Key takeaways
- AI in a dental practice falls into three groups with very different risk: imaging support, clinical notes and front-of-house admin.
- Imaging and note tools may be regulated medical devices, so check the MHRA position and the intended purpose before you buy.
- Phones, messages and recalls are usually the lowest-risk first project because a human still makes every clinical decision.
- Health data is special category data under UK GDPR, so a DPIA and a processor contract come before any pilot.
- Pick one measurable problem, run it for a fixed period and keep the dentist accountable for every clinical output.
Three places AI turns up in a dental practice
Most of the noise about AI in dentistry blurs together tools that have almost nothing in common. A practice owner weighing up a purchase will find it easier to sort every product into one of three groups, because each group has its own regulator, its own risks and its own sign-off.
- Imaging support: software that marks possible caries, bone loss or calculus on bitewings, periapicals and OPGs for the clinician to review.
- Clinical documentation: ambient note-taking that listens to the appointment and drafts the clinical record, or tools that turn a treatment plan into a patient letter.
- Front-of-house admin: answering calls and messages, booking and moving appointments, recall reminders, new patient enquiries and payment follow-ups.
The first two touch clinical judgement directly. The third touches the patient experience and the appointment book, but should never touch diagnosis. That difference decides how much evidence you need before going live and who in the practice owns the decision.
Why the front desk is usually the first project
Demand on dental teams is heavy and uneven. NHS practices in England delivered 37.6 million courses of treatment in 2025/26, still below the 38.8 million of 2019/20[1]. Of those courses, 61% were Band 1 and 11% were urgent treatments[2], and urgent callers are the ones most likely to ring repeatedly, at awkward times, while the team is chairside.
People without a regular dentist have a particularly hard time. Among adults without a dentist who had an urgent need, only 16.0% got an urgent NHS appointment and 37.4% tried but failed; when they could not get one, 69.0% took no further action[3]. For a practice with capacity, every unanswered call in that group is a patient who may simply give up.
Search matters too. Online search was the most common way these patients found a practice, ahead of the NHS website, while personal recommendation barely registered[3]. A patient who finds you through search and then reaches voicemail is easy to lose, which is why call and message handling is often where an AI project pays back first.
Use cases side by side
| Use case | What it does | Regulatory position | Who signs off |
|---|---|---|---|
| Radiograph support | Highlights possible findings on images for the dentist to confirm | Likely a medical device; check UKCA or CE marking and intended purpose | Principal dentist and clinical lead |
| Ambient note-taking | Drafts clinical notes from the consultation audio | May be a medical device if it summarises or interprets | Each dentist for their own notes |
| Treatment plan letters | Turns a plan into plain-English patient information | Not usually a device; content still needs clinical review | Treating dentist |
| Calls and messages | Answers, books, moves and logs requests within set rules | Data protection and consumer law; no clinical advice | Practice manager |
| Recalls and reminders | Sends and follows up recall invitations and reminders | PECR applies to marketing, not to service messages | Practice manager |
Clinical AI: imaging and notes
The MHRA says many software and AI products used in health and care are regulated in the UK as medical devices[4]. For dental imaging tools that is the normal position, and for note-takers it depends on what the product claims to do. A tool that only transcribes is different from one that summarises, suggests codes or flags clinical concerns.
Whatever the regulatory label, responsibility does not move. The dentist who signs the record or accepts a finding is accountable for it, so a tool that speeds up work but quietly lowers the quality of review is a bad trade. Ask any supplier these questions before a demo turns into a contract:
- What is the intended purpose, and does the marking on the product match the way we plan to use it?
- What evidence shows performance on UK patients and on the imaging equipment we actually use?
- How are suggestions displayed, and can a clinician see what the tool did not flag?
- Where is audio or image data stored, for how long, and is any of it used to train the product?
- What happens to the clinical record if the service is unavailable mid-session?
Front-of-house AI: calls, messages and recalls
Admin AI is about getting patients to the right appointment with less waiting and fewer missed contacts. A well-configured tool can answer when every line is busy, take the details of a new patient enquiry, offer slots from the book, move appointments, send recall invitations and log every interaction for the team. Ava, an AI Employee for practice phones, sits in this group: today it answers practice questions in wording you approve, explains how to get help and passes patient matters to reception, while booking directly into clinical systems is still planned rather than live.
The boundary has to be written down. A caller describing facial swelling, difficulty swallowing or uncontrolled bleeding is not a booking request; your rules should route them to a clinician, NHS 111 or 999 straight away. Any front-of-house tool that cannot follow a clear escalation list is not ready for a dental practice.
Data protection before anything goes live
Information about a patient's dental health is data concerning health, which UK GDPR Article 9 treats as special category data that can only be processed when a specific condition applies[5]. That makes the paperwork part of the project, not an afterthought.
Map the data
List what the tool collects (audio, images, names, phone numbers, notes), where it goes and who can see it.
Complete a DPIA
Record the purpose, the lawful basis and Article 9 condition, the risks to patients and the controls you will use.
Sign a processor contract
Make sure the supplier processes data only on your instructions, keeps it confidential and returns or deletes it at the end.
Update your privacy notice
Tell patients plainly which AI tools you use, for what, and how to speak to a person instead.
Set a review date
Check access logs, errors and complaints after the first month, then at least yearly.
Example: a mixed practice chooses its first project
Example (fictional): a four-surgery mixed practice has two receptionists covering phones, the desk and payments. An audit of one fortnight shows that most unanswered calls happen between 8.30am and 10am and over lunch, and that new patient enquiries are the calls most often abandoned.
In this example the owner had been shown an imaging tool and an ambient note-taker in the same month. Neither addressed the problem the audit found. The team instead piloted call and message handling for new patient enquiries and appointment changes only, with emergencies routed to the duty dentist. They agreed three measures in advance: answered calls in the busy hours, new patient enquiries converted to booked assessments, and complaints about the phones. The clinical tools went onto next year's list, to be judged on their own evidence.
What patients are likely to think
Public opinion is cautiously open. Around half of UK adults, 52%, expect AI to have a positive impact on healthcare, yet healthcare is also the area people most often say needs careful management to prevent harm, chosen by 29%[6]. In practice that means patients tend to accept AI that saves them time on admin, as long as they are told about it and can reach a person when they want one.
Your team matters as much as your patients. With 49,079 dentists and 86,562 dental care professionals on the GDC register[7], almost every practice already relies on nurses, hygienists, therapists and receptionists working closely together. Involve them early: they know which tasks eat time and which ones need a human voice.
Measuring whether it worked
Write the measures down before the pilot starts, take a two-week baseline, and compare like with like: the same weekdays, the same staffing and, ideally, the same time of year. Useful measures differ by group.
- Imaging support: agreement between the tool and the dentist on a sample of images, and any findings the tool missed that a colleague later picked up.
- Note-taking: minutes from end of appointment to completed record, and the number of edits each dentist makes before signing.
- Calls and messages: answered calls in the busiest hours, requests completed without a callback, and new patient enquiries that became booked assessments.
- Recalls: recall invitations that led to a booked examination, and opt-outs or complaints received.
Add one qualitative check for every group: ask five patients and two team members each week what felt better or worse. Numbers tell you whether something changed; conversations tell you why, and they often surface a problem, such as a confusing message or a missed escalation, long before it shows in the figures.
A short decision test
- Can you name the problem in one sentence and measure it today?
- Does the tool make or influence a clinical decision? If yes, check its regulatory status first.
- Can a patient always reach a person, and is the escalation route written down?
- Is the DPIA done and the processor contract signed?
- Is there a fixed pilot period with a decision date and an owner?
If any answer is no, pause. AI earns its place in a dental practice by solving one clear problem safely, not by being added everywhere at once.
Frequently asked questions
Is AI radiograph software a medical device in the UK?
Usually yes. The MHRA regulates many health software and AI products as medical devices, so check the product's marking and intended purpose and make sure your use matches it.
Can an AI phone tool give dental advice?
It should not. Front-of-house tools should book, move and log requests and pass anything clinical, such as swelling or uncontrolled bleeding, to a clinician, NHS 111 or 999.
Do we need a DPIA for an AI tool in a dental practice?
In almost every case. The tool will process health data, which is special category data under UK GDPR, so a DPIA helps you show the risks are understood and controlled.
Which AI project should a dental practice start with?
Start with the problem you can measure today. For many practices that is missed calls and enquiries at peak times rather than clinical tools.
References
Care Quality Commission (2026). State of Care 2025/26: dental (opens in a new tab) ↑
NHS Business Services Authority (2026). Dental statistics - England 2025/26 (opens in a new tab) ↑
Office for National Statistics (2026). Experiences of NHS healthcare services in England: September 2026 (opens in a new tab) ↑
GOV.UK (Medicines and Healthcare products Regulatory Agency) (2025). Software and artificial intelligence (AI) as a medical device (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 9: Processing of special categories of personal data (opens in a new tab) ↑
GOV.UK (Department for Science, Innovation and Technology) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
General Dental Council (2026). Registration report – October 2026 (opens in a new tab) ↑