Back to blog
Operations

What Happens When the Patient Asks for a Real Person

Every AI phone system can answer a call. Far fewer can get out of the way when the patient wants a human. Here is what a live handover looks like when it works, and why the request is a signal rather than a failure.

Anshul Mittal2 August 2026

Every clinic asks us the same question, usually within the first ten minutes of a conversation. Not about accents, not about accuracy.

"What happens when the patient just wants to speak to a real person?"

It is the right question. It is also where most AI phone systems quietly fall apart.

The three bad endings

Ask a typical automated system for a human and one of three things happens.

It loops. It restates its options, misreads the request as a new enquiry, and asks again how it can help. The patient repeats themselves twice and hangs up.

It dumps. The call is transferred blind to a line nobody is watching, which rings out into voicemail. The patient has now been passed from a machine to an empty room.

It defers. Somebody will ring you back. Sometimes somebody does, several hours later, by which point the patient has booked elsewhere.

All three share the same flaw. The system treats a request for a human as an error to be handled rather than an instruction to be followed.

Hot patching, in plain terms

Our agents hand the call over live. The patient asks for a person, and the call is patched through to a real member of the practice team while it is still in progress — no hang-up, no callback queue, no second attempt required from the patient.

The part that matters is what travels with the call. Whoever picks up already holds the context: which treatment was being discussed, what the patient asked, what was answered, what is still outstanding.

The patient never repeats themselves. The team never starts blind. A transfer that begins from zero is barely better than a hang-up.

That distinction is the whole capability. Passing a call is trivial; every phone system built in the last forty years can do it. Passing a conversation is the difficult part, and it is the only version patients actually notice.

Ask to speak to a human
Fitting, given the subject. Message us and a person will reply.

The request is a signal, not a failure

There is a temptation to measure these systems by how few calls escape them. It is the wrong metric, and it produces exactly the wrong product.

Consider who actually asks for a person.

  • Nervous patients. Often the ones considering the largest treatments, who want reassurance before they commit to anything.
  • Complex cases. Multiple treatments, previous work elsewhere, something that does not fit a standard enquiry.
  • Complaints. Without exception. Someone unhappy should reach a human immediately, every time.
  • Anyone clinically worried. Pain, bleeding, swelling, a question about medication. These are not front-desk conversations at all.

Read that list again and it becomes obvious. The conversations most likely to trigger a request for a human are the ones with the most revenue, the most risk, or the most human need attached. Suppressing those handovers to protect an automation statistic is a bad trade in every direction.

Escalation is a clinical requirement, not a convenience

Some of this is commercial. Some of it is not negotiable.

A front-office system should never be the thing standing between a worried patient and a clinician. Ours is built to recognise clinical language, distress, and complaint and to route those calls to a person by design rather than by accident — regardless of what stage of the conversation they arrive at.

The same applies to disclosure. Patients are told they are speaking to an AI assistant in the agent's opening words, before anything else is discussed. A patient who knows what they are speaking to can make an informed request for something else, which is the entire basis on which a handover works.

A useful test: ask any vendor what their system does with a caller who is in pain and upset. If the answer involves a form, a queue, or a callback window, you have your answer about everything else too.

What if nobody is free?

Practices ask this immediately, and they are right to. A handover promise is worthless during a busy Monday morning or at nine on a Sunday evening.

The conversation does not dead-end. If no one can take the call, the agent says so honestly rather than transferring into silence, captures what the patient needs, and continues the same conversation over WhatsApp — carrying the full context across, so the patient is not starting again in a second place.

The enquiry stays alive either way. It is either finished by a person on the phone or finished in writing on a channel the patient already has open. What it never does is disappear.

Put the awkward question to us
Straight to a conversation. No form, no callback window.

What to ask before you buy any of this

  1. Is the transfer live, or is it a callback? "We'll get someone to ring you" is not a handover. It is a queue with better manners.
  2. Does context travel with the call? If your team answers and has to ask what this is regarding, nothing has been saved.
  3. Who decides when to escalate? The patient should be able to ask, and the system should also recognise when to offer without being asked.
  4. What happens outside opening hours? Most handovers will be requested when the practice is closed. That is when the answer matters most.
  5. Can you set the rules? Which enquiries always reach a person, which numbers they route to, what happens if that line is engaged.

Automating the front desk, not fencing it off

An AI receptionist that cannot reach your team is not automating your front office. It is standing in front of it.

The purpose was never to keep patients away from your people. It is to make sure the routine enquiries never occupy them, and the ones that genuinely need them get through immediately, with everything already established.

Never miss a lead is the promise. Handing the right conversations to a human, mid-call and in full context, is a large part of how it holds.

Start a conversation on WhatsApp
Tell us your practice type and we will show you how the handover would be configured.