Key Capabilities
- Respond to guest inquiries 24/7 across WhatsApp, email, and booking platforms like Airbnb and Booking.com
- Access live reservation data from your PMS — check-in dates, property details, guest history
- Follow custom procedures for scenarios like early check-in requests, maintenance issues, or lost key situations
- Escalate to human operators based on configurable rules — sentiment thresholds, topic restrictions, price limits
- Learn from corrections — edit an AI response in the inbox and the agent adapts its behavior
The Response Pipeline
When a guest message arrives, Cendra’s AI processes it through five stages:1. Guest Identification
Cendra matches the incoming message to a reservation using the guest’s phone number (WhatsApp) or email address. This pulls in check-in dates, property details, guest count, and booking status from your PMS.2. Context Retrieval
The AI searches your property’s knowledge base for information relevant to the guest’s question. If a guest asks about parking, the AI retrieves your parking instructions. If they ask about restaurants, it pulls your local recommendations.3. Guardrail Check
Before generating a response, Cendra checks your configured guardrails:- Tone rules — formal, casual, or match-the-guest style
- Topic restrictions — topics the AI should never handle (e.g., refund requests, complaints)
- Escalation triggers — conditions that require human review (e.g., negative sentiment, requests above a price threshold)
4. Response Generation
The AI composes a response using the guest’s context, relevant knowledge, and your communication style. The response is grounded in your data — Cendra doesn’t hallucinate information it doesn’t have.5. Delivery Routing
Based on your automation level, the response is either:- Sent automatically (Autopilot mode)
- Queued for review (Semi-automated mode)
- Shown as a suggestion (Manual mode)
Agent Configuration
Each AI agent in Cendra is configured through the Agent Hub with these components:| Component | What it controls |
|---|---|
| Knowledge Base | Property information the AI can reference — amenities, rules, procedures, FAQs |
| Guardrails | Safety boundaries — what the AI can and cannot do, tone, restricted topics |
| Escalation Rules | When to hand off to a human — sentiment, topic, time-based, or threshold triggers |
| Labels | Conditional routing — tag conversations for follow-up, prioritization, or team assignment |
| Tools | Actions the AI can take — look up availability, send upsell offers, create tasks |
Training Your AI
Cendra AI agents learn in two ways:From Your PMS Data
When you connect your property management system, Cendra automatically builds a knowledge base from your property details, descriptions, amenities, and booking rules. This gives the AI a strong foundation without any manual work.From Your Corrections
When the AI gets something wrong, you edit the response directly in the inbox. Cendra learns from these corrections:- Edit a response → AI adjusts its behavior for similar questions
- Add knowledge → AI gains new information to reference
- Set a guardrail → AI avoids the mistake in future conversations