Amazon Launches 'Rufus' GenAI Shopping Assistant
Amazon has launched a generative AI-powered shopping assistant named Rufus. The tool is trained to answer customer questions, provide product comparisons, and offer real-time recommendations within the Amazon shopping experience, setting a new standard for AI-driven retail personalization.
- Rufus is built on a custom Large Language Model (LLM) trained specifically on Amazon's product catalog, customer reviews, and community Q&A forums. It uses a Retrieval-Augmented Generation (RAG) framework to provide up-to-date information and leverages a mix of models through Amazon Bedrock, including Anthropic's Claude Sonnet and Amazon Nova. - The system runs on AWS infrastructure, utilizing custom AWS chips like Trainium for training and Inferentia for efficient model inference and reduced latency at scale. - Advanced features include tracking 30- and 90-day price history, setting price alerts, and automatically purchasing an item when it hits a target price. It can also perform visual searches using uploaded photos and transcribe handwritten shopping lists directly into a user's cart. - Since its U.S. launch in 2024, Rufus has been used by over 250 million customers. Shoppers who engage with the assistant are reportedly over 60% more likely to make a purchase during their session. - Following its U.S. rollout, Amazon expanded Rufus to the UK and several European countries, including Germany, France, Italy, and Spain, as part of a broader global AI strategy. - The assistant faces criticism for potential bias. Independent analysis suggests its recommendations heavily favor Amazon's own products and that it recommends Amazon-branded items at a much higher rate than their market share would justify.
