It's 10:47 PM on a Wednesday. A 38-year-old woman has eaten light because a molar has been bothering her for three days. She picks up her phone and searches "emergency dentist [city]". She finds four practices and calls them in order. The first: voicemail. The second: rings out for fifty seconds. The third has a website with a contact form she fills in — she'll get a reply tomorrow morning. The fourth has a chat on its site that answers right away, explains that the practice keeps an emergency slot every morning at 9 AM, and offers to book her one for Thursday.
She picks the fourth. She didn't pick it because it was the cheapest or had the best reviews. She picked it because at 10:47 PM it was the only one listening.
The next day the other three practices will see a missed call, an empty voicemail box, a contact form they'll get to with a one-hour delay. None of them will ever know that patient existed. She existed at 10:47 PM, and she went somewhere else.
The number that changes how you read the problem
A 2026 study across thousands of dental practices put a black-and-white figure on something everyone suspected but nobody measured: dental practices miss an average of 35% of incoming calls. For smaller practices — one front-desk person, often part-time — it climbs to 38%. Inside that figure lives everything: lunch, the patient on the other line, the Thursday afternoon team meeting, the eight calls that hit the line at the same time on Monday morning right after 9.
When someone calls a healthcare practice for the first time and gets no answer, 78% don't leave a voicemail. They call the next clinic on the list. For every new patient lost this way, the lifetime value walked out the door is somewhere between €15,000 and €25,000 — because a good patient doesn't visit once, they visit for fifteen years and bring their family.
A second figure piles on. In Italy, 70% of dental appointments are still booked by phone. But of those calls, a large share — about 28% according to international studies — comes outside business hours: evenings, nights, weekends. For a practice open 40 hours a week, those are hours when no one is there to answer. The math adds up easily: between in-hours calls lost to saturation and after-hours calls that are unmanageable in the first place, the gap reaches 40%.
The fourth practice — the one that captured the 10:47 PM patient — didn't have a sharper front desk than the others. It had a system that, at 10:47 PM, was at the office.
The ten questions your front desk knows by heart
Anyone who runs a dental front desk for more than three months develops an underrated skill: answering out loud while their head is already on the next patient file. How much is the first visit. Are you open Saturdays. Do you do digital impressions or silicone ones. Are you in network with my insurance. Can I pay in installments. My son is 6, do you take him. How long does an implant last. Can I come without an appointment for a toothache. Should I bring my health card. I changed my number, can you call me on the new one.
That's ten questions. The same ten every day. None requires professional judgment, none has an answer that varies from patient to patient. And yet, for every single one, the receptionist has to interrupt what she was doing, take the call with grace, repeat in a quiet voice the same line she's already said thirty times today, and pick up where she left off — hoping the second line hasn't started ringing and someone hasn't walked into the waiting room.
A conservative estimate, based on direct measurements at mid-sized Italian practices, puts the volume of these "informational" calls between 30 and 40% of the total. It's pure receptionist time — not greeting time, not agenda-management time, not time for the patient standing at the front desk. It's repetition time.
40% isn't a guess, it's an addition
When this article talks about "cutting 40% of appointment calls" it doesn't mean having fewer appointments. It means the opposite: handling half of the phone traffic differently, so the phone rings only for requests that genuinely need a person.
The 40% breaks down in two slices. A first slice — about 28% — is after-hours calls, today lost or postponed. A second slice — about 12% — is repetitive informational questions that can be handled in self-service: hours, indicative prices, insurance partners, paperwork, directions, prep for a first visit, callback requests. The total is forty.
Nothing is taken away from the work that requires human skill. Negotiating an awkward time slot, reassuring an anxious patient before an extraction, walking through a treatment plan: those stay where they belong. What gets removed is the background noise. The constant interruptions. The line "good morning Bianchi dental practice how can I help you" repeated thirty times a day just to say "Thursday afternoon we open at 2:30 PM".
What an AI assistant actually does for a dental practice
A modern conversational assistant, built on a language model, is not a rule-based bot with menu buttons. It doesn't ask "press 1 for appointments, 2 for hours". It understands a question written in plain words and answers in the same way, drawing on the information you've predefined.
On the practice's website or on a dedicated page, to the 10:47 PM patient on Wednesday who writes "I have a molar that's been hurting for three days, can you see me?" the assistant answers with the practice's standard emergency procedure: "For emergencies we keep one slot every morning at 9 AM. I can offer Thursday at 9:00 or Friday at 9:30. Confirm?". It captures name, contact, a brief description of the symptom. The next morning the front desk finds the request in the agenda already qualified, confirms with a thirty-second SMS, prepares the chart.
For the father writing on Monday at 2:30 PM "what Thursday afternoons are you open?" the assistant answers directly: "Thursday afternoons we're open from 2:30 PM to 7 PM". No call. For the woman asking "are you in network with Unisalute?" the assistant confirms with the list of insurers the practice has direct agreements with. For the patient saying "I lost my treatment estimate, can I get another copy?" the assistant routes the request to the front desk with a note — because that's a request that needs a manual operation.
A crucial point, especially for a healthcare practice: you control what the AI can and cannot say. It never diagnoses. It doesn't comment on symptoms beyond recognizing them and triaging urgency. It doesn't discuss individual treatment plans. It doesn't comment on materials, brands, or clinical choices. Anything that requires professional judgment is routed to you or to the front desk, every time. This isn't a technical detail: for a dentist it's a matter of professional ethics.
Three scenes at three moments of the week
6:48 AM on a Monday in March. A long-time patient has noticed a small swelling under his last molar, he's worried, he's heading out to work. He writes in the practice's chat: "good morning, I have a swelling near my last lower tooth, can I come by today?". The assistant recognizes the possible urgency, offers the day's reserved emergency slots (today at 9, today at 2:30 PM), confirms the 2:30 PM slot. It captures contact and description. At 8:30 AM the receptionist opens the agenda and already has a case to handle — the patient already has a time, all that's needed is an SMS confirmation.
Saturday afternoon at 4:20 PM. A mother is planning the next checkups for her two children and remembers the practice was in network with the company health fund. She writes: "are you still in network with Fondo Est? And how do I book two checkups for two kids aged 7 and 9?". The assistant confirms the partnership, explains that a first pediatric visit takes about 30 minutes, offers two consecutive slots in the same afternoon, captures the data. Monday morning the receptionist finds two appointments already booked for two new pediatric patients. No phone picked up, no email written, no call returned.
Wednesday at 2:50 PM. A patient who hasn't been seen in four years wants to come back but doesn't remember hours, prices, and whether the hygienist is the same as before. He writes: "I'd come back for a checkup after a long time, I need a checkup and a cleaning. How much does that usually cost?". The assistant answers with the price range for the checkup (free or €40 depending on the practice), confirms the price for the cleaning, offers three slots next week. It captures consent for callback. The next day the front desk knows there's a returning patient to welcome — and already has all the elements not to call him back to reconstruct the case.
In none of the three scenes did the phone ring. In none of the three scenes did the receptionist have to interrupt what she was doing. In all three, patients got an answer in real time and felt they were being listened to.
What to check before picking one for the practice
Not all AI assistants are mature enough to live on a healthcare practice's website. For a dental practice, five criteria really matter — and they're not the same as the ones that matter for an e-commerce site or a real-estate agency.
System prompt control has to be yours. The provider must let you write exactly what the assistant can and cannot say. For a healthcare practice this isn't an option, it's a precondition. A virtual receptionist that gives advice on how to treat an infection or expresses opinions on implant materials isn't a partner to the professional — it's a deontological problem that materializes at the first misunderstanding. The system has to have an explicit line: "I don't provide diagnoses or clinical guidance, I route to the professional".
Data residency has to be in the EU. Patients who write in chat share names, phone numbers, symptoms — personal data, and in some cases health data. The provider has to store them on EU infrastructure, give you a Data Processing Agreement compliant with Article 28 of the GDPR, document deletion policies. Choosing a provider running on US servers without additional safeguards means importing a compliance problem you didn't have before.
EU AI Act compliance is mandatory from August 2, 2026. The user has to know they're talking to an AI. A serious assistant declares it in the first message, clearly and unambiguously, without trying to simulate a human being. For a healthcare practice it's also a transparency issue with the patient.
The ability to upload your own documents — price list, paperwork, internal FAQs, info sheets on the treatments you offer — makes the difference between a generic bot and a tool that feels like part of the practice. Already have a sheet with first-visit prices? Upload it and the AI uses it as a source. Have a leaflet on prepping for a professional cleaning? Upload it and the AI answers accordingly, without you having to rewrite anything in a proprietary editor.
The price has to be sustainable even in a slow month. Today complete SaaS plans start at €32/month, with a 14-day free trial and no credit card required. For a practice billing €20,000 or €30,000 a month, it's the cost of one professional cleaning. For one billing three times that, it's not even a line item.
It doesn't replace anyone. It removes the requests that don't deserve a person.
The most common fear when automation comes into a healthcare practice is that it hollows out the human relationship. The experience of the first practices that adopted an AI assistant says the opposite. The receptionist who no longer answers thirty times a day "what time do you open?" has more time to greet the person walking in with a worried face about an upcoming visit. The dentist who at 9 AM finds four already-confirmed appointments in the agenda, instead of four calls to return, starts the day treating.
The AI doesn't replace the conversation. It doesn't replace the receptionist's voice saying "come in, doctor, Mr. Rossi is on the line". It doesn't replace any of the interactions that make a patient feel well taken care of. It replaces the void. The voicemail saying "call back in office hours", which is the surest way to lose a new family. The contact form saying "we'll get back to you within 48 hours", which is the surest way to lose the urgent patient.
The difference isn't between a human practice and an automated one. It's between a practice that always answers and a practice that answers when it can. In 2026, even for dental patients, it's increasingly clear which of the two gets chosen.
Want to see how an AI assistant would work for your dental practice, configured with your prices, your insurance partners and your own rules? Paste your site URL on iperchat.ai and try it in 30 seconds — free, no signup.