How Cadence cut response times to at-risk patients by 97% with ElevenAgents

Cadence delivers clinical care through AI for patients managing chronic disease: monitoring vitals daily, supporting medication adjustments, and responding around the clock between visits. Cadence treats more than 100,000 active patients across 20+ health systems, including Hackensack Meridian Health and Yale New HavenHealth. 

Most of that care happens over the phone, and when a patient logs an out-of-range vital, it could signal a serious health event. Cadence's median time from abnormal reading to patient contact was 1 hour and 48 minutes, driven by more than 40,000 alerts a month, a touchtone IVR, and outbound calls dialed manually by clinicians. With its patient base doubling every year, hiring alone could not close that gap. Cadence built patient-facing voice agents on ElevenAgents and now reaches at-risk patients 30x faster.

Why Cadence chose ElevenAgents

Cadence tested various agentic solutions before selecting ElevenAgents on three criteria.

A HIPAA-compliant workspace 

ElevenAgents met Cadence's bar for clinical deployment with its HIPAA-compliant workspace, Zero Retention Mode. ElevenLabs never stores protected health information, end-to-end encryption in transit and at rest, and full control over what context each agent receives.

Natural, low-latency voices 

A patient calling about an abnormal reading is often anxious, and the voice on the line determines whether they continue the conversation. ElevenLabs' low-latency, natural voices meant many patients felt comfortable speaking with an AI agent. For Cadence's older patient population, interruption detection and background noise filtering keep conversations on track.

The voice was really natural and low latency, which is incredibly important. No one wants to be talking to a robot. When a patient has an important health need or question, you want that conversation to feel as natural and responsive as possible.

– Rachel Wang, Vice President of Engineering, Cadence

Speed to deploy

Cadence went from development to production calls in one month, building entirely in-house. Native integrations, such as Twilio, connected ElevenAgents to Cadence's existing infrastructure without modification, and the platform's APIs let engineers deploy agents from the tools they already use.

We tested a number of different options on the market, and ElevenAgents really has built a platform that allows you to actually roll out agents in production. You get A/B testing, you get configurability, they have APIs, they have a testing framework. It's really a fully complete solution.

– Eric Hauser, CTO, Cadence

Cadence's clinical protocols, built directly into ElevenAgents

Cadence built its clinical protocols directly in ElevenAgents rather than adapting them to a vendor's template. Its alert screening agent triggers automatically on out-of-range vitals, calls the patient, runs a symptom assessment, and warm-transfers the patient to a nurse on Cadence's 24/7 team as appropriate.

The rollout began with a human-in-the-loop pilot of roughly 2,000 calls, then scaled. Cadence ran the validation on ElevenAgents' built-in evaluation criteria, scoring calls for scope adherence and false positives. The agent held to its clinical guardrails 100% of the time and ElevenLabs' transcription matched or outperformed an alternative model on nearly every call tested, with better accuracy on medication names.

Faster patient contact, more capacity for clinical staff

A patient who once waited nearly two hours to hear from Cadence now gets a call in a median time of 3.5 minutes. The agent handles the alert workflow end-to-end, resolving 60% of the alerts it screens according to protocols without clinician intervention. With calls now arriving closer to the initial alert, patients pick up for the agent more often than they did for human outreach: ElevenAgents drives 47% more patients to answer than the human RN baseline of 41%.

When a call needs escalation, the patient arrives with their clinical need screened and verified, and nurses practice at the top of their license. With screening work removed, Cadence's Care Team has been reaching other clinical needs faster.

We also saw some very positive downstream effects. By reducing work in clinicians’ queues, we helped surface more urgent clinical escalations sooner and gave clinicians more capacity to act on them quickly. Focusing ElevenAgents on certain clinical escalations gave us even more leverage to streamline operations.

– Nick Tomaszewski, Senior Product Manager, Cadence

What's next

More than 90% of older adults in the U.S. live with at least one chronic condition, and the clinical workforce to treat them is shrinking. Cadence is expanding with ElevenAgents to close that gap, moving beyond alert screening into scheduling, device troubleshooting, appointment preparation, and patient activation. Each new agent takes on the protocol-driven calls, and every call handled end-to-end returns time to clinicians for the work that requires their judgment.

We don't have enough nurses and physicians to meet the needs of America’s rapidly aging population and growing chronic-disease burden. With ElevenLabs, our voice agents can engage patients within minutes of a concerning reading, at a scale no clinical team could sustain manually. That's what clinical AI should do: not replace clinicians, but extend their reach so every patient receives timely, human-centered care.

– Chris Altchek, CEO and Founder, Cadence