(Sponsor)
Moodboarding with Lucid
Reimagining how solo travelers plan itinerary on booking platforms through aesthetic expression and an AI-assisted flow.
AI DESIGN
DESIGN SYSTEM
WEB UI
B2C
May 2025 – June 2025
CONTEXT
How do you collect inspiration for things to do on vacation — before you even plan the trip?
You may think of different travel sites, social media, or personal recommendations. However, these resources can be separate from one another, disjointed, and hard to keep track of, making the planning process overwhelming.
Kayak is one of the leading resources travelers rely on for planning, and with guidance from a senior PM on Kayak’s advertising team, my team and I set out to reduce the disjointedness and overwhelm in travel planning — particularly through the lens of the company’s beta AI chatbot, Kayak AI.
THE SOLUTION
Powered by Kayak’s AI beta, Lucid is a web application for solo travelers that generates personalized trip itineraries from moodboard creation.
Lucid’s guided AI flow helps travelers envision a meaningful journey.
Unlike other travel tools or social media that overwhelm with options, Lucid walks users step-by-step with a single AI chat prompt to shape travelers’ ideal plans.
Lucid generates visuals to spark imagination of personalized vacation experiences.
After analyzing written AI chat reflections, Lucid creates a moodboard of vivid imagery portraying a range of itinerary items to be customized to the user’s taste.
Lucid transforms moodboard inspiration into bookable items.
Lucid connects the moodboard to real-life experiences travelers can book, including an AI justification, external reviews on other travel sites, and other details.
MY ROLE
I owned Lucid’s visual design system and component library, shaping the product’s high-fidelity look and feel.
I ensured visual brand consistency with a refined attention to detail.
I aligned our prototype’s components, color swatches, type system, and interactions with Kayak.ai’s design system to streamline development handoff.
I incorporated revenue streams without compromising on user experience.
I integrated our corporate advisor's business and design feedback into Lucid's features that strengthened our market fit and advantages against competitors.
I led both generative and evaluative research to ground my design decisions in user needs.
Collaborating with two other UX designers, I initially served as the team’s primary UX researcher before leading the visual design of our product’s mid- and high-fidelity screens.
IMPACT
After handoff to our stakeholder at Kayak, several of Lucid’s design insights were reflected in Kayak AI’s official release in September 2025.
While my level of influence is uncertain, Lucid’s core design principles clearly carried through to development.
Kayak AI’s considerable improvement in visualization and imagery was worth of investment! This speaks to the strength of Lucid’s approach.
I challenged the AI chatbot paradigm with new possibilities for interaction.
My design’s appeal demonstrates LLM’s potential for other interactions without relying on chat-only mental models.
Lucid achieved 100% positive adoption intent among surveyed participants.
While stated preferences do not always align with real behavior, our evaluative testing supports Lucid as a valuable first step toward rethinking AI interaction beyond chat.
Understanding the problem
THE PROBLEM
Solo travelers are often overwhelmed by the fragmented nature of planning tools, information and recommendations found online.
How might we reduce the stress of travel planning for solo travelers, while ensuring their resulting plans reflect both their emotional and practical priorities?
GENERATIVE RESEARCH
I conducted 6 in-depth interviews with solo millennial travelers to understand their behaviors in planning and discovering travel itineraries.
THE KEY FINDING
Visual and descriptive content is one of the highest importance aspects to solo travelers during the planning process.
One participant mentioned their primary source of inspiration was social media, while another mentioned they preferred following written blog posts. Universal was the motivation to imagine the experience with detailed information about the trip, before committing to any items.
However, before designing with this insight in mind, my team and I needed to also examine Kayak AI’s beta to learn from the swings they already had taken.
A closer look at Kayak AI
THE KEY SHORTCOMING
Kayak AI’s beta does not deliver clear, visible choices to users. It instead relies on memory and prior familiarity with the platform.
I conducted a heuristic evaluation to identify current design issues that hindered usability and failed to support users’ emotional goals.
In terms of Jakob Nielsen’s usability principles, this fails recognition over recall — a core usability principle that preaches making elements, actions, and options visible and evident to lessen cognitive overload.
EXHIBIT 1
Exploring trip micro-itinerary is difficult.
Though macro-itinerary planning (e.g., flights, hotels) is robustly handled, the platform fails to support discovery of day-to-day travel plans and items, underdelivering on user needs for end-to-end travel planning.
EXHIBIT 2
Next steps are unclear with chat-based input.
Its traditional open-ended chat interface lacks affordances or guidance, often defaulting to vague responses like “I can also do XYZ thing next,” rather than surfacing concrete next steps.
EXHIBIT 3
Visuals and structured output are limited and variable.
Current chat responses feel flat and text-heavy, missing the opportunity to present information in a dynamic, visual, or contextual way. This is especially crucial in travel planning, where users’ emotional goals are often reached from vivid images, maps, or other helpful structured content.
BUSINESS CONSTRAINTS
To leverage Kayak AI’s strengths, I examined Kayak’s existing profit models.
My team and I discussed overall profit and advertising models with our stakeholder to understand our necessary business constraints.
The stakeholder was an SME of Kayak’s click-referral commission model, where the company earns a fee each time a user clicks a travel listing and is redirected (even without completing a purchase). I kept this in mind to incorporate in mid-fidelity design iterations.
TRANSLATING INTO DESIGN GOALS
To set the stage for our design, I prioritized goals with the MoSCoW method, balancing heuristic issues, business needs, and our intended emotional goals.
The final design
THE UPSHOT
Lucid is a web application for solo travelers that generates personalized trip itineraries from moodboard creation.
AI CHAT WITH STRUCTURED FLOW
With a hybrid questionnaire + chat structure, Lucid generates richer, more focused results.
Lucid enhances Kayak’s chat-only interface by first prompting the user for key trip details (e.g., location, dates) before transitioning to AI chat.
Travelers then reflect on their individual thoughts and emotions for the ideal trip with Lucid’s chat prompt, “What matters most on your trip?” As they type, tags appear below via NLP to guide expectations and clarify responses.
VISUAL EXPLORATION
Lucid delivers on travelers’ need for rich imagery through a moodboard inspired by their AI chat input.
The interactive moodboard sets a vivid tone for travel planning, allowing the user to explore, personalize, and curate media that aligns with their ideal travel vision.
RECOMMENDATIONS
Lucid translates visual inspiration into actionable items for the traveler’s itinerary.
Once the moodboard is finalized, Lucid curates a personalized list of experiences, flights, and stays for users to explore, add to their plan, and/or book directly. With brief AI explanations to ensure transparency, Lucid’s recommendations synthesize both the user’s structured inputs and the moodboard’s expressive elements.
BUSINESS MODEL
Lucid connects users to trusted sources, adding value for both travelers and the business.
Research showed that travelers often verify the legitimacy of online suggestions with multiple sources: a key opportunity to integrate Kayak’s click-referral model into Lucid’s recommendations. The platform also provides external links to reviews, maps, and partner offers, supporting natural user behavior while generating revenue.
Design system
THE UPSHOT
I reverse-engineered and expanded Kayak’s design system to give Lucid a cohesive, pixel-perfect look.
Lucid’s strength lies not only in its interaction design and flow, but also in its visual alignment with Kayak’s existing ecosystem.
Without access to official design files, I recreated and refined components from both Kayak’s primary site and AI beta, ensuring a consistent, production-ready visual language that authentically matched the company’s brand.
Epilogue
IMPLICIT IMPACT
I believe Lucid and its research insights influenced Kayak AI’s official release in a number of ways.
My direct collaboration with Kayak’s corporate stakeholder concluded with a design handoff and final presentation in June. However, three months later, Kayak AI showed clear improvements that aligned with Lucid’s core design principles.
Guiding users with clearer instructions and interactions.
Shown in the demo video below, new elements like clearer calls-to-action and suggested prompts provide more structure to Kayak AI’s interactions, making it a more focused and effective tool.
Supporting micro-itinerary with images and external links!
In an extremely similar execution to Lucid, Kayak now enables travelers to discover new experiences and micro-itinerary by providing rich information through the chat interface.
PARTING INSIGHTS
One travel platform cannot do it all!
While fragmented sources make travel planning challenging, no single platform (Lucid included) can handle the entire planning process. Since travelers rely on multiple sources to compare and weigh each differently, effective design should acknowledge and support this observed behavior.
What if the traveler doesn’t like Lucid’s recommendation?
AI’s reputation depends on how well its recommendation aligns with its user. As a next step, Lucid would need to receive feedback on recommendations to build trust in AI interaction design.
Lucid demonstrates how AI interaction can evolve beyond just chatting.
As AI design rapidly converges around similar interaction patterns, Lucid stands out by demonstrating new, research-backed ways for people to engage and reach their goals with LLMs.