Curate the journey before you take it
My role
Lead product designer, visual designer, user researcher
Skills & technologies
Product design, UI design, AI, visual design, design systems
Industries
Travel planning, AI tools
Timeframe
12 weeks
Select final screens for Lucid.ai
What is Lucid.ai?
Lucid.ai is a travel planning tool that analyzes travelers' written input using AI to generate personalized recommendations, shown through moodboard-style imagery.
My team and I collaborated with a PM from Kayak, who provided guidance and professional expertise. This experience enabled us to ground our design in real business considerations in addition to our careful research and design iteration process.
Solo travel overwhelm
Solo travelers are both blessed and burdened by the responsibility of planning their own adventures.
They have the freedom to curate journeys tailored precisely to their wants and needs, but the fragmented nature of available tools and information often leads to frustration and stress, rather than inspiration and excitement.
Key insights from initial problem research
Travelers value both well-planned micro-itineraries, as well as a structured framework for their overall trip.
Travelers are drawn to personalized unique experiences, though they may be hard to plan for.
A key feature of an effective travel planning process is visualization of the trip (mentally, descriptively, literally).
What if travel planning was like creating a personal moodboard…
where different kinds of visual, written, and spoken inputs could be used both as self expression AND as a way to create a cohesive plan for a desired vacation?
Concept & Vision
Gentle nudges, customizable content, and a multi-step process mean that users have autonomy over the information they provide and the information they're given via the platform.
We've designed the interactions to be unique, but familiar, to bring a sense of ease and joy to the user experience.
Lucid.ai is situated within the numerous service touchpoints involved in travel; from hotel managers backstage, to service professionals in cafés, to shows and performances users might experience in a new city.
Iterative design process
To gauge the effectiveness of our design and how it aligned with user needs, mental models, and expectations, we conducted concept tests and cognitive walkthroughs at low- and mid-fidelity prototyping stages.
Low-fidelity prototype testing
We started with grayscale low-fidelity for our concept testing.
This first experience prototype walked our users through selecting where they're at in their planning process, uploading different forms of media for their moodboard, and using their uploads in combination with AI to create the moodboard, followed by trip recommendations.
I designed the key screens for the first part of our hero flow: the uploading media and moodboard editing screens.
Method
Semi-structured interviews
4 participants
Objectives
Get early reactions to our AI-assisted moodboarding
Assess user perceptions of integrating AI into travel planning
Low-fi testing insights
Need to strengthen our correlation between inputted user information and resulting moodboard & itinerary
Users were reluctant to upload their own media
Users generally consented to use of AI, but recommendations must be viable.
Solo travelers gain confidence in recommendations from multiple different sources.
Changes to make
Fewer recommendations (one flight, one stay, three recommended experiences) instead of full itinerary
Text box with similar affordances to existing LLMs instead of uploading assets
Destination entry instead of selecting “planning” phase
Reformatting moodboard
Calls to action to external sites for more information
Mid-fidelity prototype testing
Mid-fi took our concept testing into account, and we simplified the onboarding process to make it easier for users to dive in. We introduced tags and a text-entry section, reminiscent of using an LLM, to help people generate moodboard content.
We skipped the uploading media step and went straight into a generated moodboard based on text and trip plan input so that users could see and edit their moodboard more quickly; from this screen, they could upload media, edit individual tiles, or regenerate the whole thing.
I pulled Montréal-specific images to use in this second iteration of the moodboard visualization, and created a layout that would be familiar and easily discoverable for users, which felt important in this new format for creative travel planning.
Method
Hybrid concept testing + think-aloud protocol
3 participants
Objectives
Hear thoughts on changes to the tool’s flow, mid-fi visual representation of moodboard, edge cases
Mid-fi testing insights
Too few recommendations creates skepticism
Users expected a final page summarizing selected itinerary—then expecting to pull out payment information to book
Hierarchy of dates first, then flight, then stay in beginning screen
Changes to make
Subtle informational indicators before moodboard to help user anticipate AI and moodboard visual
Additional recommendations to allow for user exploration
Adjustments to card layout and size, to minimize dark patterns and unintended bias (size of cards)
Changes and additional calls to action (add to plan), and a finalized trip screen
Additional details for each recommended booking, more evident AI justification
Input fields to enter logistical information at beginning of flow
Lucid.ai Final Design
Curate the journey before you take it
Lucid.ai is a travel planning tool that analyzes travelers' written input using AI to generate personalized recommendations, shown through moodboard-style imagery.
Add as much or as little information as you like
The final prototype simplifies the onboarding: just an optional entry of destination, dates and pre-made bookings. Most of the user context is captured with this LLM-style text entry input, where user can say as much as they wish and the tool would recognize and tag their preferences in real time.
I helped my team pick a color scheme and put together a cohesive design system so that our final product would be visually striking, give a sense of calm to our users, and be easy to navigate using familiar patterns in a novel way.
Visualize your ideal trip
These preferences then shape up their moodboard. A visual way to dream up their trip with a way to iterate on it. Add or delete preferences, upload anything of your own, remix and regenerate until your vision is realized.
I was responsible for the design of our moodboard visual design and of overall experience we created in this phase of the planning. I made sure to create opportunities for customization, without being overwhelming or confusing.
The limited number of images and media on the moodboard screen was a deliberate choice; I felt that in order to create a curated, personalized feeling, users needed to be able to see all the content at once—no mindless scrolling.
Personalized, customizable itinerary all in one place
After finishing the moodboard, users will receive personalized suggestions for experiences, flights, and stays that they can choose from to customize their full trip plan.
Reflection
One travel planning platform can't do it all
Though we cited disjointed sources and information as a root of the experience problem, our qualitative research also reported users gain trust from verifying legitimate plans from multiple different sources.
How good is the recommendation?
AI’s reputation depends on how good their recommendation is, and how closely it aligns with the user.
Though it’s currently out of scope for our experience, receiving feedback is important in building long term trust with the user and AI analysis and recommendation.
Hitting the sweet spot of number of recommendations to balance trust and autonomy
While too many choices lead to overwhelm and decision fatigue, having too few can also feel like a disjoint experience as users lose trust in a tool that gives them no choice at all, taking away their ability to compare and pick options.
Having a flexible onboarding accommodates different levels of users’ patience
Some wanted the visual content right away, while others were comfortable taking more time to provide valuable input using Lucid. This is indicative that solo travelers may have the same high-level goals but unique individual behaviors.