Chatbot for a Dental Clinic in Iraq: Answers and Costs

The same short list of questions, hundreds of times a month

A dental clinic in Erbil or Baghdad does not receive complicated messages. It receives the same short list of questions, over and over, across Instagram DM, Facebook Messenger and WhatsApp. Many of them arrive in the evening and at the weekend, after the reception desk has closed.

Here is that list. Both the ordering and the point about evenings come from clinic inboxes we have worked in. Neither is measured across Iraq, so read them as a starting point and not as a statistic.

Three things are visible in that table. Most of the list is administrative. A small part of it is clinical and must never be answered by software on its own. And the line between the two cannot be drawn in the moment by whoever is typing. It has to be drawn before the system goes live, and built into how the system works.

That is what this page is about. If you want the mechanics of how the replies themselves are written, see chatbots built in Iraqi Arabic and Kurdish Sorani.

The price question is really three questions

“How much is a cleaning” has one correct answer, it is a fixed number, and the clinic already knows it. The system answers it in one message and moves on.

“How much is an implant” does not have one answer. It depends on the case, the brand of the fixture, whether a bone graft is needed, and whether the crown is included. The correct reply is a published range, a plain statement of what moves the price inside that range, the payment terms if you offer them, and an offer of a consultation. It is not a quote. A system that gives a final implant price over Instagram is creating an argument for the day the patient arrives.

Braces sit in the same category, with one addition: the most common follow-up is about duration, not money. Give the usual range in months and say plainly that the treating dentist sets the real number after an examination.

The appointment question

اكو موعد اليوم is a question about your calendar, not about dentistry. The system reads the real availability, offers two or three actual slots, holds the one the patient picks, writes the name and phone number down, and sends a reminder the day before. If nothing is free today, it offers the next two open times rather than saying no. You can see how the booking flow behaves end to end on the workflows page.

This is the part that is easy to underestimate. A missed evening message is not a lost reply. It is an appointment slot that stayed empty the next morning.

The emergency question

Pain, swelling, bleeding, a broken tooth, anything after surgery. The system does not triage these. It does not suggest a painkiller, a dose, a rinse, or a wait. It does one thing: it says a person from the clinic is being contacted now, it gives the emergency number, and it pushes the conversation to a named human with the message history attached.

If the clinic is closed and there is no on-call arrangement, the system says exactly that and gives the nearest emergency option the clinic has approved in advance. Silence at 11pm on a pain message is the worst outcome available.

Clinical content is reviewed by a qualified human before it reaches a patient

This is the requirement that shapes the whole build. It is not a setting. It is the structure.

The system does not compose new clinical sentences while a patient waits. It has an approved answer library, and anything falling outside that library triggers a handover instead of an improvisation. New clinical answers go into a review queue and reach patients only after your dentist has signed them off.

Escalation is triggered on both keywords and intent: pain, bleeding, swelling, fever, pregnancy, a child, any named medication, anything post-surgical, and any message with a photo attached. Patients often send photographs of their own teeth. The system does not attempt to read them. It routes them to a person.

The reason this is structural rather than a toggle is simple. A toggle gets switched off by whoever holds the login on a busy night. A build where clinical answers exist only as pre-approved text cannot be switched off quietly, because there is nothing else for the system to say.

The AI disclosure stays on

Every conversation opens by telling the patient they are talking to an automated assistant, and that a person from the clinic is available. Clients ask us to remove it. The answer is always no, on every build, for every tier.

It is also the thing that makes the handover work. A patient who knows they are talking to software accepts “let me get someone from the clinic” as a normal step. A patient who thought they were talking to the receptionist experiences the same sentence as being brushed off.

Language, and the voice note problem

Replies are built in Iraqi Arabic and Kurdish Sorani, not Modern Standard and not Gulf phrasing. A patient who writes شكد and gets back كم يبلغ سعر has already worked out they are talking to something imported.

Voice notes matter more in dental than in most sectors, because people in pain record rather than type. Voice messages are transcribed and handled through the same rules, which means most of them escalate to a human, which is the correct outcome.

Which channel to start on

Instagram DM and Messenger first. There is no Meta per-message fee on either. WhatsApp Cloud API requires business verification with registration documents, which is doable for a licensed clinic but adds time before anything is live, so it is the second step rather than the first. See WhatsApp Business API setup, handled for you for what that process involves.

For scale: Instagram potential ad reach in Iraq is 21.5 million, 45.4 percent of the population and up 15.0 percent year on year, with Facebook at 20.8 million and Messenger at 14.9 million and slightly down. Source: DataReportal Digital 2026 Iraq, data snapshot October 2025.

What it costs

Growth is the usual fit for a dental clinic. A clinic needs more than one channel, a live calendar connection, and an approved answer library with real depth behind implants, braces, whitening and aftercare. Starter suits a single-channel clinic with a short price list and no calendar integration. Pro is for multi-branch groups.

For comparison, the platform subscriptions people usually price against, in monthly fees alone before any setup work:

Fees as published by each platform on its own pricing page, converted at 1,500 IQD to the dollar. ManyChat cut its free plan from 1,000 contacts to 25 on 2 March 2026, so the do-it-yourself-on-free route is closed. Every price we charge is on the published rate card.

What we need from you before the build starts

The clinical sign-off is the item that delays projects. Everything else is administrative.

What this system does not do

It does not diagnose. It does not quote a final price for implants or braces. It does not replace your receptionist. It takes the repetitive part of the day and covers the hours nobody is at the desk. And we do not promise a number of extra bookings, because we cannot know your traffic, your prices or your competition before we look.

What it does do is answer the patient who messages at 10pm asking شكد سعر التنظيف, and route the patient who messages at 10pm saying سني يوجعني to a named person at the clinic.