Behind the Agent: Reaching At-Risk Patients When Minutes Matter

As remote patient monitoring scales, so does the volume of abnormal readings that need a response. You can add staff, optimize queues, and redesign workflows, but there’s still a gap between the alert firing and actually reaching the patient. Cadence took a different path. Rachel Wang joins ElevenLabs for a behind-the-scenes conversation on the voice agent that now calls patients automatically the moment a vital goes out of range - and what it took to get it safe enough to talk to a patient. Rachel Wang, Vice President of Engineering, Cadence Caleb Wright, Enterprise Deployment, ElevenLabs John Durovsik, Growth, ElevenLabs What we'll cover: - What the workflow looked like before: alerts routed through a touchtone IVR, clinicians dialing patients by hand, and why hiring couldn't scale with the patient base - How the agent triggers on an out-of-range reading, runs a symptom assessment, and either resolves the alert per Cadence's protocols or warm-transfers to a nurse on the 24/7 team - Real metrics: 97% reduction in time to patient contact, 47% higher answer rate than human outreach, and 250,000 calls handled in the first seven months - Why patients answer an agent more often than a person, and what that changed about how Cadence designs outreach - What had to be true on clinical safety and escalation design before the agent was allowed to speak to a patient - Where Cadence draws the line between what the agent resolves and what always goes to a clinician