Too many tabs
Travelers switch between maps, weather, blogs, and booking tools.
Case study
Sherpa AI Travel Co-Pilot.
An AI travel assistant that turns vague trip intent into contextual itineraries.
Plan smarter trips with LLM reasoning, location context, and weather-aware recommendations.
Role: Product, Design, BuildStack: Next.js, Claude API, Google APIs, OpenWeatherStatus: Live productFocus: AI assistant UX, API orchestration, travel planning workflows
Plan your trip
Skip the 14
browser tabs.
browser tabs.
Where to?
e.g., Lisbon, Portugal
FoodieHistory nerdNature seeker
The Problem
Travel planning is fragmented.
People jump between search, maps, weather, blogs, notes, and group chats before they get to a usable plan.
Generic recommendations
Most suggestions ignore trip intent, timing, weather, and personal context.
Planning lacks flow
Users move from inspiration to itinerary through a messy, manual process.
Target User
Built for independent travelers
Sherpa is designed for people who know roughly where they want to go, but need help turning intent into a useful plan.
01
Explore quickly
User need
Explore a destination quickly
Sherpa response
AI-generated trip overview
02
Plan with context
User need
Plan around time and weather
Sherpa response
Weather-aware itinerary logic
03
Find relevant places
User need
Find relevant places
Sherpa response
Google Places and Maps context
04
Avoid generic lists
User need
Avoid generic lists
Sherpa response
Intent-aware recommendations
05
Move into a plan
User need
Move from idea to plan
Sherpa response
Structured itinerary output
Core Flow
From vague idea to usable itinerary
The product flow turns loose travel intent into a structured plan with contextual recommendations.
Step 1
Trip Intent
Step 2
Destination Context
Step 3
Weather Check
Step 4
AI Reasoning
Step 5
Itinerary
Step 6
Refine Plan
Product Decisions
What shaped the experience
Four product decisions made Sherpa feel more like a co-pilot than a static itinerary tool.
Decision 01
Assistant-first
Start with intent, not forms. The product asks for enough structure without turning planning into admin.
Decision 02
Context layer
Combine maps, weather, and LLM reasoning so recommendations respond to place, timing, and user style.
Decision 03
Output design
Make plans structured, not chatty. Users need a usable itinerary, not a long assistant monologue.
Decision 04
Low friction
No login-first experience. Let the traveler test planning quality before asking for commitment.
Metrics
What I'd measure
A simple product health framework across plan quality, speed, refinement, usefulness, and repeat planning.
01
ActivationItinerary generation completion rate
Why it matters
Shows whether users can move from trip intent to a generated plan.
02
Speed to valueTime to first useful plan
Why it matters
Measures how quickly Sherpa delivers something the user can act on.
03
RefinementRegeneration or edit rate
Why it matters
Shows whether users are shaping the output instead of abandoning it.
04
UtilitySaved or shared itinerary rate
Why it matters
Measures whether the plan is useful enough to keep or send to others.
05
RetentionRepeat planning sessions
Why it matters
Shows whether users return for future trips or planning iterations.
Tradeoffs
What I intentionally did not build
The MVP stays focused on planning quality before expanding into a full travel platform.
No booking layer
Kept the MVP focused on decision quality before adding transactional complexity.
No heavy onboarding
Travelers can test value quickly without building a profile first.
No social/group planning yet
Collaboration matters later, but the first wedge is one traveler reaching a usable plan.
What's Next
Where Sherpa could go next
The next layer is about memory, collaboration, and agentic trip execution.
Now
AI itinerary generation with destination, weather, and location context
Prove that Sherpa can turn loose intent into a useful first plan.
Next
Saved trips, user preferences, editable itinerary blocks, and better recommendation controls
Give users more memory and control without bloating the first-run experience.
Later
Group planning, budget-aware routing, booking handoffs, and agentic travel workflows
Move from planning assistant to travel execution layer.
Try it