---
title: "Voice bots in a clinic: the questions I ask before any pilot"
description: "Voice AI on a clinic phone line can work. Whether it does depends on questions most vendors are not asked until after the contract is signed."
date: "2026-08-31"
updated: "2026-09-16"
topic: "Operations automation"
canonical_url: "https://rolstonpaul.com/writing/voice-bots-in-a-clinic-the-questions-i-ask-before-any-pilot"
author: "Rolston Paul"
reading_time: "6 min read"
---

# Voice bots in a clinic: the questions I ask before any pilot

Voice AI on a clinic phone line can work. Whether it does depends on questions most vendors are not asked until after the contract is signed.

Voice AI on a clinic phone line is one of the few automation cases where the business argument is easy. Calls are missed, evenings are uncovered, and a large share of the conversations are the same six exchanges repeated all day.

Here is the order I would use for the evaluation. Each question should produce evidence or an explicit condition to resolve before the pilot.

A voice pilot for international enquiries should be tested across the languages and time zones the clinic actually serves. I would include a caller describing a concern after returning home and check whether the system routes it to a staffed clinical pathway. India will not lead aesthetic medical tourism through an impressive demonstration if the handover fails when a patient needs a person.

## Before the vendor: what am I actually fixing

What is the number this is meant to move? Missed calls, first response time, appointments attended, agent hours on repetitive work. One number, named before any demo.

Collect a baseline over a representative operating period before evaluating the tool. Include variation by hour and staffing pattern. If the problem can be addressed through the rota, test that option in the same business case rather than assuming technology is required.

Is the bottleneck the phone or the diary? If calls are being missed because there are no consultation slots to offer, a faster answering system just delivers disappointment more efficiently.

## Scope: what may it do, what may it never do

I split every use into three buckets before the first demo.

Fine: outbound reminders and confirmations, rescheduling, directions, opening hours, basic qualification questions on an inbound enquiry, post-visit satisfaction checks.

Only with a human close by: pricing conversations, first clinical enquiries, anything where the caller sounds uncertain.

Never: clinical advice, symptom assessment, medication questions, results, complaints, or any call where the person is distressed. Not because the model cannot generate a fluent answer. Because it can, and a fluent wrong answer in a warm voice is worse than a hold queue.

Write the never list down before the pilot, in the contract if possible. It is much harder to add later, when the containment rate is being celebrated.

## The questions I put to the vendor

Language and accent coverage for my actual patients, tested on recordings from my own line, not their showreel. Indian English across regions, code switching mid sentence, background noise from a clinic reception. Every vendor demos beautifully on clean audio.

Latency, measured end to end, on my network, at my busiest hour. Anything above roughly a second of silence and callers start talking over the bot, which degrades everything downstream.

Interruption handling. Can the caller cut in? What happens when two people speak at once? This is where natural conversation actually lives.

Escalation. How does a caller reach a human? How many turns does it take? Does the human receive the transcript and the context, or does the patient start again? A transfer that resets the conversation is worse than no bot.

Data. Record where audio is sent, where it is stored, who can access it, how long it is kept and whether it is used for model training. Have the responsible privacy and legal advisers assess the applicable requirements before the pilot. Patient permission and a supplier's claims need to match the actual arrangement.

Failure behaviour. What does the system do when the model is unavailable, when confidence is low, when it does not understand twice in a row? "It apologises and tries again" is the wrong answer. The right answer is that it hands over.

Pricing shape. Per minute, per call, per resolution, per seat. Model it at three times your expected volume and again at a third. Vendors price for the volume they hope you reach.

Who owns the prompts and the call flows. If you build six months of refinement into their platform and then leave, what comes with you?

## The pilot design

One line, one use, one comparison. I would start with outbound appointment confirmation, because the failure is cheap and the baseline is easy.

Run it for at least eight weeks and make sure that window includes a bad week. Every system looks good in a calm week. What I want to see is the Monday after a holiday when volume triples.

Measure containment, transfer reasons, abandonment, attended appointments and complaints against the baseline. I would examine whether ending more calls inside the bot coincides with missed appointments or failed escalation. A vendor's containment figure is not enough to judge the service.

Listen to fifty calls yourself. Not a sample the vendor picks. Fifty, chosen at random, including every transfer. Nothing on a dashboard tells you what fifty calls will.

## The human veto, written down

One paragraph in the governance page. Something like: this system may confirm, reschedule and inform. It may not advise, assess or decide. Any caller may reach a person by saying so once. Named owner, reviewed quarterly.

That paragraph is what turns a technology decision into a clinical one, which is what it always was.

## Where I land

I would run the pilot. I am not sceptical about voice AI in a clinic, and the version of this technology available now is genuinely good at the narrow, repetitive, outbound end of the work.

What I would not do is buy it to solve a problem I have not measured, or let it near a distressed caller, or sign before I know where the audio is stored. Those three refusals cost nothing and they are the whole difference between a pilot that teaches you something and a subscription you cannot explain in a year.

## Questions people ask

### Should a clinic use a voice bot to answer patient calls?

For outbound reminders, confirmations and simple qualification, yes, with a fast route to a human. For inbound clinical enquiries and anything involving distress, symptoms or complaints, no. The cost of a wrong answer is not symmetrical.

### What should a voice AI pilot measure?

Containment rate, transfer rate and why, abandonment, appointments actually attended, and complaint volume. Compare all five against a baseline you collected before the bot existed, over a period long enough to include a bad week.

### What is the biggest risk with voice AI in healthcare?

A confident wrong answer to a clinical question, delivered in a natural voice that the patient trusts. The second risk is a distressed caller trapped in a menu with no way to reach a person.

---

Canonical page: https://rolstonpaul.com/writing/voice-bots-in-a-clinic-the-questions-i-ask-before-any-pilot
Topic guide: https://rolstonpaul.com/topics/operations-automation
Published 31 August 2026 by Rolston Paul. Contact: info@rolstonpaul.com