Inspiration

Booking flights online often feels frustrating. Travelers usually know what they want, but traditional travel websites force them through complex forms, filters, and confusing options. This leads to abandoned searches and poor user experience.

We were inspired to build a conversational AI travel agent that allows users to book flights naturally—by simply chatting, just like talking to a travel assistant.

What it does

Our Conversational AI Travel Agent lets users search and compare flights using natural language. Users can say things like “I want to fly from Seattle to Miami next month for around $400”.

The system understands the intent, extracts key details, fetches matching flights, and presents the best options in a friendly, conversational format—making booking faster and more personal.

How we built it

We used Vertex AI Agent Builder to understand user intent and extract entities like location, dates, and budget. A backend service running on Cloud Run connects to flight booking APIs to retrieve available options. Gemini then converts the raw flight data into clear, conversational responses that help users choose easily.

Challenges we ran into

Designing natural conversation flows instead of form-based logic

Handling incomplete or vague user inputs

Presenting multiple flight options clearly without overwhelming users

Accomplishments that we're proud of

Built a working conversational booking flow

Reduced complex flight search into a simple chat experience

Successfully integrated intent understanding with real-time data

What we learned

We learned how conversational AI can dramatically improve user experience. We also gained hands-on experience with Vertex AI, Gemini prompting, and cloud-based backend design.

What's next for Build a conversational AI travel agent

Next, we plan to add hotel and package bookings, real-time price alerts, user profiles, and voice-based interaction for an even more seamless travel experience.

Built With

  • apis
  • cloud-services
  • databases
  • frameworks
  • platforms
Share this project:

Updates