AI for dentists and dental practices: recalls, quotes and scheduling

The AI for Dentists delivers the most value in the steps between the dental chair and reception: hygiene recall messages sent at the interval decided for that patient, the treatment-plan quote the patient takes home without calling back, the crown returning from the dental laboratory that needs an appointment, and messages arriving while the assistant is at the dental unit. These recurring tasks are already recorded in the practice software. An AI assistant connected to the calendar can prepare or complete them according to the practice’s rules.

Stay out of the clinical part. A generic chatbot must not read an orthopanoramica, evaluate a swelling or collect consent to an implant. The Italian law on artificial intelligence puts it in writing: in the health sector AI systems "constitute a support", the decision does not pass to them and the patient has the right to know when they are used (law 132/2025, art. 7).

Who answers the phone while the assistant is at the chair

In many studies the same person passes the instruments to the meeting and answers the phone, and not by disorganization.Dental Study Assistant (ASO), transposed with the DPCM 9 February 2018, includes assistance to the dentist ‘during clinical performance’ and, in the same profile, activities related to ‘accepting customers and managing the secretariat and relations with suppliers’. The course for the qualification lasts at least 700 hours, 400 of which are internships: it is staff trained to be at the meeting, and each phone call takes it away mid-sitting.

An AI assistant on WhatsApp, on the site chat or on the phone can take requests that do not need her: confirm or move an already set session, give schedules and directions to arrive, say with which health funds the study is agreed, respond to "when do I have the next hygiene?". For the entry the pattern is that of a AI switchboard, for messages that of a chatbot WhatsApp.

Booking from scratch is another matter. In a dentist's office the capacity is counted in gatherings and operators: hygiene goes on the hygienist's chair on days when there is, a first visit may require an orthopanorami, an implant surgery occupies a long band and a dedicated assistant. The prudent choice is to leave the AI assistant only the types of appointment you described, with durability and armchair, and keep the rest in the reception.

Then there is the list of what immediately passes to a person: pain, swelling, a broken tooth after a fall, bleeding after an extraction, questions about a drug. The assistant asks no clinical questions and advises nothing, recognizes the request and turns it around, with a number to call if the study is closed. The Recommendations of the FNOMCeO ICT Group on emails with patients, which the document proposes as a starting point also for WhatsApp, they go in the same direction: ask not to use these channels for emergencies, set days, times and a maximum response time, send an automatic confirmation of receipt.

Hygiene calls: Intermission is decided by the dentist, not the software

The six-month recall equals for all is a habit more than a clinical rule. The guidelines of the NICE, the British institute that draws them up for the English health service, ask to set the interval for each patient according to the risk, from a minimum of 3 months to a maximum of 12 under eighteen years and 24 for adults, and to record it in the documentation of the study. For an automation it means one precise thing: the interval is read from the patient's card, where the visitor wrote it.

Even those who perform the session work on the recommendation of the clinician. The Regulation on the profile of dental hygienist (D.M. 137/1999) provides that it operates ‘on the advice of dentists and surgeons licensed to practice dentistry’. Automation complies with this order: it takes the deadline, searches for the hygienist's free spaces, sends the message and records the response. The general logic of automatic calls is explained in the guide to recall customers with AI.

Think of a study with a hygienist present two days a week. Every Monday, automation pulls out patients with the recall expiring, proposes to each two times in the days of the hygienist and updates the agenda when the patient chooses. Those who do not respond within the time limit you set end up in a short list that the assistant calls back to voice, and the front desk stops scrolling through lists of names.

Be careful what you put in the message. The Privacy Guarantor clarified that the healthcare professional kept secret "must no longer request the consent of the patient for the treatments necessary for the healthcare provision", while they ask for explicit consent, among others, the treatments made by healthcare professionals for commercial purposes (clarifications of 7 March 2019). A reminder that adds "this month whitening costs less" changes nature: it becomes a commercial message, on WhatsApp it ends up in a marketing category model (this is explained by the guide to WhatsApp Business templates) and contains a discount, i.e. the kind of content on which the Health Advertising Act puts a limit. Keep the two messages separate.

The estimate of the plan of care that the patient brings home

A dental care plan has phases, sessions, and costs that the patient wants to reread calmly, often with someone in the family. Many quotes remain on the table at home: the patient does not say no, he simply does not call back. It is the flow where a well-written reminder helps and where a commercial tone does the most damage.

The rules are already there. The Medical Deontology Code, which applies to doctors and dentists, asks Article 54 to communicate "preventively to the assisted person the fee". Law 219/2017 wants the documented informed consent "in written form or through video recordings" and inserted in the medical record, and establishes that "the time of communication between doctor and patient constitutes time of care". The AI can prepare that interview, not take its place.

The technique of reminder is that of automatic follow-up, with a more sober register: no invented deadline, no time offer, always an easy way to talk to a person.

Crowns, prostheses and bites: when work returns from the laboratory

Crowns, bridges, prostheses and bites made by the dental laboratory are tailor-made medical devices. The European Medical Device Regulation provides that they do not bear the CE marking but a declaration from the manufacturer, with the name of the person who prescribed them and the indication of the patient for whom they are intended, which is ‘made available’ to that patient (Article 21 and Annex XIII). In the studio all this becomes a round of dates: fingerprint or scan, sending, re-entry, proof, delivery.

The ride breaks down at predictable points. The work comes back and no one calls the patient, or the test was already set and the lab is late. An automation reads the lab emails, links the order number to the patient card and updates the status; when the assistant marks the job as arrived, the patient receives two times for the test on the right chair, and on delivery the study receives the reminder for the statement to be made available to the patient.

A concrete example: a patient waits for a bite. The lab confirms the shipment with an email that reports the order number, the automation moves the status to "incoming" and prepares the message for delivery. If the lab reports a delay, the patient knows it before leaving the house, not in the waiting room.

X-rays, OPTs and medical history: where generic AI does not enter

Endoral, orthopanoramic and cone beam are health-related data, one of the particular categories of Article 9 of the GDPR. Uploading them to a public chatbot to get an opinion means handing them over to a provider without a contract for data processing and without knowing where they end up: that's the problem described in Business Data and AI.

Software that analyzes images to report tooth decay or bone loss exists, but they are another category of product. For the Medical Device Regulation, software providing information for diagnostic or therapeutic decisions is at least Class IIa (Annex VIII, Rule 11), and AI systems of this type subject to the verification of a notified body are at high risk for AI Act, with obligations that the Digital Omnibus on AI moved to August 2, 2028. Law 132/2025 cited at the beginning adds that the decision remains "always remitted to the medical profession".

The same criterion applies to questionnaires. If you automate your medical history before the first visit, the questions are written to you by the dentist, one by one, and each one must serve the cure. The Guarantor has sanctioned with 20,000 euros a dentist who asked every patient, already in acceptance, if he had infectious diseases such as HIV, for violation of the principles of Article 5 of the GDPR, including minimization (Ordinance of 10 June 2021).

Post, before and after photos, reviews: what AI can write for the studio

A language model left free writes like an advertiser: "the smile of your dreams in one day". For a dental practice is the wrong register, and there is a precise norm. The law 145/2018, art. 1, paragraph 525, in the text in force, applies to private health facilities and to members of the registers, including dental companies, and excludes from their communications "any element of an attractive and suggestive character, including communications containing offers, discounts and promotions, which may result in the improper use of health treatments". The code of ethics adds to Article 56 that advertising must be ‘true, correct and functional to the object of the information’, and entrusts its control to the competent Order by territory.

Use AI to draft an article on how to prepare a hygiene session or what to expect after an extraction, then reread each sentence. The photos before and after of a recognizable patient are health data: the Guarantor has sanctioned with 20,000 euros a doctor of aesthetic medicine who had published them on Instagram without consent (measure of 12 December 2024). The case was not about a dentist, but photos of a smile before and after a treatment pose the same problem.

For reviews on Google the rule to be written in the instructions is only one: the answer never confirms that the author is a patient and does not cite treatments. With that constraint, the demo for Respond to Google reviews with AI It also works for a dental practice.

Which flow to start from in your dental practice

The four streams don't weigh the same in every study. The table helps you choose the first one by looking at the data you need and who sets the rules.

1. Look at what your management software already does. Many dental practice management systems cover recalls and reminders: AlfaDocs indicates automatic calls and communications via email, SMS and WhatsApp, XDENT reminders and recalls sent on WhatsApp by the management software, and a voice AI assistant to fill in the clinical record and treatment plan. Check your software’s official page: custom AI helps where the software stops, namely freely written messages, laboratory emails and treatment plan summaries.
2. Choose a stream only, the one that weighs the most on the reception today, and notes how it goes now before touching anything.
3. Fix the data side. Contract for the processing of data with the provider (Article 28 of the GDPR), place of data retention, updated information: the points to be checked are in contract with an AI provider.
4. Write with your dentist what passes to a person and what phrases the assistant never uses: diagnosis, clinical reassurances, discounts.
5. Have the assistant tell him he's an AI. The AI Act calls for systems that communicate with people to inform them that they have been interacting with an AI since August 2, 2026; the other obligations are in AI Act for an SME.
6. Measure for a few months and add the second stream only if the first one worked.

The framework for all private health is on the page AI for Private Health. If you want to understand which of these flows is convenient in your studio and how to link it to the management software that you already use, Giallo Studio starts from the map of a single process and tests it in small before extending it.