AI

AWS launches Agentic Shopping Assistant to help retailers build AI tools

Amazon Web Services Inc. today introduced a new offering designed to help retailers integrate artificial intelligence features into their online stores.

AWS Agentic Shopping Assistant, or ASA, combines several of the cloud giant’s AI products with professional services. 

ASA is derived from a tool called Alexa for Shopping tool that Amazon.com Inc. added to its e-commerce marketplace this month. The tool enables users to generate product comparisons, check how an item’s price has changed over time, and perform related tasks. It replaced an existing set of AI features that Amazon says drove nearly $12 billion in incremental sales last year.

One of the AWS services that underpin ASA is Amazon Bedrock. It provides access to cloud-based foundation models developed by the Amazon unit and its partners. Last year, AWS extended the service with an AI agent development toolkit called Amazon Bedrock AgentCore. The company stated today that ASA makes use of the toolkit.

The newly launched service also incorporates OpenSearch, an open-source search engine. AWS offers a managed version of the software that removes the need for developers to maintain the underlying infrastructure. AI shopping assistants like those ASA is designed to power can use OpenSearch to find product information requested by users.

According to AWS, the professional services included in ASA will be provided by systems integrators and a business unit called the AWS Generative AI Innovation Center. The unit will help retailers customize the service for their requirements.

Staffers from the Generative AI Innovation Center can adapt AI assistants to a company’s design guidelines, brand voice and product catalog. Additionally, the unit will implement guardrails to prevent chatbots from generating irrelevant output. An electronics retailer, for example, may wish to prevent an AI assistant built into its website from generating erroneous device repair advice.

The Generative AI Innovation Center will also develop other code customizations. According to Amazon, it can build workflows that enable an AI assistant to personalize responses based on a shopper’s chat history. Additionally, a retailer could commission analytics tools that monitor the output quality of its ASA-powered chatbots.

Building a custom AI application from scratch can take years in some cases. According to Amazon, ASA compresses that process to about 60 days.

ASA is the latest example of Amazon commercializing key building blocks of its e-commerce business. Earlier this month, it launched a service that enables other companies to ship merchandise via its logistics network. A few weeks earlier, Amazon floated the possibility that it may sell robots to industrial companies and consumer brands. The AWS parent uses more than 1 million robots in its fulfillment centers. 

Image: Amazon