Tour Booking Chatbot (WhatsApp + n8n)
24/7 support and bookings with accurate pricing — based on real variables (boat, season, experience, duration and guests).
The Problem (The Gap)
Tour and experience companies lose revenue when users ask for price and availability and don’t get an immediate answer. The real challenge isn’t “chatting well” — it’s computing the correct price every time, because it changes by boat, season, experience, duration and number of guests.
The Engineering (The Tech)
We designed this system as a reliable agent: instead of throwing a spreadsheet into a prompt, we turned pricing rules into deterministic tools (code) and used AI only to understand intent, collect variables, and drive the conversation.
Users ask for a price on WhatsApp → the bot collects the minimum variables → validates available durations → returns the exact amount and guides the user to booking. Less friction, fewer mistakes, more conversions — without relying on human support.
What it does
- 24/7 WhatsApp support (text and voice)
- Understands intent: pricing, questions, booking, changes
- Collects variables (experience, boat, date, guests, duration)
- Validates duration availability for that experience/date
- Calculates the exact price using the client’s rule set (no guesswork)
- Logs leads and conversation into a spreadsheet
How it works
- Receives a WhatsApp message (text or voice)
- Normalizes input and aggregates messages to avoid duplicate replies
- Agent guides the flow and collects required variables
- Deterministic tools: season check, duration listing, price calculation
- Final response is “humanized” and split into WhatsApp-friendly chunks
Pillars
Zero-guess pricing (deterministic)
Pricing must be exact. Rules become code and the agent only answers when all required variables are present.
- Variables: boat, season, experience, duration, guests
- Validation: only allows durations that truly exist
- Always exact (no hallucinations)
Robust message pipeline
Fixes the classic WhatsApp issue: users send 3 messages and the bot replies 3 times.
- Message aggregation with a wait window
- Short-term context memory
- Chunked delivery for better UX
Scalable architecture
Modular, step-based workflow with isolated tools — easy to adapt to other booking businesses.
- Input: text + voice (transcription)
- Tools: season / durations / pricing
- Logs and lead capture
Security & compliance
- Conversation data can be minimized and stored with controls
- Deterministic tools reduce hallucination risk for pricing
- Separation of concerns: AI handles dialogue; code handles numbers