Y2Z Travel
Role
Founding Product Designer
Team
1 Designer · 2 Engineers · CEO
Duration
8 months
Platform
PWA

Overview
Y2Z Travel is a bootstrapped AI travel planning platform built to help travellers research, personalise, plan and collaborate on trips in real time. We replaced the exhausting, multi-week research burden of modern travel planning, spread across 7+ disconnected tools, with one intelligent product that plans and creates an itinerary the way a great travel agent or an experienced traveler would, asking the right questions, sequencing the logic, and generating highly customisable, personalised itineraries instantly.
As Founding Product Designer, I worked alongside the CEO on the 0-to-1 roadmap, leading everything from early field research to design and GTM strategy while balancing bootstrapped constraints.
Background
Every piece of the trip, ideas, notes, tickets and expenses lived in a different app. What we were missing was traceability. Notes tied to a place. Research tied to a day. Context we could track without losing it or hunting across a dozen tools.
We suspected the setup was bad. The data proved it. Our app activity logs showed that a single day of trip planning involved over 220 instances of opening travel-related applications, constant context-switching across a dozen apps that exposed the real gap.
We researched what tools people currently use to plan journeys, and tried the fixes ourselves: Stippl, Notion and our own Sheets hack.
Every option forced the same trade. Usability or customisation, never both. Stippl sat closest to usability. No tool occupied the zone we needed, the Ideal Zone.

Four tiers, each testing the same question from a different angle: social signals, macro data, field immersion, and synthesis, so no single source could carry a false signal on its own.
This research was compiled from 50+ traveler interviews, 30+ industry reports, financial market reports on travel, conversations with travel agents and analysis across online communities.
Research & Discovery
The cognitive load remained high, the traceability stayed low, and the overall experience was exhausting, even with dedicated tools.
Through investigative work, we wanted to know: was this friction unique to us, or were we looking at a systemic failure felt by every traveler? We needed to understand how and why the problem occurred, what effect it had on users, and the magnitude of the opportunity ahead of us.
To find out, we built a multi-tiered discovery framework, inspired by exploratory research, the Double Diamond model and a Triangulation Design Matrix.
Field research, synthesis and interviews surfaced more friction than expected. We mapped 70+ findings it into a sunburst, organising what we found on paper, reports, forum threads, industry data, and tracing each one back to its root.
The Challenge
The multi-tiered framework gave us the answer. Community data surfaced the friction. Industry data told us it was real at scale. The tools already on the market hadn't solved it either. Together, that was enough to act on. It was time to pick the challenges that would solve our tooling needs and the traveler's top issues at once. We prioritized carefully around what people actually needed.
Here's a version folding that in:
As a team, we ran affinity mapping to bring in the human layer, grouping what we heard directly from user interviews and conversations. This helped us surface the most critical problems alongside the initial direction, which collectively shaped the HMW questions shown below.

Challenge 1
Challenge 2
Challenge 3
Process
The three problems weren't separate. They were interconnected. So we started with personas. Understanding who we were building for grounded everything that followed.
We took a first-principles approach, going back to what was true across all three challenges, and used that shared foundation to build a strong user profile. Design, the AI system, and personalization developed in parallel from there. Here's how each one unfolded.
Personas were built grounded in real user data, then the journey was mapped across four stages: inspiration, research, planning and booking.
Storyboard kept us anchored to those needs, ensuring every current and future feature stayed aligned to what users actually needed.
We chose design first, the most feasible starting point to build and develop in parallel. Competitor analysis and MoSCoW shaped what to build, then prototyping with Bolt.
We had built a small community early on, 27 travelers willing to talk about their planning frustrations. That group became our testing ground, shaping the product's direction through several rounds of feedback.
To hit the 90-day fragmentation problem, we chose AI, backed by real signal: 71% of tourism experts agreed AI would transform the industry, and generative AI use for trip inspiration jumped from 50% to 75% among active users. We explored the market, compared models, and tested prompts before landing on a tuned OpenAI API for MVP speed.
Nearly 3 in 5 travelers didn't have a destination in mind when they started. Personalisation had to begin at question one. We studied travel agents' techniques, narrowed 20 candidate questions to 6 that carried 79% of planning confidence, then fed answers into a Variable Prompting Pipeline, tested across variants before finalising.
Early Explorations
Design didn't stop once personas and the process took shape. Three early prototypes gave us a starting point. What follows is a glimpse of that early exploration, on the way to an MVP built to test PMF.
70% of travelers preferred planning on laptop over mobile, so the direction shifted to laptop-first from that point on.

It was time to recreate the experience for a much larger audience than the small community we'd built it with. To move fast at that scale without burning our limited runway on custom components, we chose Soft UI, an existing React UI library we had purchased long ago.
Feedback from the prototype was collected first. We prioritized what could realistically be added, then looked for design references. Maps and Mindtrip came up most often as the apps people already trusted.
That reference shaped our direction: suggestive itineraries, draggable ordering, sharing. We sketched, tested variations, and improvised the suggestive algorithm, then moved into build. We built the working prototype in Replit.
This Changed How It Worked
Designing for travelers is genuinely hard when the product itself keeps changing shape with every conversation. Trial and error wasn't optional, it was how we found out what actually worked.
One traveler put it plainly: plan every stop, time and distance, arrival, check-in, how far next. We rebuilt the card around that: weather, mode of travel, everything an activity actually needs.
This is where the product started to become its own thing. Distinctly ours, evolving from that point on.
Live routing API calls got expensive fast. We shifted to model-based estimates instead, same accuracy, same card, 80% cheaper.
Solution
This is the most satisfying planner we ever built.

An intelligent AI travel planner that works like magic.

Result
Not a guess. A number someone was willing to commit to.
Impact
beta users said they'd pay




