IVR, missed-call lists, callbacks, a voicebot: which of them actually answers the phone in a clinic

When a clinic tells me it "already has the phones sorted", I ask what exactly happens to the fourth call when three lines are busy. The answers vary and nearly all of them are true. The menu routes it. The phone system logs it. Someone will call back. The answer I rarely hear is: it gets picked up by someone who can book the appointment.

In the piece on voice assistants I wrote that most telephony solutions record the call instead of answering it. That is true, but a little unfair to IVR and callbacks, because each of them solves a real problem. Just not the same one. This piece is about what each of these tools does well, where it stops, and how to tell which one a clinic actually needs.

Every telephony tool does one of four things with a call. It routes it to the right place. It makes it wait. It records it for later. Or it handles it, meaning there is someone or something on the other end that can close the matter.

Only the fourth one increases the number of requests the front desk gets through in an hour. The first three change what the queue looks like, not how fast it shrinks. That does not make them useless. It means each helps with a different symptom, and it is worth knowing which.

A recorded menu with keypad choices: "for public-payer bookings press one, for private bookings press two, for test results press three".

A note for readers outside Poland: many Polish clinics run two kinds of appointments side by side, ones funded by the public payer (NFZ, the National Health Fund) and private, paid ones, often handled by separate desks. That split is where IVR earns its keep most often.

What it does well. It spreads calls across people or teams who already exist. If a clinic has two booking desks and private patients kept landing in the public-payer queue, IVR fixes that cheaply and immediately. It is also useful for messages that would otherwise eat up calls: opening hours, days a given doctor is not seeing patients, the address of the blood draw point.

Where it stops. IVR does not create a new person to talk to. If there are two receptionists behind option one and both are on the phone, a patient who presses one waits exactly as long as before, just after a longer introduction. We spoke with a clinic that had introduced a public/private split, wait-time measurement and a callback list. Traffic got split, the workload did not go down, and the queue at peak still reached twenty people.

What to watch out for. Every menu level is a dozen or so seconds before the patient even joins the queue. Two, at most three options per level. A menu you have to hear to the end to reach "all other matters" puts people off more than an engaged tone does.

The phone system logs every call nobody answered. During a break or at the end of a shift, someone opens the list and starts dialling.

What it does well. Nobody disappears without a trace. For a clinic that did not know how many calls it was losing, the list alone can be a wake-up call, and that is real value. It is also the raw material for the measurement I describe below.

Where it stops. The list has a number and a time, but no reason. The person calling back does not know whether they are phoning someone who wanted to book, cancel or ask about a result, so every callback starts from zero. Some entries are the same person who tried five times.

There is also a less obvious point. A list of phone numbers with times, exported to a spreadsheet on a shared drive, is personal data that someone has to protect, and once a reason for the visit is added, it is health data too. I wrote about this in the piece on phone bookings in a clinic covered by NIS2.

Two things go by the same name here and they need separating.

The first is manual callbacks from the list, usually hours later. The second is an automatic callback from the queue: the system tells the patient they are fourth in line and offers to hang up, and the phone system rings them back as soon as a line frees up. The patient keeps their place and does not sit on hold.

What it does well. The second version genuinely helps when the peak is short. If the problem is forty minutes after opening and then the phones calm down, a callback within a quarter of an hour is enough and nothing more is needed.

Where it stops. Callbacks do not reduce the number of conversations, they only move them in time. If the front desk is busy all day, the callback queue grows all day too, and the patient gets the call when they have already booked somewhere else. One number tells you which situation you are in: how long it takes from a missed call to the callback. Minutes are a solution. Hours are a formality.

A voice assistant picks up, talks, checks the calendar for a slot and books the appointment. If the matter is beyond it, it hands it to a person with a description, not just a number. I covered this in detail in a separate piece, so here only what matters for the comparison.

What it does well. It adds someone to talk to at exactly the moment the patient calls, including at peak, in the evening and at the weekend. It takes repetitive requests off the front desk: booking, rescheduling, cancelling, questions about opening hours.

Where it stops. It is only as good as its access to the calendar. Without booking into the system where the front desk later sees appointments, it is just a missed-call list with a nicer voice. It does not assess health and does not decide urgency, so medical matters go to people anyway. It also takes more preparation than anything above: integration, conversation scripts, sorting out recordings and legal bases, and telling the patient they are talking to an AI system.

The honest answer is that it depends on the shape of your calls, not on what is newest. A few typical symptoms and what answers them.

I will also say it plainly: an extra person on the front desk during peak hours is a perfectly valid option, not a failure. They know the clinic and the patients, and they will handle matters no system should touch. In many of the clinics we talk to, the problem is that such a person cannot be found, not that nobody thought of it.

These tools are not mutually exclusive either. In a larger clinic a sensible setup is often a short IVR that splits traffic, a voice assistant that picks up what people cannot get to, and a callback list that only holds requests with a description.

Everything above comes down to a few numbers that nearly every phone system already collects.

After a week you have an answer no salesperson will give you, us included: whether your problem is order, a peak, or a shortage of hands.

If it turns out you are short of hands, the natural question is what a voice assistant costs and how to work out whether it pays for itself. That is a topic for a separate piece, where I do the sums on the cost of front-desk work rather than on promises.

A voice assistant is only one part of the layer between the patient and the practice management system. Online booking, text reminders and reviews are covered on our page for healthcare providers.

If you already have these numbers from your phone system, or you do not know where to get them, write to me and we will tell you what they point to. I usually run that conversation myself, and if it turns out a rollout is needed, it is done by the team the clinic deals with directly. Sometimes the best advice turns out to be a two-button IVR, and then that is exactly what we will say.