Clinic AI Front Desk

An AI front desk for clinics and appointment based businesses in Nigeria. It answers customer enquiries on the clinic's website any time of the day, saves every lead in a database and alerts the staff on Telegram within about a minute, so no customer is left waiting or forgotten.

The problem it solves

Many clinics lose customers because messages come in at night, on weekends or when the front desk is busy. By the time someone replies, the customer has already gone to another clinic. This system replies in seconds and makes sure the staff know about every new enquiry.

How it works

1. A customer opens the clinic website and chats with the AI front desk.
2. The AI answers questions using only the clinic's own information (treatments, prices, opening hours). It does not make up answers.
3. When the customer is ready, the AI collects their name, phone number, the treatment they want and their preferred time, and saves it in the database.
4. Medical questions are not answered by the AI. They are marked for a staff member to handle.
5. Every minute, a second workflow checks for new leads and sends the staff a Telegram alert with the customer's details.
6. Each lead is alerted only once, so the staff never get the same alert twice.

Tools used

Part	Tool
Automation	n8n (self hosted)
AI model	Google Gemini
Database	PostgreSQL 16
Running everything	Docker Compose on Ubuntu (WSL)
Staff alerts	Telegram Bot API
Public access	Cloudflare Tunnel
Demo website	Netlify

Files in this repo

File	What it is
docker-compose.yml	Starts n8n and PostgreSQL together
ai-front-desk.json	The n8n workflow for the AI chat
staff-alert.json	The n8n workflow that sends Telegram alerts
index.html	The demo clinic website with the chat bubble
notes.md	My learning notes

Passwords and keys are kept in a .env file that is never uploaded to GitHub.

Leads table

Column	Meaning
name, phone	Customer details
treatment, preferred_time	What they want and when
channel	Where the enquiry came from
status	new, or handoff when a staff member must reply
alerted	Whether the staff have been notified
followed_up	Whether the staff have called the customer

What is next

1. Stricter AI rules, so it only saves treatments the clinic actually offers.
2. Move from my laptop to a small cloud server, so it runs all day without my computer.
3. A WhatsApp channel, since that is where most Nigerian customers are.

Built by

Thompson Nduke Dominic, Duke and Demand Studios dukeanddemandstudios.com