Radha Healthcare Chatbot

The problem

Radha operated four primary care locations in the US Southeast. Their front desk teams were overwhelmed with inbound calls — the majority asking repetitive questions about appointment availability, insurance acceptance, preparation for upcoming visits, and post-visit prescription status. Clinical staff were being interrupted for questions that didn't require clinical judgment.

They needed to deflect the high-volume, low-complexity interactions to an automated system, freeing front desk staff to focus on in-person patient experience and the calls that genuinely required human judgment. Patient data sensitivity and HIPAA considerations shaped every design decision.

What we built

An AI chatbot deployed on their website and as an SMS channel, trained on their specific knowledge base: insurance panels accepted at each location, appointment type preparation guides, post-visit FAQ responses, and general practice information. The chatbot handles intake pre-screening questions, appointment-related FAQs, and post-visit follow-up prompts.

We were deliberate about what the chatbot never touches: it cannot access patient records, cannot confirm specific diagnoses or prescriptions, and escalates immediately to a human for anything clinical. The escalation path routes to staff via secure messaging with full conversation context, so patients don't repeat themselves.

All conversation data is stored with encryption at rest and in transit. No PHI is processed by third-party AI models — we used a self-hosted LLM for the deployed chatbot to meet their compliance requirements.

Technology

• Self-hosted Mistral LLM (HIPAA-conscious deployment)
• RAG knowledge base built on clinic-specific documentation
• Web chat widget (custom React component)
• SMS channel via Twilio
• Escalation routing to staff via encrypted internal messaging
• All data encrypted at rest (AES-256) and in transit (TLS 1.3)

Results

45% reduction in front desk inbound call volume within 60 days. Patient satisfaction with the chatbot interaction rated 4.2/5 in post-interaction surveys. Front desk staff redirected time to in-clinic patient experience and complex scheduling coordination. Zero HIPAA incidents in the first year of operation.