A practice manager asked me last month whether MedPhone was basically an answering service with a robot on it. It is a fair question and I gave a bad answer, so I have been thinking about the right one since.
The difference is not the voice. It is not even that one is cheaper. It is what exists at the end of the call.
What an answering service actually does
A human answering service picks up, takes the caller name, a number, and the reason for the call, and puts that in a queue. Some deliver it by email, some by portal, some page your on call at 2am if the message is flagged urgent. That is the product, and it is a good product. It has been around for fifty years because it solves a real problem: nobody hears a ring out.
What it does not do is finish anything.
The patient who called at 7pm to move Thursday does not have a new appointment. They have a message saying they would like one. Somebody on your staff opens the queue at 8am, calls back, gets voicemail, leaves a message, and now there are two open loops where there was one. The patient calls again at lunch. Your front desk is on another line. Round three.
An answering service converts a missed call into a callback. It does not remove the work. It moves the work to tomorrow and adds a round trip to it.
What changes when the thing answering can write to your chart
An AI receptionist that is actually integrated with your EHR does the same first ten seconds, and then keeps going. It finds the patient. It looks at what that provider genuinely has open, not a generic grid. It books the slot, writes the appointment back, and tells the caller what it just did.
The patient hangs up with an appointment. Nobody on your staff finds out about it until they look at the schedule, and there is nothing for them to action.
That is the whole difference, and it decides everything downstream:
- Callback queue. An answering service produces one every night. An integrated agent produces one only for the calls it deliberately hands off
- Time to resolution. Minutes, on the call, instead of a day and a half of phone tag
- Patient experience. The caller who got an appointment at 7pm does not know or care what answered. The caller who left a message and waited two days does
- Double booking. A message queue has no idea what your schedule looks like. Two staff working the same queue can promise the same slot twice
- Documentation. A summary written into the chart at the time of the call, rather than reconstructed from a message the next morning
Where an answering service still wins
I am not going to pretend this is one sided. There are cases where a human service is the better buy and you should know what they are before you sign anything.
It works with any system. There is no integration, which means there is nothing to build, nothing to approve, and no dependency on what your EHR vendor allows through its API. If you are on a system nobody has integrated with, that matters a great deal.
It can be live next week. Integrated agents take longer because the integration is the product.
And a person is better at a distressed caller. A patient who is frightened, or grieving, or calling about a result they have just read, should reach a human, and every serious vendor in my category builds rules to make sure they do. If most of your after hours volume is that kind of call, buy the human service.
The cost comparison, honestly
This part is harder to write than it should be, because most answering services do not publish a price. What you will be quoted is either per minute or per message, sometimes with a monthly minimum underneath it. Two things follow from that structure and both are worth knowing before the sales call.
The first is that the busier you get, the more it costs. Volume is the thing you want to grow and it is also the thing being metered. The second is that a per message model quietly rewards short messages, which is not the same as rewarding resolved problems.
Ask for the per minute rate, the per message rate, the monthly minimum, and what happens in a month when you go over it. All four exist. None of them are usually on the website.
Our own pricing is published, including the overage rate, because I think a category where nobody publishes anything is a category that is easy to be dishonest in.
See what MedPhone costs, including overage
The question that decides it
Forget the technology for a second and ask what problem you are actually buying against.
If the problem is that your phone rings out and patients hear nothing, an answering service solves that completely, and cheaply, and next week.
If the problem is that your front desk starts every morning with a stack of callbacks, that the 7:30am rush goes to voicemail while your staff are still parking, and that you cannot tell how many of those callers gave up and booked with the practice down the road, then a message queue is not the fix. It is the same work, timeshifted.
One more thing worth saying out loud: this is not a replacement argument. Nobody I have sold to has fired a receptionist. What they have done is take ten to fifteen hours a week off the phones and put those hours into billing and collections, which is where the money in a practice actually is.
Questions people ask
Can an AI receptionist transfer a call to a person?
Yes, and the rules should be yours rather than the vendor default. Clinical questions to the nurse line, urgent calls to whoever is on, and a different rule after hours if you want one. If a vendor cannot show you where those rules are configured, that is a real answer to a different question.
What happens when the AI does not understand the caller?
It should transfer, and it should do it before the caller has to repeat themselves three times. The failure mode to watch for in a demo is an agent that keeps trying rather than one that gives up gracefully. Ask to hear a call that goes wrong, not just one that goes well.
Do we have to drop our answering service to try one?
No, and for the first month I would not. Run the AI on your daytime overflow and leave the human service on nights until you have listened to enough calls to trust it. Overlap for a month costs you one month of a bill you were already paying.
Is an AI receptionist HIPAA compliant?
It depends entirely on the vendor, which is a boring answer to an important question. The thing to check is whether they sign a BAA before deployment, and whether every subprocessor that touches the audio is also under one. That second part is where most of the category is thin.
Which is better for a dental practice specifically?
The same reasoning applies, but the call mix is different. Dental volume skews toward hygiene recall and rescheduling, which is exactly the kind of routine, high volume, low ambiguity work an integrated agent handles well. The deciding factor is usually whether the vendor connects to your practice management system rather than anything about the voice.
Written by
Agni Patel is the founder and CEO of MedPhone. He has now had this conversation with enough practice managers to write it down.