AI Receptionist for UK Dental Clinics: NHS, Private and Hand-Offs

Set up a UK dental clinic's AI receptionist by writing routing rules before you buy: private and routine bookings the AI may complete, NHS eligibility, charge and urgent questions it passes to staff, and clinical calls it escalates. Add a spoken AI and recording notice, and test with your own diary.

The vendor pages tend to dominate search results and describe features. What they leave out is the decision only you can make: which callers get which treatment. A mixed NHS and private practice has two rulebooks running in one phone line, and an AI will follow whichever one you give it, including a half-written one.

Follow me on Instagram@sagnikteaches

The work runs in six steps: sort your calls, write the NHS and private rules, set the notice, connect the diary, test, then go live in stages. For the money side, the companion tutorial on what an AI receptionist costs a dental practice has the wider comparison. Nothing here is legal advice; confirm data-protection points with your own adviser.

Connect on LinkedInSagnik Bhattacharya

Step 1: sort a fortnight of calls into four groups

Print or export the phone log for ten working days. Tick each inbound call into one of four groups. The sample below is invented for a four-surgery mixed practice, to show the shape.

Subscribe on YouTube@sagnikteaches

Your own split will differ. The point of the exercise is that the AI's job is the first row and a safe version of the second. If the first row is under a quarter of your calls, a smarter voicemail may be enough.

Step 2: write the NHS and private routing rules

Write these as plain if-then lines the vendor can load. Here is a filled-in draft, which you should adapt to your contract and your clinicians.

IF caller wants a check-up AND says they are a private patient
   -> offer private check-up slots; book; send confirmation text.
IF caller asks "do you take NHS patients?"
   -> give the approved sentence: "We see NHS patients on [days/with
      [dentists]]. For registration and eligibility I'll take your
      details and the team will call you back within [one working day]."
IF caller asks the cost of NHS treatment
   -> read the published patient-charge wording on the practice website;
      do not estimate for a specific course of treatment.
IF caller is an existing NHS patient wanting a routine recall
   -> book the recall slot type configured as NHS.
IF caller mentions pain, swelling, bleeding, injury, fever or trouble
   swallowing -> follow the urgent route (Step 3).
IF caller asks for a complaint, a refund or a dispute -> transfer.

Three decisions sit inside those lines and need a named person. First, who decides whether a new caller can join the NHS list, since the AI must not promise a place. Second, whether NHS and private slots are separate appointment types in the diary; if they are mixed, the AI cannot tell them apart. Third, what wording the practice has approved for charges, because the AI repeating a figure from memory is a common way to give a wrong one. Use your published charges page and nothing else.

A before-and-after example of a rule that went wrong

An illustration. Before: the practice let the AI "book any available slot" for new callers. A caller asked for a filling, said she was on the NHS, and the AI booked a private dentist's 20-minute slot, because it was the first free. The reception team found it at 5pm and had to ring her. After: the rule says new NHS-seeking callers are never booked directly; they get a callback task with name, number and reason. One extra line stopped the repeat.

Step 3: the urgent route, in three layers

1. Trigger words. List the phrases patients actually use: "face swollen", "can't open my mouth", "bleeding won't stop", "knocked my tooth out", "temperature", "can't swallow".
2. In hours. A warm transfer to a nurse or receptionist, with the AI passing a one-line reason so the patient does not repeat it. If nobody answers in 30 seconds, take a message flagged urgent and text the duty clinician.
3. Out of hours. A fixed, tested message: your emergency arrangement, and where to go for serious symptoms such as swelling affecting breathing or swallowing. The NHS website tells patients they can call 111 or use 111 online, and to go to A&E for severe swelling with difficulty breathing or heavy bleeding that will not stop; private practices often run their own on-call rota. Write down which applies to you.

I would not let the AI judge severity. Its job is to detect that the call sounds urgent and hand off. ViveoAI's pricing page says its product does not diagnose, triage or advise on treatment, and that urgent and clinical calls are escalated under rules the practice approves before go-live. That is the right division of work whichever vendor you pick. The deeper after-hours question is in whether AI can take emergency calls.

Step 4: the opening notice and UK GDPR basics

Callers must be told they are speaking to an AI, and that the call is recorded or transcribed, in plain words at the start. A workable version:

"Thank you for calling [practice]. I'm the practice's automated
assistant, an AI. This call is recorded and transcribed to manage
your appointment; details are on our website privacy page. I can
book appointments and take messages, or you can say 'reception'
at any time to speak to the team."

Behind the notice, three pieces of paper. A contract with the vendor that makes it your processor, with instructions, security, deletion and sub-processors set out. An update to your privacy notice describing the AI assistant, recording, retention and the lawful basis. And a short written risk assessment of putting health-related calls through a third party. DentalFlow's page says it offers UK data residency, a full data processing agreement and encrypted recordings; those are vendor claims, so ask for the agreement itself and for the list of sub-processors. Dental information is health data, so ask also what happens to audio after transcription and how long summaries are kept.

Ask where a transcript ends up

An AI call produces up to four records: the audio, the transcript, a summary email and an entry in your diary. Each can sit in a different system with a different retention period. Draw them on a sheet of paper, mark who can open each, and set one retention rule for all four, such as 90 days for audio and transcript, with the diary entry following your normal patient-record rules. If you keep the summary email in a shared inbox, that inbox is now part of the data map.

Step 5: connect the diary, and be honest about the level

Dentally describes itself as cloud-based practice software with NHS integration, and it runs a partner marketplace of integrations. ViveoAI lists Dentally, EXACT, CareStack, Sensei and Aerona among the software it works with. Neither statement tells you what your clinic can do on day one; integrations come at three levels.

• Message only: the AI emails or texts a request; staff book it.
• Look-up: the AI finds the patient and confirms an existing appointment.
• Read and write: the AI offers free slots and writes the booking in.

Ask the vendor to book, move and cancel a test patient in your own Dentally account while you watch. ViveoAI notes that new practice-management integrations are quoted separately and are excluded from its free pilot, so ask early whether yours is already supported. Check appointment types too: if "NHS check-up" and "private check-up" are separate in the diary, the AI must pick the right one from what the caller said.

What UK vendors publish

Prices below are as published on vendor pages on 4 November 2026 and may change. Whether tax is included is not stated on the pages I read, so ask.

DentalFlow's page gives a 30-day money-back guarantee and says no long contracts, with cancellation any time after month three; extra locations add £100 to £250 each to its setup fee. ViveoAI says month-to-month and a 14-day pilot of up to 100 minutes. Compare like for like: DentalFlow's Starter covers business hours only, so it will not answer at 7pm.

A worked example: a mixed practice, first month

Illustration only. A three-dentist practice with two NHS-heavy days and three private-heavy days gets about 400 calls a month. The manager counts 55 unanswered calls, of which perhaps 20 were new enquiries and 12 were private. Suppose the AI is on a £349 overflow plan and books 8 of those 12 private enquiries at an average first-visit value of £110, a figure to replace with your own.

• Recovered private value: 8 x £110 = £880.
• Plan: £349. Margin before staff time: £531.
• NHS-leaning enquiries: 8 callback tasks created, handled by reception the next morning.

Two lessons come out of the numbers. The NHS side shows up as work, not revenue: the AI turns a missed call into a callback task, which helps patients but adds to the morning list. And the private side pays for the plan only if the AI books, not merely answers. Track "private bookings from AI calls" separately from "calls answered", as the 90-day measurement guide suggests.

A sample NHS call, read line by line

An invented transcript, with the points I would check in each line.

AI: Thank you for calling. I'm the practice's automated assistant, an AI...
Caller: Hi, I'd like to register as an NHS patient. Are you taking people on?
AI: Thanks for asking. We see NHS patients on some days, and I can't confirm
    a place myself. May I take your name, number and postcode so the team can
    ring you back by 10am tomorrow?
Caller: It's [name], and my number is... it's a bit urgent, I've had a
    toothache since Sunday.
AI: I'm sorry you're in pain. Let me put you through to the team now.
    I'll tell them it's a toothache since Sunday.

Three checks. The AI did not promise a place on the list, which is correct. It spotted "toothache" mid-call and changed route, which depends on your trigger list covering words that arrive after the first request. And the transfer passed the reason, which the receiving person must see on screen. If the caller instead says "no, it's fine, I just want registration", the AI should still log a callback. A good test is to read this exact script aloud to the vendor's demo and see whether the route changes when the toothache appears, not just when it is the first thing said.

The practice sheet the AI answers from

Alongside the routing rules, the AI needs a short sheet of approved facts. Fill it in once and review it each quarter. An illustrative version:

Opening hours: Mon-Fri 8.30-17.30, Thu to 19.00, Sat closed
NHS sessions: Mon, Tue and Thu with Dr A and Dr B; list status: callback
Private: all days; new-patient examination 40 minutes
Hygienist: Wed and Fri; private or NHS referral by dentist only
Payments: card and bank transfer; payment plans: staff to discuss
Cancellation: 24 hours' notice; fee may apply (staff decides)
Access: step-free entrance at the rear; parking: two bays
Never answer: treatment advice, medication, prices for specific plans

Keep the "never answer" line in the sheet as well as in the rules, so the instruction exists twice. When a fee or day changes, edit the sheet on the same day, and ask the vendor how quickly the change reaches the live agent. If the answer is "next sync tomorrow", plan changes for the evening.

Staffing sums: what the AI frees up and what it adds

An AI receptionist is not free labour, because someone supervises it. A rough weekly time budget for an illustrative clinic in the first month:

• Reading ten transcripts: 40 minutes.
• Working through callback tasks the AI created: 90 minutes, replacing roughly the same time previously spent on voicemail.
• Updating the practice sheet and rules: 20 minutes.
• Vendor check-in or ticket: 15 minutes.

That is about three hours a week, falling to one or two by month three. Against that, reception stops being interrupted by the phone during check-in, which is the benefit front desks tend to report first, though I would measure it with a simple count of calls dropped during peak times, not a feeling. For the staffing alternative, AI against more front-desk hours sets the sums side by side. ViveoAI's pricing page compares £25k to £35k a year per full-time hire with a monthly plan, which is a claim worth recalculating with your own wage and on-cost figures.

Where this fails in a mixed practice

Patients who switch status mid-call

"I'm NHS, actually, can I have the cheaper one?" A caller moving from private to NHS halfway through a booking creates a mismatch between appointment type and funding. Make the rule explicit: any change of status after a slot is offered ends the booking and creates a callback task.

Children, carers and consent to share

A parent books for a child, a daughter calls for her father, a carer for a resident. Decide what identity checks your practice requires before the AI discloses or changes anything, and let it take messages only where the check cannot be done by voice.

Accents, names and noisy lines

Names and dates of birth are the fields most often misheard. Read transcripts for spelling errors, and have the AI read details back, spelling the surname letter by letter and asking "is that right?", before it confirms. ViveoAI says its voice can be tailored in tone and regional accent; test with callers from your own patient base, not the demo voice.

A call that is really a complaint

Complaints about charges often begin as a routine question about a bill. Use trigger phrases such as "charged", "refund", "not happy" and "complaint" to transfer, and make sure the transfer reaches the practice manager, not a general line that rings out.

Step 6: test, then go live in stages

1. Test calls (day 1 to 3). Twenty scripted calls: 8 private, 5 NHS, 4 urgent, 3 odd ones (a supplier, a wrong number, a complaint). Record what the AI did each time.
2. After hours only (week 1 to 2). Forward calls after closing. Read every transcript the next morning.
3. Overflow in hours (week 3 to 4). Calls ringing more than four times, or when all lines are busy, go to the AI.
4. Review (end of month 1). Count wrong bookings, missed urgent flags and patient complaints. Decide whether to widen or tighten.

The hand-off should be a feature you can hear. In test calls, say "I want to speak to someone" and time the response. If it is more than a few seconds, or the AI argues, change the setting before go-live. The call scripts and escalation rules tutorial has wording for the awkward moments.

Checking it worked after 30 days

At the end of month one, put five numbers on one page and compare them with the fortnight of calls you sorted in Step 1.

• Calls answered by the AI: how many, and how many of them were completed without a person.
• Bookings written to the diary: split private, NHS-routine and recall. Check a sample of ten against the diary for the right type and dentist.
• Callback tasks created: and how many your team closed within the time the AI promised callers.
• Urgent transfers: how many, how long each took, and whether any urgent call was missed. One missed urgent call is a reason to tighten the trigger list at once.
• Patient comments: anything said at reception about the phone, good or bad.

If bookings are correct and urgent transfers are clean, widen the AI's hours. If callbacks are piling up unanswered, the problem is staffing the morning list, not the AI, and adding more AI calls would make it worse. That is the usual reason a good pilot sours in month three.

Scripts for the three moments that decide patient trust

The transfer. "I'll put you through to the team now. I've told them you're calling about a swollen face, so you won't need to explain again." The summary has to arrive with the call, or the promise is false.

The callback promise. "I've passed your details to reception, who will ring you before 10am tomorrow. Is this the best number?" Do not let the AI promise a time your team cannot keep; set the time in the configuration to match your actual morning routine.

The decline. "I can't advise on treatment, but I can book you with the dentist or take a message for the clinical team." Short, honest and not apologetic.

Who owns the system once it is live

Name one person, usually the practice manager, to read a sample of transcripts weekly for the first two months, update the FAQ sheet when fees or hours change, and be the single contact for the vendor. Without an owner, the rules drift, and the AI keeps telling callers last year's answer. If you are comparing vendors that book straight into the diary, the Weave, Zo and generic agent comparison may help, and dental AI mistakes on consent and over-automation is a good checklist to read before signing.

Practical questions from clinic owners

Can an AI receptionist book NHS appointments?

Technically it can if it writes into your practice software, but the rules about who qualifies for NHS care, charges and urgent access are yours to set. Most mixed practices let the AI book private and routine slots and pass NHS eligibility questions to staff.

Does Dentally work with AI receptionists?

ViveoAI lists Dentally among the systems it works with, and Dentally runs a partner marketplace of integrations. Whether a particular vendor reads and writes your diary depends on the integration level, so ask for a live test booking in your own account.

What must the call-recording notice say?

That the caller is speaking to an AI, that the call is recorded or transcribed, why, and how to reach a person. Put the longer explanation in your privacy notice and link to it by text or on the website.

Should the AI take emergency calls?

It should recognise them and route them, not manage them. Define urgent phrases, an in-hours transfer, and an out-of-hours instruction that matches your NHS or private emergency arrangements.

How long does set-up take?

Vendors quote around two to four weeks from signature to live. DentalFlow says two weeks; ViveoAI says most practices are live within around four. Add time for your own policy writing.

Further reads

• Which Dental Front Desk Tasks Can AI Take Over? — Which front-desk tasks AI can take over.
• How Dental Practices Use AI for Recalls and Appointment Reminders — Recall and reminder messages after calls are covered.
• What to Do When an AI Receptionist Gets a Booking Wrong — What to do when a booking comes out wrong.
• Winning Back Lapsed Dental Patients With AI Messages — A follow-up job once calls are handled.
• Explaining Dental Treatment Plans to Patients With AI — Helping patients understand plans in writing.
• AI Receptionist for Dentists in Ireland: GDPR, Disclosure and Cost — The same questions under EU law.
• AI Voice Agent for a Dentist: Minutes, BAA and Exit Terms to Negotiate — The contract terms a dental practice should check before signing for an AI voice agent: data agreement, minutes, consent, number, export and exit.
• Virtual Receptionist for Doctors: A Worked Call Flow and Script Set — A call flow, opening script, emergency wording, hand-off rules and check-call routine for a human or AI virtual receptionist.
• AI Notes for Dentists: Hygiene, Restorative and Emergency Visit Templates — One AI note template per visit type: hygiene, restorative and emergency, with sample outputs, review steps and the consent points to settle first.
• AI Receptionist for Medical Clinics With Two Sites: Routing and Hand-Offs — Route calls between two clinic sites with one menu or two numbers, site fact sheets, closed-site transfer rules and a shared calendar check.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ViveoAI pricing and home pages, DentalFlow pricing section, Dentally website (NHS integration and partner marketplace), NHS website page on emergency dental appointments, checked 4 November 2026.