Quick overview
An AI-powered WhatsApp assistant for real estate teams. Answers property questions
from a live Airtable database, qualifies leads as hot, warm, or cold, checks
calendar availability before booking viewings, and escalates to a human agent
when a lead is ready to move.
How it works
- WhatsApp Trigger receives a text or voice message from the customer
- The message is validated and routed by type, text goes straight through, voice notes go to transcription first
- Voice notes are downloaded and transcribed using Google Gemini before reaching the agent
- The AI agent processes the message, using Airtable to look up real property data and Google Calendar to check availability
- A structured output parser returns the reply along with the lead status, hot, warm, or cold, and whether escalation is needed
- The reply is sent back to the customer and the lead is logged to Google Sheets
- If the lead is hot or needs a human, the agent is notified, with duplicate alerts prevented. If the AI agent fails, the customer still gets a fallback reply and the developer is alerted separately
Setup
- Add your WhatsApp Business Cloud API credentials to the WhatsApp Trigger, Send WhatsApp Reply, Notify Agent, and Send Unsupported Notice nodes
- Add your Google Gemini API credentials to the Gemini Chat Model and Analyze Audio nodes
- Connect your Airtable account and select your Properties base and table in the search_properties tool
- Connect Google Calendar and select your calendar in check_availability and book_appointment
- Connect Google Sheets and select or create a Leads sheet in Log Lead to Sheet
- Replace the placeholder phone numbers in Notify Agent and Notify Developer with your own
Requirements
- WhatsApp Business Cloud API access
- Google Gemini API key
- Airtable base with a Properties table
- Google Calendar account
- Google Sheets account
Customization
- Update the knowledge base section of the system prompt to match your own region or business rules
- Adjust business hours and viewing duration inside the calendar tool descriptions
- Swap Airtable for a different database if preferred, keeping the same tool description logic
Additional info
Session memory is in-memory and resets if the n8n instance restarts. For production use at scale, consider a persistent memory store like Redis.