Amazon has unveiled Alexa+, the latest iteration of its voice assistant, designed with a new architecture that leverages generative AI. This upgrade allows Alexa+ to connect with various large language models (LLMs), agentic capabilities, services, and devices, enhancing its conversational abilities and personalization features.
The development team faced numerous technical challenges in creating Alexa+. A significant advancement was the construction of a new architecture to orchestrate APIs at scale. This system enables seamless integration with services like GrubHub, OpenTable, Ticketmaster, Yelp, Thumbtack, Vagaro, Fodor’s, Tripadvisor, Amazon, Whole Foods Market, Uber, Spotify, Apple Music, Pandora, Netflix, Disney+, Hulu, Max and smart home devices from companies such as Philips Hue and Roborock.
In addition to API integration, Alexa+ can handle multifaceted requests by stringing together multiple API calls. "The result is an experience more like one you’d have with a real-life personal assistant," according to the announcement. For instance, users can ask Alexa+ to make lunch reservations and share plans with friends through text messages.
To ensure accuracy in responses and avoid unpredictable answers from LLMs, Amazon has implemented grounding techniques. Partnerships with over 200 news outlets including the Associated Press and Reuters provide Alexa+ with accurate real-time information.
Latency was minimized by using a sophisticated routing system that matches customer requests with suitable models from Amazon Bedrock's suite of tools like Amazon Nova and Anthropic Claude. This ensures a fast yet accurate user experience.
Maintaining Alexa's personality was crucial for Amazon. The AI assistant retains its smart and empathetic nature while also being personalized based on user preferences such as favorite music or disliked foods. "We optimize each model built into our architecture to ensure they reflect Alexa’s personality," stated the release.
Furthermore, agentic capabilities were added so that Alexa+ is not limited by existing APIs alone. This enhancement aims to make it an even more versatile AI assistant capable of solving daily problems and keeping users entertained.
###