Detour
AI travel itinerary orchestration
Kevin Gao · Sole Product Designer · Owns end-to-end · 5 weeks
PROBLEM
Multiple apps.
One cancelled train.
No answers.
Illustration:
6 app icons on phone
You're in a ryokan lobby in Hakone. Your Shinkansen just got cancelled.
PROBLEM
The real unit of disruption
isn't a single booking.
Bookings have dependencies, but no product
manages the relationship layer.
RESEARCH
Diary Study
6 participants · Multi-city travelers · Recent trips
2–3
daily adjustments
on average
"I didn't realize how often I was adjusting until I wrote it all down."
"Most changes were small. 10 minutes here, swap this for that."
RESEARCH
Critical Incident Interviews
5 participants · Major travel disruptions
The hardest part wasn't finding alternatives.
It was not knowing what else would break.
"I rebooked the flight, then realized my airport pickup and first night hotel were both wrong now."
"I spent two hours fixing things that broke because of one delay. Each fix created another problem."
RESEARCH
Competitive Experience Audit
TripIt
Aggregates data,
no intelligence
Google Calendar
Shows events,
no relationships
Expedia Romie
AI assistant,
only Expedia bookings
No product orchestrates across all your bookings.
From scattered points to a connected graph
Flight
Hotel
Train
Activity
Restaurant
→
Flight
Hotel
Train
Activity
Restaurant
Existing tools treat bookings as isolated points. Travelers think in connected systems.
Detour
Understands relationships between bookings.
Re-orchestrates when things change.
Stress test
→
Daily adjustments
→
Disruption recovery
Target user: Self-planned multi-city travelers
DESIGN DECISIONS
Exploring interaction directions
direction-exploration.html screenshot
Solution-first cards won: fastest path to decision under stress.
DESIGN DECISIONS
Three tiers of delegation
Trust is built through transparency, not settings.
DESIGN DECISIONS
Key tradeoffs
Could have�
Minimal pre-trip friction
→
Chose
Proactive stress testing.
2 min before saves 2 hours during.
Could have�
Show all details upfront
→
Chose
Solution first, details on demand.
Speed to decision under stress.
Design
9 screens · 2 prototypes · 1 orchestration engine
Onboarding: Connect and go
D0 Phase 1
Connect email
D0 Phase 2
Scanning
D0 Phase 3
Trip found
No registration. Email = identity. Missing bookings can be added later.
Pre-departure: See your trip as a system
D1
Trip Overview
D2
Conflict Detail
Stress test catches issues before they become emergencies.
On the road: Daily view and gentle nudges
D3
Day View
D4
Weather Adjustment
Same engine, lighter touch.
When things break: Impact analysis
D5
Notification
D5
Impact List
6 downstream plans affected by one cancelled train.
Recovery: Two paths forward
D6
Plan A
D6
Plan B
Must-haves preserved in both plans. Cost difference: ¥6,800 vs ¥18,500.
Execution and resolution
D7
Execution
D8
Updated Itinerary
Three tiers in action: auto-searched, confirm purchase, draft your message.
See it in action
Prototype A
Disruption recovery flow
D5 → D5/D6 → D7 → D8
Prototype B
Pre-departure setup
D0 → D1 → D2
Interactive prototypes Link
VALIDATION
Scenario-based usability testing
5 participants · Full Prototype A flow · Think-aloud
Finding 1 4 of 5 users navigated back from impact analysis to recovery plans at least once. Average time from disruption alert to plan selection: 58 seconds.
Finding 2 3 of 5 users tapped completed actions to verify they actually happened. 2 users asked "how do I know the tea ceremony was really rescheduled?"
After iteration: plan selection time dropped to 31 seconds. 0 of 5 users questioned completed actions.
ITERATION
Solution first, details on demand
→
Plans come first. Impact analysis becomes optional expansion.
ITERATION
Making automation visible
→
After
V2 D7
(live execution)
Users watch actions complete one by one. Trust through visibility.
OUTCOMES
Tying it back
31s
disruption to decision
Average time from disruption alert to plan selection, down from 58 seconds before iteration. Users stopped bouncing between impact analysis and recovery plans.
← No answers
0 / 5
questioned automation
After making completed actions visually trackable, zero users asked whether rebookings actually happened. Trust was built through visibility, not settings.
← Multiple apps
5 / 5
completed full recovery
All participants navigated from disruption notification through plan selection to execution without external tools. One product handled what previously required six.
← One cancelled train
What's next: Traveling together
See the same thing
Shared trip view via invite link. Both travelers see disruptions and recovery plans in real time.
Protect what matters to each of you
Each traveler marks their own must-haves. Detour considers both when generating recovery plans.
Decide together
Both travelers react to plans. Aligned? One-tap execute. Disagreed? Detour highlights the tradeoff.
No app download required for co-travelers. Web-based shared view.
Same scenario,
different experience
Your Shinkansen is cancelled. But you don't open six apps.
Detour already analyzed the impact, generated two plans, and started executing.
You confirm two tickets and one hotel message. Done.
Thank you
Kevin Gao
gaoux.com · huawengao@email.com