I led the 0-to-1 design of KAYAK’s AI travel assistant, successfully establishing it as a core company-wide feature. After validating user value through a standalone experiment, we scaled the AI across all major travel verticals.
Ultimately, we integrated this framework directly into KAYAK’s core platform using a highly scalable design system.
Highlight #1: Seamless Sync
The side-chat architecture allowed us to simplify the widget into a lightweight shortcut card.
Clicking a widget card instantly updates the main KAYAK view, merging conversational ease with familiar manual controls.
Highlight #2: Context-Aware AI
The side-chat acts as a true companion by reading the context of the user's current screen.
By understanding specific hotel details,
it provides immediate answers to build user confidence right before booking.
Highlight #3: Flexible Experiences
To support different user behaviors, the side-chat offers a smooth transition to a full-screen immersive mode. This allows the UI to scale dynamically from quick contextual nudges to content-heavy travel planning.
Act 1
Foundation & Scaling
Project Timeline & Evolution
In 2024, I led the design of an AI travel search tool that eventually became a core company feature. We validated its value with a standalone assistant, then expanded it across major travel verticals. Finally, we integrated it into KAYAK’s core platform using a scalable design system.
Problem & Opportunity
KAYAK offers powerful features across the user journey, but they often exist as fragmented touch points. We leveraged AI to connect these dots, creating a seamless experience that drives confidence and conversions.
From that journey mapping, I extracted three core questions that drive every trip: Where to go, When to go, and What to book. To handle open-ended natural language, I mapped how unpredictable prompts connect to KAYAK’s structured live data.
Wireframe: Search Widget
To differentiate from generic LLMs, I designed a modular 'widget'. This interface seamlessly combines real-time KAYAK travel data with conversational AI responses.
Scaling: Atomic UI System
Standardized with a fixed header and flexible body, this modular framework accelerated our shipping velocity across 10+ travel verticals.
Business Signal
A 40% conversion lift among AI users provided a strong business signal, becoming the catalyst for KAYAK’s company-wide AI roadmap.
Act 2
The Pivot: From Destination to Companion
Uncovering the Growth Ceiling
Despite strong conversion signals, testing revealed three roadblocks to mass adoption: transparency, accessibility, and familiarity. Users still needed traditional UI controls to verify AI suggestions.
Transparency: Thinking step & AI snippet
We added a 'thinking step' to reassure users, and injected contextual snippets directly into results to explain why specific options were recommended.
Accessibility: AI-Assistant Gateway
After A/B testing multiple entry points, we implemented a split-view AI gateway for maximum flexibility across the entire user journey.
Architecture Shift: From Destination to Companion
Pivoting from an isolated chat page, we introduced a side-chat split-view. This redefined the AI from a standalone tool into a contextual travel companion.
Familiarity: Enable AI Everywhere
With this architecture change, users can now browse the familiar KAYAK interface while accessing AI on demand, organically multiplying our user engagement touch points.
Act 3
The Expectation Gap
Engagement Bet: The Front-Door Experiment
To maximize engagement, we placed the AI gateway directly on KAYAK's front door page. I built functional prototypes using Claude Code to rapidly validate input models and align with engineering.
The Expectation Gap: AI Capability Overshooting
While open-ended search drove high engagement, failure rates spiked. The blank interface inadvertently shifted user expectations from simple search to complex travel delegation.
Solution #1: Calibrating the Front Door
To manage expectations, we pivoted to a hybrid structure. We kept the traditional 'Origin' input but added AI suggestions exclusively to the 'Destination' field.
Solution #2: Context-Driven Nudge
This taught us that interaction context dictates the user's mental model. Instead of relying solely on the homepage, we distributed context-driven AI nudges exactly where users experience friction.
Learning: Principles for AI Product Design
Contextual cues drive user intent far better than static forms. Furthermore, establishing clear guardrails for non-deterministic AI is essential for building trust.
Scaling the Team Impact
Beyond product execution, I acted as a multiplier. I hosted 10+ workshops on AI prototyping workflows to accelerate the entire design organization's velocity.
