Amazon Nova 2 Sonic: How to Build a Voice Agent for Patient Scheduling

Amazon has released a detailed guide to building a voice agent for healthcare providers, based on Nova 2 Sonic and Bedrock AgentCore. The agent calls patients on its own, confirms appointments, and reduces the no-show rate without an operator on the line.

What the voice agent can do

The agent takes over routine reminder calls about upcoming visits. It holds a full voice conversation: it authenticates the patient, clarifies the appointment details, and records the response in the scheduling system. The entire conversation runs autonomously, and front-desk staff step in only when a situation falls outside the standard scenarios.

Key capabilities of the agent:

• Voice authentication of the patient before the conversation begins
• Confirming, rescheduling, or canceling an appointment
• Collecting information about how the patient is feeling the day before the visit
• Escalation to a live operator for non-standard requests
• Parallel handling of hundreds of calls at scale

This architecture lets a clinic cover its entire volume of reminders without growing its staff, and at any time that suits patients, whether evenings or weekends.

How it is built

At the core of the stack is Amazon Nova 2 Sonic, a model with native support for real-time two-way voice. Classic systems first recognize speech as text, then process it, then synthesize it back into audio. Nova 2 Sonic works with the audio stream directly, which reduces latency and makes the conversation feel more natural.

Bedrock AgentCore handles orchestration. It manages calls to external functions, such as checking the schedule, writing changes, and retrieving patient history, and it decides the next step of the conversation based on the data it gets back.

For development and testing, the guide includes a browser-based interface, so a developer can walk through the entire scenario without real telephony. For production, AWS recommends integrating with Amazon Connect, but the tutorial leaves that layer out of scope and focuses on the agent logic and voice interaction.

Why it matters for healthcare

"You handle routine calls at scale, which helps reduce the rate of

missed appointments," the official description of the solution states.

No-shows are a chronic problem in healthcare: by various estimates, 15–30% of appointments fall through without warning. Clinics lose revenue, and patients with a real need are left without a slot. Automated voice reminders that let a patient reschedule immediately during the call reduce this rate without adding load on staff.

An additional benefit is collecting information about how the patient feels before the visit. If a patient reports that their condition has worsened, the doctor gets context in advance: they can adjust the appointment protocol or offer an alternative format, such as telemedicine or an earlier slot.

What this means

A ready-made template from AWS significantly lowers the barrier to entry. A reminder voice agent can now be deployed without a team of ML engineers: it is enough to adapt the example to your own scheduling system and telephony provider. Healthcare is becoming one of the first sectors where voice AI agents are moving from pilot projects into standard operating practice.