Development of Ayre: An AI Travel Planning Agent
thesyphon.xyz
Ayre - Conversational Agent that helps you discover & plan your next vacation.
Ayre is a conversational travel agent that helps people turn loose travel ideas into personalised trip plans through natural dialogue — making vacation planning feel more like talking to an expert than using a booking tool.
Results🚀
Designed and built a Gemini-powered conversational travel agent
Developed end-to-end trip planning flows from discovery to email delivery
Defined agent behavior, prompt architecture, and human escalation logic
Prototyped an AI-native travel planning system using Figma, Cursor, and Next.js
Ayre landing page with nudges to start a trip plan
The problem
People plan travel through conversation & half-formed ideas, but most travel tools are built around forms and filters.
Travellers often arrive with vague ideas, evolving preferences, and a need for guidance, while operators need those conversations to translate into real availability, qualified leads, and bookings. This creates a gap between how people naturally plan trips and how travel systems are designed to support them.
The Solution
Conversational AI that reads intent
Ayre is a conversational travel planning agent designed to make discovering your next vacation feel like speaking to an expert, not using a booking tool.
Through guided dialogue, it helps travellers move from vague ideas to personalised itineraries, while turning each conversation into structured, bookable intent.
Decide through dialogue your vague travel idea
Using Gemini-powered dialogue, a structured TripDraft system, email summaries via Resend, and escalation paths for edge cases, Ayre helps travellers go from vague intent to personalised vacation plans without navigating traditional booking funnels. The result is a planning experience that feels consultative on the front end and operationally grounded on the back end.
The Agent Design
Conversational discovery → understand traveller intent
Trip drafting engine → structure dynamic itineraries
Delivery layer → send curated plans via email
Human escalation layer → route complex cases seamlessly
Design Methodology
Conversation Design × Service Design × Human-in-the-Loop AI
Ayre was designed through an intent-first methodology that combined conversational UX, structured AI orchestration, progressive disclosure, and service design thinking — treating travel planning not as a chatbot flow, but as an end-to-end advisory experience.
Conversational interface that allows users to iterate
Transparency in Agent steps
Agent builds user responses into structured plans
Agent builds the travel itinerary tailored to the users needs
The agent stores chats itineraries for users to revisit
Design System
The design system was built on Figma with a detailed visual style guide & component guidelines.
Figma UI flow
Design System documentation
AI-Native Build Process
Stack: Figma · Cursor · Next.js · Gemini · Claude
Ayre was designed and built as an end-to-end prototype, moving from conversational UX exploration to a functioning AI travel planning agent.
1. Prompt & Agent Architecture
Before building the product logic, we defined Ayre’s role as a travel advisor, using prompt architecture to shape tone, behaviour, decision logic, and how the agent gathers preferences, recommends destinations, and handles edge cases.
2. AI-Assisted Product Development
Using Cursor as a development copilot, we translated the concept into a working Next.js application — building the chat interface, trip planning logic, email flows, and escalation pathways while iterating directly through code.
3. LLM Orchestration
With the Gemini API, we connected the conversational reasoning layer, enabling the agent to interpret intent, ask contextual follow-ups, and generate structured travel plans from natural dialogue.
4. Prototype to Functional Agent
The result was more than a prototype — a functioning AI agent system where UX, prompts, model behaviour, and product logic were designed in parallel.
End-to-end MVP flow
Future Direction
Future iterations would extend beyond trip discovery into deeper booking intelligence, operational integrations, and assisted conversion.
Intent-aware travel advising
Further evolve Ayre’s conversational intelligence to understand guest intent beyond keywords — interpreting nuance, preferences, and evolving travel needs through more natural dialogue.
Real-time inventory and pricing
Integrate Ayre with Property Management Systems (PMS) and supplier inventory to surface live availability, dynamic pricing, and instant travel quotes directly within the conversation.
CRM-powered lead orchestration
Connect conversations to HubSpot to automatically generate leads, contacts, and deal pipelines from qualified travel inquiries, turning planning conversations into actionable demand.
Assisted booking flows
Extend Ayre from trip planning into conversion by enabling guided self-booking experiences — or allowing the agent to initiate bookings on behalf of travellers when appropriate.
Human-in-the-loop escalation
Deepen support handoff capabilities so complex or high-touch requests can be routed to human agents with full conversation memory and trip context preserved.
Toward an autonomous travel agent
Longer term, evolve Ayre from a conversational assistant to an agentic travel system — capable of discovery, planning, booking coordination, and support within one continuous experience.
Like this project
Posted Apr 29, 2026
Ayre, is an AI travel agent for personalised trip planning. Prototyped an AI-native travel planning system using Figma, Cursor, and Next.js
