vivo AI Travel Assistant
Transforming unstructured social content into actionable itineraries via AI & Motion.
Project Overview
How might we use system-level AI to bridge the gap between fragmented social inspiration and actionable travel plans?
Travel planning was a broken experience.
The "App Silos" Problem
Users constantly switched between social apps for inspiration and utility apps for booking and maps, leading to high cognitive load and abandoned plans.
The "Screenshot Graveyard"
Users habitually saved ideas as screenshots from Instagram, TikTok, and XiaoHongShu — but this data was trapped and never turned into reality.
User Friction
Manually extracting POI names and addresses from images and videos was tedious and error-prone.
The status quo: users juggle 5+ apps to plan a single trip, losing inspiration along the way.
Since screenshots are the primary way users save inspiration, we made the screenshot gesture the trigger for the entire AI workflow.
Bridging 3rd-party inspiration to 1st-party execution.
The "Zero Input" Strategy
Instead of copy-pasting text, users simply drag content — images or text — to the side. No typing, no switching apps.
System-Level Trigger
Utilized the "Atomic Island" (Origin OS's dynamic island equivalent) as the entry point. It acts as a bridge, allowing users to drop content without interrupting their current browsing flow.
Design Principle
"Bridging the Gap" — connecting 3rd-party inspiration directly to 1st-party execution through the OS layer.
Drag → Drop → AI Analysis → Result. A seamless pipeline from social inspiration to structured itinerary.
Designing trust through fluid motion.
Addressing AI Latency
The GenAI model takes 2–3 seconds to process data. Static loading spinners cause anxiety and erode trust. We designed a fluid morphing animation (Lottie/JSON) for the Atomic Island — a "skeleton loader" that mimics AI thinking, providing visual feedback and masking the delay.
Motion as Communication
Every animation serves a purpose: the Atomic Island expands to signal readiness, pulses during processing, and morphs into results. No decoration for decoration's sake.
Structured output, not just raw answers.
Contextual Header
Destination name (e.g., "Pike Place Market") paired with real-time weather data for immediate context.
Day-by-Day Navigation
Tab-based structure (Day 1, Day 2, Day 3) breaks the itinerary into digestible chunks, reducing cognitive load.
Smart Details
Each POI includes recommended duration (e.g., "15 min"), condition tags (e.g., "Rainy"), and distance from previous stop.
What you see is what you get.
Building User Trust
Users needed to see immediate, tangible results to trust the AI. The WYSIWYG approach was crucial — showing the generated itinerary instantly rather than hiding it behind loading states.
Route Optimization
The system automatically optimizes the sequence of spots based on location proximity and opening hours, reducing total travel time.
Human Control
Users can drag & drop to adjust the itinerary, add or remove spots, and override AI suggestions. The AI proposes; the human disposes.
Impact & Learnings
Steps Reduced
From copy/switch/paste/search to a single drag & drop gesture
Pixel-Perfect Motion
Lottie motion system delivering smooth animations across all vivo devices
Time Saved
Eliminated hours of manual research and organization.
AI Ecosystem Pattern
Established the first cross-app intelligence pattern for Origin OS