Amazon Web Services (AWS) announced seven new features for its Agentic Shopping Assistant, an artificial intelligence solution aimed at retailers. The updates, announced on Saturday (26), include long-term memory, voice and image searches, and shopping cart management within the conversation itself.
The change allows online stores to offer a more personalized shopping experience, from product discovery to order completion, without relying on external assistants.
One of the main new features is long-term memory, which allows the assistant to retrieve previous conversations and recognize preferences across different visits. The system can also use data provided by the retailer, including purchase history, favorite categories, sizes and price ranges.
Another update is multimodal search. Consumers can send photographs to find similar products or talk directly to the assistant by voice, using Amazon Nova Sonic technology.
The system also now provides a conversation history organized by date, allowing the resumption of interrupted searches and purchase decisions.
AWS integrates shopping cart and expands retailers' control
The integration with the cart allows adding and removing products, changing quantities and selecting variations through natural language commands. The assistant can also identify duplicate items, recommend complementary products and suggest alternatives compatible with the budget.
The solution works with the cart infrastructure already used by the retailer and serves both authenticated customers and visitors. A new integration with Shopify, based on the Universal Commerce Protocol (UCP), simplifies catalog connection and purchasing operations.
For companies, AWS also launched an administrative dashboard that allows modifying assistant instructions, updating catalogs and connecting external tools through the Model Context Protocol (MCP), without relying on code changes.
Another new feature is the business intelligence dashboards, which present demand trends, catalog gaps and conversation performance indicators.
AWS states that the solution works within the account of each retailer, which maintains control over the code and customer data. Billing is based on the cloud services used, rather than a specific license for the assistant.
