Secretaria — AI Front Desk for Clinics
Customer service automation for clinics: an AI agent on WhatsApp that handles appointment scheduling, patient questions, and day-to-day front-desk work.
About the project
Clinics spend a large part of the day answering repetitive messages: scheduling, rescheduling, confirming appointments, and clarifying questions about procedures and preparation. The front-desk team becomes a bottleneck, patients wait, and no-shows go unmanaged.
Secretaria was born to automate this routine through the channel patients already use every day: WhatsApp.
What it does
An AI agent talks to patients on WhatsApp in natural language: it books, reschedules, and confirms appointments directly against the clinic's calendar, answers frequently asked questions grounded in the clinic's own information, and sends reminders to reduce no-shows.
When a conversation requires a human — a sensitive case or an unusual request — the agent hands it off to the team with the full context summarized, so the patient never has to repeat themselves.
How it was built
The agent is built in Python with FastAPI, connected to the WhatsApp API. The LLM orchestration layer gives the agent tools to act on the clinic's scheduling system — checking availability, creating and updating appointments — always with validation steps before confirming anything with the patient.
Clinic-specific knowledge (procedures, preparation instructions, policies) is served through a RAG pipeline, so answers stay grounded in official information instead of model guesses. Every conversation is logged and monitored, feeding continuous improvements to prompts and flows.