- Industry: Healthcare (US-based telehealth platform)
- Size: 200+ employees
- Challenge: Scaling patient support without proportional headcount growth
A growing US telehealth platform was handling 15,000+ patient inquiries per month across chat, email, and phone. Their support team of 25 agents was overwhelmed. Average response time had climbed to 4 hours, patient satisfaction scores were declining, and the cost per support interaction was $18 — unsustainable as the patient base grew 30% quarter over quarter.
QUYNT’s AI chatbot didn’t just reduce our costs — it actually improved the quality of support our patients receive. The AI handles routine questions instantly, and our agents can now focus on the patients who really need a human touch.
The company needed to dramatically reduce support volume without sacrificing quality. Previous attempts with rules-based chatbots failed because healthcare inquiries are nuanced — patients ask complex questions about medications, insurance coverage, appointment scheduling, and symptoms that require contextual understanding and empathy.
QUYNT designed and deployed a generative AI chatbot trained on the platform’s knowledge base, clinical guidelines, insurance policies, and 2 years of historical support transcripts. The system uses RAG architecture to provide accurate, sourced answers and includes safety guardrails that automatically escalate clinical questions to licensed professionals.
- Phase 1 (Weeks 1–3): Data preparation, knowledge base construction, and safety guardrail design.
- Phase 2 (Weeks 4–6): Model training, RAG pipeline development, and integration with the telehealth platform.
- Phase 3 (Weeks 7–8): Pilot with 10% of traffic, human review of all AI responses, and iterative refinement.
- Phase 4 (Weeks 9–12): Full rollout, agent training on AI-assisted workflows, and continuous optimization.
- 52% reduction in support costs (cost per interaction dropped from $18 to $8.60)
- 70% of inquiries handled autonomously by the AI chatbot
- Average response time reduced from 4 hours to 12 seconds for AI-handled queries
- Patient satisfaction increased from 3.6 to 4.4 out of 5
- Support team refocused on complex cases, improving job satisfaction and retention
