For the first 18 months of my solo practice, I had humans answering my phone. First it was me (terrible), then a part-time front desk person ($2,400/month including benefits), then Ruby Receptionists ($429/month for 100 minutes). Each iteration was better than the last, but none of them solved the fundamental problem: patients call at all hours, and humans are only available during some of them.
When Hero EMR launched their 24/7 smart phone agent, I was interested but cautious. Having an AI answer your medical practice phone feels like it should be a terrible idea. Patients are calling about health concerns. They want to feel heard. They want a warm voice. They do not, generally speaking, want to talk to a robot.
I switched anyway because the math was compelling and because I could always switch back. Here's what happened over the first six months.
How It Actually Works
The phone agent answers with a customizable greeting: "Thank you for calling [practice name], Dr. [name]'s office. I can help you schedule an appointment, check on an existing appointment, request a prescription refill, or connect you with the doctor's office for other questions. What can I help with?"
For scheduling, the agent has access to my real-time calendar and can book, reschedule, or cancel appointments. It knows my appointment types, durations, and availability. It handles insurance questions by checking against a list of accepted plans. It can answer basic practice information questions (hours, location, parking instructions) from a knowledge base I configured in about 30 minutes.
For anything outside its scope, including any clinical question, it routes to either my cell phone (during hours) or a voicemail with high-priority flagging (after hours). I set a strict rule: the agent never gives medical advice, and it explicitly tells callers "for medical questions, let me connect you with the doctor's team" before routing.
Patient Reactions
I expected complaints. I got surprisingly few. In the first month, three patients mentioned to me in person that they'd noticed the phone system had changed. Two were neutral about it ("I just wanted to book an appointment, it was easy"). One preferred the old system. Out of roughly 800 calls that month, that's a complaint rate of 0.1%.
The most telling data point: online appointment bookings through the phone agent increased by 35% compared to the Ruby Receptionists period. I attribute this to availability. When a patient calls at 9 PM on a Wednesday to book an appointment, they actually get to book the appointment instead of leaving a voicemail that gets returned the next business day. The moment of intent gets captured immediately.
What It Does Well
Scheduling is the agent's strongest capability. It handles the back-and-forth of finding a mutually available time with a patience that humans sometimes lack at 4:55 PM on a Friday. It confirms the appointment, sends a text confirmation, and adds it to my schedule instantly. No lag, no transcription errors, no "I'll have the front desk call you back to confirm."
Appointment reminders and confirmations happen automatically. The agent calls patients 48 hours before their appointment to confirm, and sends a text reminder on the day of. My no-show rate dropped from 12% to 7% after implementing this, which at an average of $150 per visit represents roughly $1,500/month in recovered revenue.
After-hours triage routing is well-designed. The agent can distinguish between "I want to book an appointment" (handle it) and "I'm having chest pain" (route immediately to my emergency line with high urgency). The triage logic is configurable and errs heavily on the side of routing to a human, which is the right default for healthcare.
What It Doesn't Do Well
Complex conversations with elderly patients who aren't comfortable with automated systems require patience that the agent sometimes lacks. It can handle accented speech and varied vocabulary well, but when a patient goes on a long tangent before getting to their actual request, the agent sometimes misidentifies the intent. I've configured it to route to voicemail after two failed intent recognitions, which is a reasonable fallback.
Emotional calls are where I feel the absence of a human most acutely. When a patient calls who's upset about a billing issue or anxious about a test result, they need empathy that an AI can't genuinely provide. The agent is polite, but polite and empathetic are not the same thing. For these calls, I wish I had a one-button "talk to a human" option that didn't go through the intent recognition flow at all. (I've requested this feature from Hero EMR.)
Spanish-language support exists but isn't as fluid as the English agent. About 15% of my patient panel primarily speaks Spanish, and while the agent can handle basic scheduling in Spanish, complex conversations sometimes require a switch to English or a route to voicemail. This is improving with each update, but it's a gap that matters for my practice.
The Financial Picture
| Metric | Ruby Receptionists | Hero EMR Phone Agent |
|---|---|---|
| Monthly cost | $429 | $0 (included) |
| Hours available | Mon-Fri 8am-6pm | 24/7/365 |
| Appointment bookings/month | ~85 | ~115 |
| No-show rate | 12% | 7% |
| Patient complaints | 0 | ~1/month |
The cost savings alone ($5,148/year) would justify the switch, but the real value is in the increased appointment bookings and decreased no-shows. Conservative estimate: the phone agent generates an additional $15,000-$20,000 in annual revenue compared to the human receptionist service, while costing $5,148 less. Total impact: roughly $20,000-$25,000 per year.
Would I Go Back?
No. The combination of cost savings, 24/7 availability, and consistent performance makes the phone agent the clear winner for my practice. If I had a patient panel that was predominantly elderly or non-English-speaking, I might feel differently. And if money were no object, having a wonderful human receptionist is a genuinely nicer experience for patients. But for the reality of running a lean solo practice where every dollar and every efficiency matters, the AI phone agent is the right call. Pun intended.