Back to projects
Product Building Completed

Fliggy AI Trip Designer

An interactive short-trip planning prototype built around AI recommendation, user choice, and system validation.

Context
Product internship extension · Independent prototype
Role
Product definition, interaction structure, and prototype development
Scope
2–4 day domestic trips · Mobile web

This is not a tool that generates a complete itinerary and asks the traveller to accept it. It explores a different question: when AI participates in travel decisions, how can the user retain agency while the plan remains executable?

The problem

Short-trip planning is fragmented across guides, maps, booking platforms, and accommodation services. A one-shot AI itinerary may look complete, yet remain difficult to revise and weak on budget, route, opening-hour, and reservation constraints.

The prototype therefore treats planning as a sequence of decisions that can be understood, revised, and undone.

How it works

It supports both travellers who have not chosen a destination and those who already know where they want to go. The system progressively structures origin, dates, party size, budget, interests, and pace before moving through destination, transport, stay area, and day-by-day planning.

Every stage follows the same principle:

  • AI offers a small set of comparable options;
  • the user accepts, skips, or changes them;
  • the system continuously checks budget, route, time, and intensity;
  • high-impact changes require confirmation and remain reversible;
  • unavailable live information is labelled as demo or estimated data.

A central decision

Budget is not a number that appears only on the result screen. It is a constraint carried through the entire choice architecture. Overspend is graded by severity; a plan cannot be treated as ready when it exceeds the budget by more than 10% or contains unresolved blocking issues.

AI may explain results that have already passed deterministic constraints. It cannot alter the budget rules, restore filtered destinations, or silently replace confirmed choices.

Outcome

The first release produced a complete mobile web prototype, a product requirements document, and a deterministic fallback. Automated checks cover budget conversion, destination scoring, routing, draft recovery, and the principal planning flows.

Open the interactive demo

Boundary

The public demo uses demo and estimated data. It is not an official Fliggy product and does not connect to live inventory, maps, weather, payments, or orders. The mainland China mirror runs the core journey in the browser and is not presented as a live AI service.

Reflection

The most important output is not an itinerary. It is a product position: AI can organise possibilities, but it should not quietly acquire the final decision.