toward integrating generative AI into the core of the retail customer journey, offering a template for agents that can understand preferences, recommend products and manage inventory data while leaving the final purchase decision to the customer.