AI voice intake that cut patient hold time from 94 to 22 minutes

Patients calling the network's intake line waited an average of 94 minutes. Staff triaged calls by hand and typed the same demographic and symptom details into the network's medical record system again and again.

The network needed intake that kept up with call volume without adding staff and without moving protected health information outside a controlled environment.

The team the founder led built a voice agent on Amazon Bedrock that answered calls right away. It verified the caller, ran a structured intake conversation, and wrote the details into the medical record workflow, so staff reviewed records instead of typing them.

Calls the agent could not or should not handle (clinical questions, distress, anything unclear) went to a person, with the transcript and the details already captured attached.

Average hold time went from 94 minutes to 22, about 77% less. The figure is Khavion's own; the client is under NDA, so there is no public source to link.

The system went into production and ran inside the client's own cloud accounts. This page describes it as it was when the founder left that job.

The founder's employer at the time delivered this in a HIPAA-eligible environment under a signed Business Associate Agreement. There is no HHS-recognized certification for HIPAA, so this page describes the safeguards instead: protected health information was encrypted in transit and at rest, access went through IAM with MFA, and audit logs were kept for six years.

We would design the handoff to a human first: distress, clinical questions, anything the agent should not decide. We would also make sure the intake data has a clean place to land before quoting any hold-time target.