The Challenge
A multi-clinic network in Delhi with 12 locations was overwhelmed with repetitive patient queries — appointment bookings, prescription refills, doctor availability checks. Their support team of 8 handled 400+ calls daily, with 60% being routine queries that could be automated.
Our Solution
We built a custom RAG chatbot powered by GPT-4o integrated with their EHR system. The chatbot was trained on their clinic policies, doctor schedules, and medical FAQs. It was deployed on their website, WhatsApp, and an in-clinic kiosk.
How We Built It
- 1
Audited 6 months of support tickets to identify the 10 most common query types (appointments, refills, timings, billing, insurance, etc.) and mapped them to exact data sources.
- 2
Set up a document ingestion pipeline that cleaned and chunked their clinic policies, doctor schedules, insurance forms, and FAQ documents into a vector store.
- 3
Built a RAG chain with GPT-4o that retrieved relevant chunks for each query and generated a response with a citation to the source document.
- 4
Integrated the chatbot with their existing EHR system via a secure API so it could retrieve live appointment slots and confirm bookings in real time.
- 5
Deployed to three channels: the clinic website (embedded widget), WhatsApp Business API, and an in-clinic kiosk with a touch interface.
- 6
Ran a 2-week shadow test where the chatbot handled queries alongside the human team, measuring accuracy and handoff rate before going live.
Integration & Architecture
The chatbot connected to the existing EHR system via a read-only API for appointment slots and a write API for booking confirmations. No patient data was transferred outside the clinic's network — the vector store was deployed on-premise.
The Results
- 70%Fewer Support Tickets
- 4.8★Patient Satisfaction
- 60%Less No-Shows
- ₹8LAnnual Savings