tRAVEL TECH

B2B2B

booking Platform redesign

AI Sales agent

TravelSales: From a 2-day phone quote to booking in one session

TravelSales: From a 2-day phone quote to booking in one session

The only company that installs AI sales agents directly for bus operators, gives operators a real-time booking and inventory platform, and lets customers choose their exact unit with instant pricing.

TIMELINE

2026 Jun – 2026 Sep

ROLE

Product Designer

SKILLS

Claude Design, Figma, Lovable

Scope

  1. Operator calendar and quotation page

  2. Travel agent booking flow

TEAM

Founder, Product Manager, Software Architect, AI Specialist, Product Analyst

LONG STORY SHORT

Turned a confusing MVP into a shipped booking flow for vehicle operators and travel agents

Turned a confusing MVP into a shipped booking flow for vehicle operators and travel agents

Turned a confusing MVP into a shipped booking flow for vehicle operators and travel agents

I redesigned three broken operator tools and built a new booking form for travel agents. All of it shipped.

For operators, the calendar hid times and status, the booking form buried required fields across five tabs, and the quotation page changed depending on how you opened it.

01 Research

Studied the pain points and motivations of vehicle operators and travel agents and how they interact with each other in the ecosystem.

02 Collaboration

Worked closely with the founder and product manager to define product strategy, the initial PRD, & feature priorities.

03 Design

Used AI tools including Claude Design and Lovable to generate prototypes and iterate based on leadership and user feedback.

BUSINESS overview

Helping small & medium family transport businesses survive

Helping small & medium family transport businesses survive

Helping small & medium family transport businesses survive

There are four foundational areas to understand before diving into the designs.

What the business does

Connects Latin American charter bus operators with travel agencies through AI-powered booking

Revenue model

A single-digit commission per completed booking, with no subscription

Why it exists

A few big platforms dominate the market, leaving family operators without the tools to compete

What Maya AI Agent changed

Days of WhatsApp and manual pricing became one AI-captured, priced request.

UNDERSTANDING THE USERS

Accuracy, speed, & transparency are the traditional pain points

Accuracy, speed, & transparency are the traditional pain points

Accuracy, speed, & transparency are the traditional pain points

The founder has a decade of experience in the charter transport industry, and he offered insights about how TravelSales can help a family transport company in Latin America compete with the same tools as multinationals.

Calling operator

no record, no trace

Jun 17

47 pax

Hotel Reforma

1) The call

The agent calls and reads out address, dates, and passenger count.

?

2) The notes

The operator writes it by hand. Accuracy depends on note-taking discipline.

Need a price for Jun 17

1–2 days of silence

3) The wait

One to two days later, the operator calls back with a hand-calculated quote.

agent books elsewhere

4) The drop

Conversion drops. The agent already has alternatives or hesitates on price.

OPPORTUNITIES

Two users, one trip, different mental models

Two users, one trip, different mental models

Two users, one trip, different mental models

Every booking moves from the tour group to the operator, and the two people I designed for read the same trip in opposite ways.

End customers

End customers

Need group transport

Travel agents

Travel agents

Purchase on behalf of the tour group

Maya AI, TravelSales

Maya AI, TravelSales

Checks availability, prices, and routes to operators

Vehicle operators

Vehicle operators

Approve, block, & dispatch

Bus operators

Bus operators

Travel agents

Travel agents

Think in vehicles and dates

Think in groups and routes

Value their long term reputation over raw profits

More direct interactions with end customers

Own the fleet and the risk

Always rushed, no time to type line by line

Plan multi-day, multi-stop trips

Think in vehicles and dates

Think in groups and routes

Value their long term reputation over raw profits

More direct interactions with end customers

Own the fleet and the risk

Always rushed, no time to type line by line

Plan multi-day, multi-stop trips

💭

How might we help vehicle operators turn a booking request into a quote within a minute, and help travel agents submit booking requests without hesitation?

How might we help vehicle operators turn a booking request into a quote within a minute, and help travel agents submit booking requests without hesitation?

Tension

Where I pushed back, and why

Where I pushed back, and why

Where I pushed back, and why

The hardest calls came from stakeholders pulling in different directions.

Availability calendar for operators

Product manager: "Why not add a daily view on the calendar, in addition to the weekly view?"

My call: Ship the weekly view first. Operators care most about how full their fleet is, so the daily view moved to the roadmap.

Manual booking form for travel agents

Travel agent: "Not sure how I can get started with the AI sales agent."


Founder: "AI agent is our core feature, and we have to save it no matter what."

My call: Add a form next to AI, not instead of it. AI stays first, traditional agents get a familiar way in, and form bookings cut API costs down.

COMPETITIVE ANALYSIS

Competitor UI for the operator calendar

Competitor UI for the operator calendar

Competitor UI for the operator calendar

Competitor research surfaced a requirement I'd missed: Busify builds a cleaning buffer into every booking, while Betterbus showed how to handle long trip details in limited space. Both fed directly into the design iterations.

Busify (US-based): Cleaning period at the end of each booking

Betterbus (Germany-based): Fitting long trip details into tight spaces

DESIGN PROCESS

Stress-testing design variations based on founder and PM inputs

Stress-testing design variations based on founder and PM inputs

Stress-testing design variations based on founder and PM inputs

Every screen went through multiple rounds, cutting steps for operators and friction for travel agents.

FOR BUS OPERATORS

FOR BUS OPERATORS

FOR BUS OPERATORS

Booking blocks on calendar overview

  • Smaller time fragments were added to each daily block to accommodate bookings that span multiple days, e.g. overnight trips

  • Key details on every block: quote number, times, status, and cleaning buffer, with anything cut off shown on hover

Create new availability

  • One screen: dates, customer, and unit are all it takes to create a quote

  • Progressive disclosure: route, drivers, and costs move to the trip page

  • Less input: dates from the calendar arrive prefilled, and dropdowns replace long lists

Quotation page

  • Stage-aware actions: block unit leads until the bus is secured, then the magic link takes over

  • Built to close: every paid magic link earns TravelSales a commission, so once the unit is blocked, it's the most prominent action on the page

  • One trip page: the quote looks the same no matter where operators open it

FOR TRAVEL AGENTS

FOR TRAVEL AGENTS

FOR TRAVEL AGENTS

New feature: a manual form alongside the existing AI booking feature

  • Consolidated entry point: Maya AI agent and the manual form share one landing screen

  • Choice, not a chat: agents who'd rather not talk to an AI switch to a form in one tap

Manual booking form that handles both direct and multi-stop trips

  • One form for every trip: "Add a stop" turns a direct trip into a multi-stop itinerary

  • Less repetition: round trip is a single toggle, and passenger count is set once

  • Trip at a glance: drive and arrival times update as stops are added

FINAL DESIGNS

Shipped, end to end

Shipped, end to end

Shipped, end to end

From an agent's booking request to an operator's confirmed booking, every step shipped.

Artifact 01

How operators create availability on the calendar

Operators now create availability directly on the weekly calendar, one row per vehicle, and see the whole fleet at a glance. Dates, customer, and unit are all it takes to create a quote. Dates from the calendar arrive prefilled, and dropdowns replace long lists. What took a few minutes takes under one.

Artifact 02

Hover state on calendar

When a booking is too short to fit its details, hovering reveals the quote number, times, and status.

Artifact 03

How operators handle quotation from the Maya AI agent

Blocking the unit comes first, then the magic link takes over. Every paid magic link earns TravelSales a commission, so once the unit is blocked, it's the most prominent action on the page. Operators can still manually edit the AI-generated cost before it's confirmed.

Artifact 04

How travel agents submit a booking request via manual form when they choose not to use the AI agent

Agents enter pickup and drop-off, add stops anywhere along the route, and price the whole trip in one tap.

ouTCOMES

Shipped, and simpler by design

Shipped, and simpler by design

Shipped, and simpler by design

The operator calendar, quotation page, and agent booking form all shipped in early fall 2026.

5

1

Tabs to create a quote, now one screen with four required fields, so operators can respond before agents look elsewhere

1

2

ways for agent to book, so traditional agents stay in the funnel instead of picking up the phone

3 in 1

Direct, round, and multi-stop trips handled in a single form, so travel agents never pick the wrong one

LEARNINGS & REFLECTION

What I took away

What I took away

What I took away

🔍

Protect the differentiator, then listen

AI sales agent stays the first thing agents see, but traditional agents who prefer forms still have a way in.

🔄

Design for users' main motivations

The PM wanted a daily view for precise booking times, but operators mostly ask which buses are free in a given week. Anchoring the debate in that question made the weekly view an easy call instead of a standoff.

⚖️

Shared elements make two platforms feel like one product

Every quote on the operator side shows whether it comes from the AI agent, so both platforms read as one connected system.

Matthew Liang

Product Designer