Answering every call at a practice, without touching the clinical ones

What the numbers say, and what they do not

Start with the honest caveat. The widely quoted patient-behaviour statistics, the ones claiming two in three callers ring a competitor after a voicemail, trace back to marketing pages with no published method. We will not repeat them. Definitions vary too: some counts treat any unanswered call as abandoned, others count only a caller who waits in a queue and then hangs up, and the two produce very different numbers from the same phone system.

One dataset is worth looking at. A 2025 analysis of 4.7 million calls across 297 practices found roughly 30 percent abandoned during business hours, rising to 46 percent on weekday evenings and 48 percent at weekends. It is a vendor's own data, so treat it as directional rather than settled. Against that, the commonly cited healthcare call-centre average sits near 7 percent, with physician practices running 10 to 20 percent at peak.

Those two pictures disagree sharply, and the gap is the point: nobody can tell you your abandonment rate from an industry average. What survives the disagreement is the shape. Calls are lost most at the hours when nobody is at the desk, and that is a staffing problem with a technical answer.

The line that does not move

We said it in August and it has not softened. We do not put an agent on a line where the first ninety seconds carry a clinical consequence. A patient describing chest pain, a parent describing a febrile child, anyone whose words need a trained ear in the first breath: that call reaches a person, immediately, and the agent's only job is to get out of the way fast.

So the honest version of this article's promise is narrower than the headline anyone else writes. Every call gets answered. Not every call gets handled.

What that looks like in practice is a rule written before a phone number is configured. The agent opens, identifies why the caller is ringing, and either completes an administrative task or transfers. The transfer trigger list is written with the clinical lead, not with us, and it is deliberately generous: anything about symptoms, medication effects, test results, deterioration, or distress goes to a human without the agent attempting a judgement about severity. An agent that tries to decide whether chest pain is urgent is a clinical device, and it should be regulated as one.

Two further rules we insist on. Any caller who asks for a person gets one, on the first ask, without a retention script. And the agent says what it is in its opening line, because a patient discovering halfway through a call that they have been talking to software is a trust problem you will spend a year repairing.

What is left, and why it is most of the volume

Strip out everything clinical and a surprising amount remains. At most practices it is the majority of inbound calls, and it is the part that currently eats the front desk.

Booking, rescheduling and cancellation. The single largest category nearly everywhere, and the one where an abandoned call turns directly into an empty slot that could have been filled from the waiting list.

The same four questions. Opening hours, parking and directions, what to bring, whether you take a given insurer or payment method. These never need a person and they interrupt one constantly.

Repeat prescription requests, taken and routed. Taken down accurately and passed to the clinician or pharmacy queue. The agent records the request. It does not approve anything.

Reminders and the waiting list. Outbound confirmation the day before, and the call that offers a cancelled slot to the next suitable patient before it goes to waste.

Pick the first one carefully. Booking is usually where the money is, and it is also the one that fails most visibly if your appointment rules live in somebody's head rather than in the system, so a practice with inconsistent booking conventions should fix those before automating them.

Measure your own line for one week

Every number above describes somebody else's practice. Yours is the only one that decides anything, and you can have it by Friday.

Pull the call report your phone system already produces. Most VoIP platforms will export inbound volume, answered, abandoned, and average wait by hour, for the last ninety days, in about ten minutes. You are looking for the hours where the abandoned line rises, not the headline rate.

Listen to twenty calls with a pen. Mark each one administrative or clinical. That ratio is the single most useful fact in this whole exercise, because it tells you what share of your volume an agent could even legitimately touch.

Count the empty slots. How many appointments went unfilled last month while the waiting list had people on it. That number, not the abandonment percentage, is what a booking agent is actually competing against.

Then decide with three real figures in front of you. If your administrative share is low, your abandonment is already under control, or your booking rules change by clinician and nobody has written them down, the honest answer is that an agent is the wrong purchase this quarter and we will say so. If a third of your calls are dropping at the hours nobody is at the desk and most of those callers wanted an appointment, the arithmetic makes itself.

Install it with the handover rules written first

AI Operations Launch wires voice, messaging and web chat into one system in thirty days. On a medical line the order is strict: the transfer triggers are agreed with your clinical lead and signed off before a number is configured, scope and data handling are written down before anything is built, and the system is measured in week four against your own call report rather than an industry average.

See the thirty-day install →

← All insights