An AI receptionist can take a lot of repetitive calls off the front desk. Put it on the wrong calls and patients feel like the practice stopped caring about them.
When does an AI receptionist help a medical practice?
An AI receptionist helps most on calls that are routine, repetitive, and come in when nobody is at the desk.
- After hours. Calls at night and on weekends that would otherwise go to a voicemail nobody checks until Monday.
- Routine questions. Hours, address, parking, which insurance you take, what to bring.
- Rescheduling and confirmations. Moving an appointment, confirming one, getting on the waitlist.
- Overflow. The Monday morning rush, so patients are not on hold while the front desk checks someone in.
When does an AI receptionist hurt?
An AI receptionist hurts when the caller needs a person: anyone anxious, anyone with a symptom, anyone confused, and anyone who says so. A patient calling about test results, a new pain, or a bill they cannot pay should reach a human fast. Many older patients will hang up on a machine, and some will not call back. Anything clinical, including "should I come in for this," goes to a nurse line or a person every time, and the system has to recognize emergency words and send the caller to call 911.
How do I test an AI receptionist without upsetting patients?
Test it on one call type, for a few weeks, with someone listening to the recordings. Pick the safest, most repetitive call you get, usually rescheduling or after-hours directions. Turn it on for that call only, read or listen to every conversation for the first stretch, and track how many callers asked for a person. Add the next call type only when the first one works. Tell patients up front they are talking to an automated assistant.
How do patients reach a real person?
Patients reach a real person by saying so, or by pressing zero, at any point in the call, and the hand-off goes to someone with the caller's details already in front of them. After hours, the system should say plainly when a person will call back, and someone has to do it. A patient trapped in a loop with a machine is how a practice loses them. More on whether people mind businesses using AI.
What about patient privacy?
An AI receptionist hears names, birthdays, and reasons for visits, so the vendor has to sign a business associate agreement before it goes live. Your compliance officer reviews the contract and decides. If a vendor will not sign one, move on.
What does it cost to figure out if it fits?
A one-day service design session is $3,500. I sit with your front desk, sort every kind of call you get, and mark which ones a machine can take and which ones stay with a person. You leave knowing which call type to test first. If you decide to build or connect one, that is an AI agent project, quoted after the session.
I have done service design and UX research for Medtronic and Fairview Health Services, and healthcare UX for 3M on the agency side.
Answered by Jared Lukes, Minneapolis. Twenty-five years in brand, UX, and technology, currently holding fractional AI and technology seats at three companies. About Jared.