A health plan needed to know if AI could replace the manual process of calling provider offices to schedule member appointments. We delivered a working prototype in two days that proved it could.
A health plan offers an appointment scheduling service where staff call provider offices on behalf of members to book appointments. It's a high-value service that members love, but the process is entirely manual and phone-based: call the provider office, wait through the IVR menu, wait on hold, speak with a receptionist, negotiate available times, and confirm the appointment.
Each call averages 10 to 15 minutes, with significant variability depending on the provider office. Some offices have complex phone trees with multiple layers. Others have long hold times. Some require specific information (insurance ID, referral number) before they'll schedule. The actual human conversation at the end is just a fraction of the total call time.
Because the process is so labor-intensive, the health plan can only offer it in limited circumstances. They wanted to explore whether AI could handle this workflow, which would allow them to scale the service across their entire member population without the headcount that would otherwise make it impossible.
The health plan asked us to prove the concept. We delivered a working AI voice agent in two days that could conduct the entire outbound scheduling workflow from dial to confirmation.
The agent receives a scheduling request with member details, insurance information, preferred providers, and scheduling preferences (dates, times, location). It dials the provider office directly.
The agent listens to the automated menu, comprehends the options (whether spoken or described), and navigates using DTMF tones or voice responses. It handles multi-level phone trees, "press 0 for operator" shortcuts, and common IVR patterns. When placed on hold, it waits patiently. It has no queue of other callers getting frustrated.
When connected to a receptionist, the agent conducts a natural conversation: identifies itself as calling on behalf of the member and health plan, provides required information (member ID, insurance, date of birth), requests available appointment times, and selects the best match based on member preferences. It handles common questions around new patient status, referral requirements, and insurance verification.
Once an appointment is booked, the agent captures the confirmed date, time, location, and any preparation instructions. This is structured and reported back to the health plan's scheduling system, and the member is notified of their appointment.
In two days, BusyKeys delivered a working prototype that proved an AI voice agent could navigate real-world provider phone systems, interact with human receptionists, and book appointments. The agent handled IVR trees of varying complexity, maintained natural conversation flow, and accurately captured appointment details.
The prototype gave the health plan a clear answer: this service can scale. What was previously only offered in limited circumstances due to the labor required can now be extended to their entire member population. The AI agent handles the repetitive phone work, which means the health plan can offer appointment scheduling as a standard member benefit rather than a restricted service.