Travel advice on LLM-powered Alexa+
Giving Alexa users the best travel advice in the world: Architecting the Multimodal UX for Alexa’s first Expert Advisor experience
Project context & scope
Role & Duration
Target modalities & Devices
Lead UX Designer, Expert advisor experiences
~8 months
Voice, Text, Visuals
Echo devices, Web, Mobile, FireTV
Team complexity
Cross functional: Product management, engineering, QA, Business Development, Marketing
Cross organizational: Local Info, Alexa web, Alexa mobile, Device & Design Group(visual design team), Multimodal developer tools, Smart briefings
3P content partners
Partners considered: Fodors, Trip Advisor, Lonely Planet, Travel & Leisure, US News & travel, GetYourGuide
Problem context
Desk research revealed that travel planning is defined by information overload and friction. The average traveler views 141 pages of content across a mix of online and offline resources before making a booking decision. This fragmented experience presented two key problems:
High Cognitive Load: The current experience is exhausting, requiring the user to manually synthesize disparate pieces of information (reviews, guides, booking sites) into a coherent trip plan.
Lack of Personalization: With 55% of travelers starting without a specific destination in mind, the existing tools are reactive and non-customized, failing to serve the user who needs inspiration and personalized direction. 57% of US travelers feel that brands need to tailor information based on personal preferences
The strategic goal: Make Alexa an Expert Advisor (not a search engine)
The strategic goal was to leverage Alexa's unique capabilities—the accessibility of voice and the generative power of the LLM—to replace this fragmented, non-customized process.
This required architecting a system that could uniquely:
Synthesize: Instantly combine disparate information (from 3P partners like Fodors and Trip Advisor) into a single, comprehensive itinerary or piece of advice.
Contextualize: Proactively tailor advice based on user history and conversational context, moving beyond reactive search.
Maintain Continuity: Seamlessly bridge the journey from travel research to planning and booking across the Echo, Web, and Mobile modalities, always using the right modality depending on the fidelity of information and interaction required.
The Big Win
The Minimum Lovable Product (MLP) earned exceptional user trust, scoring 6.19/7.0 CSAT and securing 88% positive sentiment from the initial beta cohort.
Achieved an exceptional 2.66% Customer Perceived Defect Rate (far surpassing the 15% launch target) with 96% LLM Accuracy (API Invocation Accuracy/Expert Selection Accuracy), validating the quality of the new UX architecture.
Conceptual design & System architecture
Based on the desk research and an audit of potential travel partner websites, I translated the strategic goal of “expert advisor” into a cohesive, functional system that could manage state and context across multiple devices and follow the users as they go about the trip planning experience with touchpoints on Echo, mobile and web.
Data modelling
The complexity of the LLM and multi-modal experience needed a clear definition of the data model. These are broad strokes to understand the system inputs, outputs and LLM technology- both from the general Alexa LLM layer and the Travel partner content specific layer. I worked with engineering to define the shared state API requirements to ensure that context and history—not just search queries—were preserved, allowing the system to move seamlessly with the user and help them wherever they may be in their trip planning experience, or on whichever modality or device they prefer.
Cross-device, multi-modal user flow
From ideation sessions, I helped develop a deep understanding of the workflow and where Alexa can be most helpful to the user. The key learnings were that the trip planning experience can be broken down into 3 main phases
Exploration/Discovery
Alexa as Expert Advisor
Activities like preparing for the trip
Using these insights, I created a user flow to tie in the different phases of trip planning to which device and modality would that phase be most effective in for the user. This also resulted in identification of another key phase “task completion”. And helped define peripheral but very important features like proactive nudges and send to phone.
The Day in the Life scenario
From ideation sessions, I helped develop a deep understanding of the workflow and where Alexa can be most helpful to the user. The key learnings were that the trip planning experience can be broken down into 3 main phases
Exploration/Discovery
Alexa as Expert Advisor
Activities like preparing for the trip
Using these insights, I created a user flow to tie in the different phases of trip planning to which device and modality would that phase be most effective in for the user. This also resulted in identification of another key phase “task completion”. And helped define peripheral but very important features like proactive nudges and send to phone.
Success Metrics
-
It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more.
Quote Source
-
It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more.
Quote Source
-
It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more.
Quote Source
Quote Source
“It all begins with an idea. Maybe you want to launch a business.”
Quote Source
“Maybe you want to turn a hobby into something more.”
Our Services
Basic
Add your pricing strategy. Be sure to include important details like value, length of service, and why it’s unique.
Intermediate
Add your pricing strategy. Be sure to include important details like value, length of service, and why it’s unique.
Advanced
Add your pricing strategy. Be sure to include important details like value, length of service, and why it’s unique.
Dignissim suspendisse in est ante in nibh mauris cursus mattis. Malesuada fames ac turpis egestas maecenas pharetra convallis posuere. Leo a diam sollicitudin tempor id eu nisl. Proin sagittis nisl rhoncus mattis rhoncus urna neque viverra justo. Adipiscing diam donec adipiscing tristique. Ultrices tincidunt arcu non sodales. Amet purus gravida quis blandit turpis cursus. In hac habitasse platea dictumst. Cras fermentum odio eu feugiat pretium nibh ipsum consequat. Nisi est sit amet facilisis magna etiam tempor. Ipsum nunc aliquet bibendum enim facilisis gravida neque convallis.