Global Clinical Research Organization:
Pioneering Agentic AI for Patient Assistance Programs
Deploying Nova 2 Sonic AI voice agents and automated SMS campaigns to automate patient support, achieving 14% AI containment and an 84.3% quality score from Day 1 with zero Sev1 incidents.
The Challenge
A major clinical research organization wanted to modernise their Patient Assistance Program (PAP) by migrating to Amazon Connect. Their objective was to introduce an AI-first strategy that could automate inbound patient support inquiries and drive proactive specialty pharmacy outreach, all without sacrificing the empathy and quality of human care.
- Patient Assistance Program required AI-first automation without compromising empathy or care quality
- Specialty pharmacy prescription management was entirely manual and resource-intensive
- No automated patient engagement—all outreach was handled by human agents
- Project timeline compressed by one month due to customer mandated deadlines
- Changing and inconsistent test data during build and UAT phases
- Complex provider queries involving multiple patients per call required careful AI handling
- AI response accuracy impacted by background noise and varying call quality
The Solution
CloudInteract engineered the organisation's first Agentic AI workload, implementing Amazon Bedrock, Amazon Q, and cutting-edge Nova 2 Sonic voice bots. The solution autonomously handles inbound prescription queries while executing proactive SMS and voice campaigns—delivering a highly conversational, automated patient support experience.
- Built and deployed 2 Nova 2 Sonic AI voice agents for patient hub and specialty pharmacy
- Integrated Amazon Bedrock and Amazon Q for intelligent, knowledge-driven responses
- Deployed outbound SMS and voice campaigns for proactive patient engagement
- Built SMS bot for inbound prescription refill requests and outbound status updates
- Integrated with the client's Engagement Platform (CRM) and Enterprise Pharmacy System
- Delivered Apollo by CloudInteract for AI call analytics, containment tracking, and QM scoring
- Achieved 14% AI containment from Day 1 with a clear optimisation path identified
- Client empowered to control AI knowledge base independently via amendable articles
Why CloudInteract
CloudInteract took a risk-aware, delivery-led approach. Extensive internal testing across the company identified and remediated issues before external UAT, resulting in very few failures during client testing. A formal Risk Acceptance document addressed AI's inherent limitations and helped the client understand and accept the technology. The project was re-baselined and critical scope decoupled when the timeline was compressed by one month.
Project Timeline
Analysed common patient queries via Apollo CX Insight and built knowledge base architecture
Built conversational IVR, integrated CRM & pharmacy systems, and deployed Nova 2 Sonic agents
Extensive company-wide testing across scenarios, latency, accents, and edge cases
Client UAT with locked test data set and formal AI risk acceptance documentation
Successful Day 1 with zero Sev1 incidents, only 7 tickets logged post go-live (3 due to incorrect client data)
Ongoing AI tuning with intent-level analysis to improve containment beyond initial 14%
Technology Stack
Key Results
Related Case Studies
Global Clinical Research Organization: Executing a Time-Critical Cloud Divestiture with Zero Disruption
Migrating 1,200 agents to AWS in just 4 months to beat severe deadline penalties, while setting the stage for AI-driven patient assistance and achieving a 25% reduction in operating costs.
Read case studyDigital Healthcare Provider: Delivering Faster, Smarter, and More Trusted Patient Interactions
Breaking down data silos to radically accelerate patient verification and improve clinical routing, reducing talk time by 20% while achieving over 90% faster identity verification.
Read case study