Personalized chatbot that allows NFL fans to connect with Alexa as a sports buddy and have in-depth conversations about various sports topics
Designed Amazon's first generative AI chatbot experience, influencing the future of Alexa's conversational strategy and paving the way for 'Let's Chat'
Since launching in November 2014, Alexa has expanded dramatically from a basic proof-of-concept to millions of deployed devices. Early versions supported only fundamental voice commands: music playback, weather updates, alarms, and simple factual queries.
The system required users to learn rigid command structures and specific phrasing, as it lacked natural language processing capabilities, context awareness, and multi-turn conversation abilities. Every interaction required the wake word "Alexa”, creating an unnatural, command-driven communication style rather than conversational dialogue. Additionally, low user engagement with the Alexa app highlighted the need for more intuitive interaction methods, driving Amazon's investment in advanced conversational AI capabilities.
Our goal was to create a richer, more natural conversational experience with Alexa by leveraging generative AI advancements and determine whether this improved overall engagement on the Alexa app and represented a worthwhile investment.
As the project's lead designer, I spearheaded comprehensive design and research initiatives for the chatbot experience, encompassing initial research, design development, and user testing.
Conducted 3+ user interviews, designed 2 surveys by partnering with the Customer Insights team to understand user needs, pain points and evaluate designs
Defined design strategy by partnering with senior leadership across functions such as Science, Engineering and Product to ensure a balance of user needs and business requirements. Led an ideation session to identify new ideas, prioritize features based on user needs and state-of-the-art technology
Designed experiences for product launch and north-star vision that helped streamline conversations and drive decision making. Designed user flows, defined interaction patterns and created low-fi and high-fi prototypes for research studies and evalutate concepts.
Collaborated with multiple stakeholders across 7 teams from product managers and engineers to research scientists and legal partners. Presented design concepts back by research study data to senior leadership (VPs and SVPs) to invest in emerging technology like generative AI thereby influencing future products.
This experimental generative AI sportsbot focused exclusively on NFL content due to model training limitations
The MVP aimed to deliver knowledge-grounded contextual Q&A, multimodal image responses, proactive news article engagement, casual conversational capabilities, and robust handling of acronyms and typos
Some of the features that the bot supported include:
Multimodal responses embedded with images, video and links
General knowledge based contextual Q&A where Alexa infers from past conversations
Proactive engagement hooks such as related news articles to continue chats
The product was soft launched in the US in Jan 2023 on the iOS version of the Alexa App. Within a month of launch, we saw:
through organic discovery
average number of turns
used proactive pills for interactions
This experiment led to the start of a new initiative called "Let's Chat" which influenced the future strategy and direction of Alexa. It was announced by Dave Limp, the former VP of Amazon Devices during the Amazon Devices Fall Event September 2023.
