The complete guide to an AI receptionist for dental practices.

Reception is where a dental practice quietly leaks work. The desk is staffed while you are open, and the people staffing it are also greeting patients, taking payments, chasing recalls and answering the phone. Every enquiry that arrives while all of that is happening waits — and every enquiry that arrives after the door is locked waits longer.

Two things go wrong, and they go wrong differently. Someone in pain does not wait politely. A broken tooth on a Sunday evening sends a person down the search results until a practice responds. The practice that answers is the one that gets the patient. The quieter loss is the shopper — the person comparing two clinics on what a check-up costs and how registration works. In the UK that list starts with whether you take NHS patients. In Australia it starts with which health funds you accept and what the gap comes to. They book wherever they get a straight answer first.

Underneath both sits the repeat-question tax. Opening hours, parking, NHS versus private, what a first visit involves, whether you offer a particular treatment. The answers already sit on your website. Reception recites them anyway, one caller at a time, all morning.

An AI receptionist, in the form Adoomi builds, is a chat assistant on your own website that answers as your practice. There is a ready-made setup for dental practices specifically. It is trained on your website, your published fees and your FAQ pages, and it answers strictly from that approved content. Ask it something you have not published and it says so and hands the conversation to a human rather than improvising. It works around the clock and replies in the patient's own language.

The practical wins are narrow and real. The registration and fees questions get answered from your own pages by the website chatbot, not by a receptionist with a patient at the desk. The Sunday-evening enquiry that used to sit unread until Monday arrives as something your desk can act on. An Australian practice points it at the same pages: the fee page, the new-patient page, the health funds you claim through at the desk.

The useful mental model for a practice is a hard line drawn at the surgery door. On one side is the desk list — the administrative questions a receptionist answers all day without consulting anybody. On the other side is everything clinical, which is a handover, always, with no exceptions and no judgement calls made in chat.

This is also what makes the review step at setup worth the time. You are not tuning a personality; you are deciding what the practice is willing to say in writing, overnight, to somebody it has never met.

Dental enquiries cluster exactly where the practice is not: the evening after work, the weekend, the bank-holiday Monday. That is when a crown comes off, when a parent starts worrying about a child's tooth, and when somebody finally sits down to compare two local clinics.

The assistant keeps answering the desk list through all of it. What it must never do is decide how urgent a symptom is. Instead it surfaces what your own emergency page already carries: your out-of-hours arrangement, the number to ring, the safety net. Then it flags the enquiry to the person you named. The instruction is yours; the assistant only relays it.

That published instruction needs no software at all, and it is worth writing whether or not you ever add an assistant. If your site does not carry one, adapt something like this and put it on your contact page. It works tonight, and it is exactly the content an assistant would later surface. The advice line and the emergency number differ by country, so the template carries the UK and Australian ones.

Most enquiries arrive unusable. "Do you have anything this week?" tells your desk nothing — not who is asking, not whether they are registered, not what for. The point of lead capture is that the assistant asks the boring questions before a human spends a minute on it. Contact details, the need and the context all land on one structured enquiry.

• Name and best contact. Mobile for WhatsApp, plus email.
• The reason in the patient's own words. "Filling fell out on the bottom left" carries more than any dropdown, and it reaches the clinician unedited.
• New or registered. Registration status changes who handles the enquiry and how long it takes.
• NHS or private. Asked as your practice frames it, so nobody is told something your pages do not say.
• Urgency in their own terms. Captured as the patient described it — never rated, scored or triaged by the assistant.

Then the handoff. The patient can continue with your team on WhatsApp, and the original question plus every captured detail travels with them. Nobody opens with "hi, what's the problem?" at someone who has already explained it twice. Email alerts do the internal routing: fees and registration to reception, clinical wording to the clinician you named. Anything safeguarding-adjacent — a child patient, a vulnerable adult — reaches one named person rather than the general inbox.

The enquiry arrived on Sunday evening and was answered on Sunday evening. Your desk opened Monday to a named patient, a question in their own words, and a clinical flag already sitting with the right clinician.

In dentistry an assistant that oversteps is not an embarrassment, it is a risk. So the limits are the product, and they are worth reading before the features:

• No clinical advice, ever. It will not read a symptom, name a cause of pain, comment on an X-ray, suggest a painkiller or antibiotic, or say how soon someone should be seen. Each of those stops the conversation and emails the clinician you named.
• No invented fees. Figures come from your published fee list or not at all. Treatment set at consultation waits for the consultation — the assistant captures the enquiry instead of estimating.
• No booking, holding or moving appointments. It does not touch a diary and cannot see one. It captures who needs what and when they are free; your desk books.
• No NHS promises. Availability changes by practice and by season. It repeats what your page says today and nothing beyond it.
• No phone answering. Restated because "receptionist" implies otherwise: your phone line is untouched today, and the AI phone agent remains waitlist-only.

One more boundary, on data. Chats are EU-hosted, GDPR export and deletion are built in, logs hold no message content, and the chat window sets no tracking cookies. An Australian practice weighs that EU location against the Privacy Act 1988 and the Australian Privacy Principles, as it would any overseas supplier. None of that makes your practice compliant with anything — compliance is your registration, your policies and your processes. It just means the assistant is not quietly working against them.

Two products reach the same practice manager under names a search engine barely separates. A dental virtual receptionist is normally a trained person in another building, answering your line under your practice name, priced by the call or the minute. An AI dental receptionist of the kind Adoomi builds is software on your own website, priced flat. They meet different patients, and a person is better at some of this.

Read down that table and the split is clean. A person covers the spoken call and the improvised answer. Adoomi covers what patients type, which is where the repeat questions live and where the evenings and weekends go unanswered today. That is the half a dental practice can hand over now, for a flat figure, without adding anyone to the desk.

Some practices are phone-first, and the assistant does not change that. Adoomi answers on the practice website, not on the phone line, so a list that rings by habit also needs a telephone answering service. The assistant only meets patients who land on your website, so search traffic has to be there first. Where it is, Adoomi covers the patient who searched instead of ringing. They read the fee page on a Sunday evening, type a question and get an answer there and then, at any hour, without anyone at the desk touching it.

Otherwise, run the arithmetic from your own records rather than anyone's brochure:

1. Count last month's misses. From the phone log, tally the evening and weekend calls nobody answered, then add the website and email enquiries that waited a day or more for a reply.
2. Price the phone side. If you already pay an answering service, multiply those calls by your own per-call or per-minute rate from your tariff sheet — not an assumed market figure.
3. Price the website side. Set that against a flat £8/month — Adoomi's Local plan, 200 message credits. The two catch different people: the service catches whoever rings, the assistant catches whoever searched, landed on your page and never picked up the phone.

Whatever step one shows arriving through the website is the share £8 a month covers. Those patients get answered without the desk touching it, and the fee and appointment questions that fill Monday morning arrive already handled. Plenty of practices run both, a person on the line and the assistant on the site, because each reaches patients the other never meets.

Setup takes about ten minutes: connect your website, review the answers the assistant proposes to give, and add one line to your website. For a practice, the review step is the whole job. Read the out-of-hours wording as a patient in pain would read it. Confirm that anything clinical hands straight over. Check every fee against your published list, and check the NHS wording says what your page says — not more.

Pricing is flat and practice-shaped. The Free plan is £0 with 50 message credits a month, and one credit covers one AI answer. Then £8/month (Local, 200 credits), £20/month (Growth, 500), £75/month (Scale, 4,000) and £250/month (Advanced, 14,000). The full plan comparison has the detail. Hosting is EU-based, with GDPR data export and deletion built in and no tracking cookies in the chat window.