Grace — Production AI Patient Outreach at Scale
A fully automated AI system that calls and texts patients to check on their wellbeing, scores every conversation in real time for clinical risk, and hands off patients-in-need to live staff — operating under HIPAA and clinical-safety constraints from day one.
Most AI demos never reach production. Grace did — calling and texting real patients in a regulated clinical environment, scoring conversations for clinical risk in real time, and escalating patients who need help to live staff for immediate follow-up.
The full write-up will cover:
- Designing a novel dialog engine on top of third-party telephony and speech services
- What clinical-safety constraints actually do to an AI architecture — and why they made the system better
- Real-time conversational risk scoring and the handoff path to live clinical staff
- Evolving from voice-only to multi-modal voice + SMS: asynchronous threaded conversations, longer time-to-resolution windows, different pacing
- Measuring what matters: escalation precision, emergency-room visits prevented, hospital admission rates
Full case study coming soon.