How a private hospital can get an afternoon a day back from its front desk

Walk into the reception of almost any private clinic in Douala at eleven in the morning and you will see the same scene. Two people behind the desk. One is with a patient who is standing there in person. The other is on the phone. There is a second line ringing that nobody will reach in time, three unread WhatsApp messages on the clinic's phone, and a queue of four people waiting to be spoken to.

Nobody in that picture is doing anything wrong. The work is simply arriving faster than two people can absorb it.

What the phone is actually asking

Before proposing anything, we ask a clinic to do one boring exercise: for one week, tally every incoming call and message by what it was about. The result is remarkably consistent across the clinics we have looked at. Something like three quarters of the traffic is four questions:

• Are you open, and until when? Especially around public holidays, when the answer changes and nobody updates anything.
• How much does X cost, and do you take my insurance? Consultation, scan, delivery, dental work.
• Can I get an appointment on Thursday? Followed by a negotiation about times.
• Is my result ready?

None of those four require a clinical judgement. None of them require the person answering to have met the patient. They require somebody to know the opening hours, the price list, the appointment book and the status of a result — and to be free at the moment the phone rang.

The remaining quarter is a different matter entirely. Someone describing symptoms. Someone distressed. Someone asking whether they should come in now. That work is not the problem, and it is not what this article is about.

What an agent does with that traffic

An AI agent, in this setting, is software sitting on the clinic's WhatsApp number and website chat. It reads what the patient wrote, works out which of those questions it is, answers from the clinic's own price list and calendar, and writes down what happened.

For the appointment case, it does more than answer. It offers the three next available slots for the right doctor, takes the one the patient picks, writes it into the booking system, sends the confirmation, and sends the reminder the day before. The front desk finds out because it appears in the calendar, not because anyone told them.

Two design decisions matter more than the technology.

The first is the handover rule. Anything clinical goes to a person. Not "the agent tries and gives up" — a hard rule, written down before launch, that certain categories of message are never answered by the machine. Symptoms, pain, medication questions, anything that sounds like distress: the agent says a member of staff is coming, and flags it. In practice the clinic decides where that line sits, and it usually sits further towards caution than the vendor would like.

The second is that the patient must always be able to reach a person. "Talk to someone" has to work, at any point, without an argument. A system that traps people in a loop does more damage than the phone queue it replaced.

What it would actually cost, and what it gives back

Here is the honest arithmetic, and it is the arithmetic we would run with you rather than for you.

Say the clinic handles 120 inbound contacts a day between phone, WhatsApp and walk-ins. Say three quarters of that is the four routine questions, and say the agent handles most of it — not all, some conversations go sideways and get escalated. That is somewhere around 70 to 85 conversations a day that do not reach the desk.

At roughly two minutes a conversation once you count the interruption cost, that is between two and three hours of staff attention returned every day. Not two staff removed — nobody should sell it that way. Two people who now spend their day with the patients in front of them instead of choosing, forty times a day, between the person at the desk and the ringing phone.

What it costs: a build of one to four weeks depending on how the appointment book works, and a monthly running cost driven by message volume and support. For a clinic of that size it is smaller than one salary. Whether that is worth it is a conversation for the person who signs, not for us to answer in an article.

The part most clinics get wrong first

The single biggest failure we see is starting here at all.

If your appointment book is a paper diary, or a spreadsheet on one receptionist's laptop that nobody else can open, then an agent has nothing to write into. If your price list has not been updated since the last change and three staff members quote three different figures, the agent will confidently quote a fourth. These systems are only as good as the records underneath them, and most clinics do not have those records yet.

That is not a reason to give up. It is a reason to sequence properly: get the appointment book, the price list and the result status into something a system can read first. That work is worth doing on its own merits even if you never add AI on top — and if you do add it later, it takes weeks rather than months.

What we would ask you first

If you run a clinic and this sounds like your reception, the questions we would ask before quoting anything:

1. What does your appointment book actually live in today?
2. How many people can update the price list, and when did it last change?
3. What percentage of your inbound contact is on WhatsApp already?
4. Which categories of message must never, under any circumstances, be answered by a machine?

The fourth question is the one that determines whether this is a good idea for your clinic. If the honest answer is "most of them", we will tell you so.