Multichannel AI Receptionist

AI front desk for real-estate agencies in Brazil, over WhatsApp and Telegram. Multi-tenant architecture across five n8n workflows on PostgreSQL, running on my own Oracle Cloud ARM server.

My role

Author: architecture, n8n workflows, database and server.

The challenge

Real-estate agencies get questions over WhatsApp and Telegram at any hour. Answering fast, understanding what each person is looking for and booking visits takes a big part of the team’s day.

The solution

An AI assistant that talks to the agency’s clients, qualifies their interest and helps book visits. The architecture is multi-tenant: one installation serves several agencies, each with its own configuration stored in PostgreSQL.

How it works

• ›Ingestion: receives WhatsApp and Telegram messages in a single format.
• ›Brain: the AI model understands the conversation and decides the reply.
• ›Actions: does what the conversation asks for, such as saving the lead.
• ›Output: sends the reply back on the same channel.
• ›Operations: follow-up and maintenance routines.
• ›Five n8n workflows running on my own Oracle Cloud ARM server.

FLOW

STACK