At 08:00 on a Monday the phones at most practices light up at once. Two people on reception, a queue at the desk, someone asking where to park, and a line of callers who mostly want the same three things: an appointment, a change to an appointment, or an answer to a question that's already on your website. An AI medical receptionist takes the calls in that queue that are pure admin, so your team can deal with the ones that need a person.
This page is for owners and practice managers at private GP clinics, dental practices, physiotherapy and therapy clinics, opticians and veterinary practices. It explains what the phone and messaging traffic at a clinic actually looks like, which parts an AI can safely take, and which parts it must never touch. It also covers the data protection questions you need to answer before any patient detail goes into an AI tool, Klugon included. We'd rather you read that section and decide it isn't for you yet than switch something on that your data protection officer would object to.
One rule runs through all of it. The assistant is a receptionist, not a clinician. It does not triage, it does not interpret symptoms, and it does not give clinical advice. Emergencies are told to ring 999 or 111 in the UK, or 911 in the US, and clinical questions go to your staff.
What an AI medical receptionist does at a clinic
Strip away the clinical side and a practice front desk is a booking office with a long list of standing answers. That part is what an AI receptionist is built for.
Klugon's AI receptionist answers your phone line, your website chat and, if you connect them, WhatsApp, Instagram and Facebook Messenger. On every channel it works from the same material: your opening hours, your services and how long each one takes, your fee list, your policies, and the documents and web pages you give it. Concretely, it can:
- Book appointments into Klugon's built-in calendar, using your opening hours, the length of each appointment type, and which clinician offers it. It checks for conflicts per staff member, respects a minimum notice period and a booking window you set, and skips closed dates.
- Cancel appointments. On the phone it can cancel a booking and offer a new slot. In chat and messaging it can also move a booking in one step.
- Take new-patient enquiries: name, contact details, the reason for the visit in the patient's own words, how they heard about you. Each one is saved as a lead your team can pick up.
- Answer the standing questions from your own knowledge: prices, what to bring, parking, how long an assessment takes, whether you see children, whether you take a particular insurer, your cancellation policy.
- Hand the conversation to your team when it reaches something it shouldn't answer. That means it marks the conversation as needing a person, shows it in your inbox with a "Needs a person" banner, and, on plans that include Slack or Zapier, posts an alert there. It does not put the caller through to someone. More on that below, because it matters for a clinic.
- Answer at any hour. There is no business-hours switch on calls: if your number is routed to Klugon and your plan is live, it picks up at 03:00 on a Sunday the same way it does at 10:00 on a Tuesday.
It speaks English, German, French, Spanish, Italian, Portuguese, Dutch, Japanese and Arabic on calls, which helps practices with patients who are more comfortable in another language, as long as you remember it's still only doing admin in that language.
What it doesn't do is just as important for a clinic, so here it is plainly. It doesn't send appointment reminders, recall letters or texts of any kind. It doesn't send confirmation emails; the patient gets a booking reference in the conversation itself. It doesn't record calls. It doesn't make outbound calls. And it doesn't connect to clinical systems such as EMIS, SystmOne, Dentally, Cliniko or your veterinary practice management software. If those gaps rule it out for you, better to know now.
The calls a clinic actually gets
Before deciding what to hand to an AI, it's worth listing what comes in. Most practices that sit down and tally a week of calls find the same pattern: a large block of routine admin, a smaller block of calls that need a clinician's judgement, and a handful that are genuinely urgent.
Routine booking traffic
This is the bulk of it. "Can I book a check-up?" "Have you got anything Thursday after work?" "I need to cancel tomorrow, my son's ill." "Can I move my follow-up to next week?" For a physio clinic it's initial assessments and follow-ups; for an optician it's sight tests and contact lens checks; for a vet it's vaccinations, nail clips and post-op checks; for a dental practice it's examinations and hygiene appointments.
These calls follow rules you already know by heart. Initial assessments are longer than follow-ups. New patients can't be booked with the associate who's leaving in March. The hygienist doesn't work Fridays. Once those rules are written down, an AI can apply them consistently, at any hour.
New patient and new client enquiries
"Are you taking new patients?" "Do you do sports massage?" "How much is a private GP appointment?" "My dog's just been rehomed, can we register?" These callers are shopping around. If nobody picks up, many of them ring the next practice on the list, and you never find out they called. That makes these the calls most worth catching out of hours.
Admin questions with standing answers
Prices. Parking. Whether you have step-free access. Whether you accept a particular insurer, or direct billing. What to bring to a first appointment. How long a dental implant consultation takes. How to get a copy of an invoice. These have the same answer every time, and they eat reception time.
Calls that need a clinician
"I had a tooth out yesterday and it's still bleeding." "My back's worse since Tuesday's session, should I keep doing the exercises?" "My cat hasn't eaten for two days." "Can I take ibuprofen with what the doctor gave me?" "Are my results back?" Every one of these needs a person with clinical training, or at least a person who can read the patient's record. An AI must not answer them, and the next section is about making sure it doesn't.
Urgent and emergency calls
Chest pain. Difficulty breathing. Heavy bleeding. Facial swelling that's spreading. A dog that's eaten chocolate or rat poison. A sudden change in vision. These callers need to be told, immediately and clearly, where to go. That's 999 or 111 in the UK, 911 in the US, or your out-of-hours provider or emergency vet.
Busy hours and seasons
The pattern varies by practice type, but most clinics see it. There's a surge when the lines open, especially on Mondays and after a bank holiday, then lunchtime and the hour after work. After-hours calls cluster in the evening, when people get home and remember they meant to book.
Seasons shift the mix. Winter brings more illness and more same-day requests to general practice. January brings a wave of people starting new year health plans: physio for the running injury, the dental check-up they've put off. Late summer is back-to-school sight tests at opticians. For vets, fireworks season and the run-up to Christmas bring their own predictable calls, from anxious pets to things eaten that shouldn't have been.
The national picture shows why phone access matters to patients. In the 2025 GP Patient Survey published by Ipsos, 52.9% of patients said it was easy to contact their GP practice by phone, up from 49.7% in 2024. That's NHS general practice rather than private clinics, but it's a fair sign of how patients feel about getting through.
No triage, no clinical advice: the rule that comes first
If you take one thing from this page, take this. An AI receptionist at a clinic must never triage, never interpret symptoms and never give clinical advice. It doesn't matter how capable the underlying model is or how confident it sounds. A receptionist that tells a caller with chest pain to "rest and see how it goes" is a disaster. One that tells a parent their child's rash "doesn't sound serious" is too. Neither is acceptable from a person on your front desk, and neither is acceptable from software.
Klugon's assistant follows the instructions and knowledge you give it. That makes your instructions the safety mechanism, and it makes testing them your job before you go live. Here's how we recommend setting it up.
Emergencies go to 999, 111 or 911
Write an explicit instruction into the assistant's behaviour: if a caller describes anything that sounds like an emergency, or says it's urgent, the assistant tells them to hang up and call 999 now, or 911 in the US, and does nothing else. It doesn't ask follow-up questions or try to book them in. It doesn't try to judge whether the emergency is real.
For anything urgent but not an emergency, give it the right number for your patients. In England, that's NHS 111, online at 111.nhs.uk or by phone. For a vet, it's your out-of-hours emergency provider's number. For a dental practice, it's your emergency dental arrangement or 111. The NHS has a plain guide on when to call 999 that's worth reading when you write these instructions.
The key design choice: err towards the emergency route. You want an assistant that tells a caller with mild indigestion to call 111, not one that decides chest discomfort is probably nothing. Tell it, in so many words, that it can't assess symptoms and must not try.
Clinical questions go to staff
Anything about symptoms, medication, test results, treatment progress, whether to come in, or whether something is normal after a procedure is a clinical question. The assistant's job is to recognise it and hand it on, not to answer it.
In practice that means three things the assistant does with a clinical question:
- Tells the caller it can't advise on that, and that a member of the clinical team will need to help.
- Takes their name, a contact number and a short note of what they're asking about, in their own words.
- Hands the conversation to your team, so it appears in your inbox marked as needing a person.
It should also repeat the emergency route in the same breath: "If it gets worse or you're worried, call 111, or 999 if it's an emergency." That one sentence covers you if the caller underplayed their symptoms.
What "hands the conversation to your team" means
This needs spelling out, because many people assume an AI receptionist can put a caller through to the nurse. Klugon's can't. There is no live call transfer.
When the assistant hands over, it marks the conversation as escalated, logs it, and shows it in your inbox with a "Needs a person" banner. If you've connected Slack or Zapier (both need a Business plan or above), it also posts an alert there. Someone on your team then rings the patient back from your own phone.
Two consequences for a clinic. First, your handover isn't instant, so your instructions must never imply it is. The assistant should say "a member of the team will call you back", not "I'll put you through". Second, a handover is only as good as the person watching for it. Decide who checks the inbox, how often, and what happens out of hours. Also note that the assistant isn't paused after a handover: if the patient carries on talking, it carries on replying. Your instructions should tell it to stay on admin even then.
Questions the assistant must refuse
A short list helps when you write the instructions. The assistant should decline, and hand over, when a caller asks it to:
- say whether a symptom is serious, or whether they need to be seen;
- advise on medicines, doses or interactions;
- give, explain or chase test results;
- advise on aftercare beyond reading out a leaflet you've supplied word for word;
- decide whether a pet needs to come in today;
- issue or promise a repeat prescription;
- change a care plan, or say whether to carry on with exercises;
- confirm or discuss another person's appointment or records.
That last one is a data protection point as much as a clinical one. The assistant shouldn't tell a caller whether their partner has an appointment booked, however reasonable the request sounds.
Test it like you'd test a new receptionist
Before patients reach it, spend an hour trying to break it. Use the browser test during your trial. Describe chest pain casually. Ask whether a bruise after a blood test is normal. Say "it's probably nothing but" and see what happens. Ask about medication. Pretend to be a worried pet owner whose dog ate grapes. Every one of those should end with the emergency route or a handover, never a reassurance. If any of them doesn't, tighten the wording and test again. Repeat the exercise whenever you change the assistant's knowledge or instructions.
Where the assistant earns its keep
With the boundary clear, here's where an AI front desk is useful to a practice.
After-hours booking. The evening is when people sit down and sort their week out. A patient who wants a physio assessment and finds an assistant that offers Tuesday at 17:30 is booked. A patient who hits voicemail may well book elsewhere. For most clinics this is the clearest win, and you can start with only after-hours calls if you're cautious.
Overflow at peak times. Using your phone provider's call forwarding, you can send calls to Klugon only when your lines are busy or nobody answers within a set number of rings. Your team still takes the calls they can; the assistant catches the rest. Our guide to keeping your number with call forwarding walks through the options.
Cancellations. A patient who cancels at 21:00 has freed a slot that someone else can book through the same assistant at 21:15. Without that, the slot often stays empty until the morning.
Standing questions. Prices, parking and what to bring are perfect for the website chat as well as the phone. Every one the assistant answers is a call your team didn't take.
Patients who prefer messages. Plenty of people would rather send a WhatsApp at lunchtime than ring. Connecting WhatsApp needs a Meta business setup, but once it's done the same assistant answers there.
If you want to go deeper on booking specifically, our page on AI appointment booking covers how slots, buffers and staff work, and the after-hours answering service page covers the evening and weekend set-up in more detail.
Two sample conversations
Both are samples written for this page. The practices are made up, and so are the patients. They show the behaviour you should configure and test for, not recordings of real calls. Klugon's voices are AI-generated.
Sample 1: a physio booking after work
A sample call. Northgate Physio is a made-up two-room clinic. Phone call, Tuesday 19:42.
What to notice: the assistant said it was an AI in the greeting, because the clinic wrote that greeting (Klugon's default greeting doesn't, so you should write your own). It refused the clinical question without hesitating, gave the 111 route anyway, and offered a callback rather than a transfer. The booking used the right appointment length, the right price, and the clinician's real availability.
Sample 2: a dental call that has to go elsewhere
A sample call. Harbour Lane Dental is a made-up practice. Phone call, Saturday 22:10.
What to notice: the assistant didn't ask about the tooth, didn't offer an appointment, and didn't negotiate when the caller downplayed it. That's exactly the behaviour you write into the instructions and test for. The "note for the team" is a handover. Someone at the practice sees it in the inbox on Monday morning, or sooner in Slack if it's connected, and can follow up.
Patient data: GDPR, the ICO and CQC
This is the section to read with your data protection lead, or with whoever answers for data protection at your practice. Nothing here is legal advice. These are the questions you need to answer before you route patient calls through any AI tool.
Health data is special category data
Under UK GDPR, information about someone's health is special category data, which carries extra conditions on top of the ordinary rules. The ICO's guidance on what counts as special category data is clear that this includes information revealing a person's health status. That covers more than you might think. A caller saying "I'm booking because my back's gone again" has just told you something about their health. So has a booking for a "post-op review" or a "smear test".
That means you need a lawful basis and a separate condition for processing it. You need to tell patients in your privacy notice how their calls and messages are handled, and you need to be confident that every processor in the chain is suitable. The same ICO guidance sets out the conditions that apply.
Do a DPIA first
Putting an AI system in front of patients, where health data will be processed, is the kind of change where a data protection impact assessment is expected. The ICO explains when a DPIA is required and how to do one. Do it before go-live, not after. It will make you write down what data the assistant collects, where it goes, who can see it and how long you keep it. The answers will shape how you configure the assistant.
What you need to know about Klugon specifically
Here's what we can tell you, plainly, so your DPIA has facts in it rather than marketing:
- Who we are. Klugon is operated by Softaar Technologies UG, a German company, under the GDPR. Our privacy policy, linked at the foot of every page, sets out legal bases, data subject rights and retention.
- Where data goes. Conversations are processed by sub-processors that include Google Firebase, OpenAI and Telnyx, which our privacy policy lists as US-based with transfers covered by Standard Contractual Clauses. We don't claim that data stays in the EU or the UK, and you shouldn't assume it does.
- No certifications. Klugon is not certified for SOC 2, ISO 27001 or PCI, and it is not HIPAA compliant. We don't sign Business Associate Agreements.
- Separation. Each business's data is stored separately and knowledge search is filtered to your business only.
- Export and deletion. You can export leads, customers, conversations and other records as CSV and delete individual records. Full erasure is by email request. Deleting a customer doesn't automatically delete their conversations or bookings, so build that into your retention process.
- No call recordings. Calls are streamed, not recorded. Each call appears in your inbox as a conversation with its duration and message log, but the log may only show the assistant's side.
If your DPIA concludes that you can't send patient health information to these processors, take that seriously. A cautious set-up is still possible: use the assistant only for general questions and booking routine appointments, tell it not to ask about symptoms or reasons for visits, and point anything else to a callback from your team. Some practices will decide even that is a step too far. That's a legitimate answer.
Collect the minimum
Whatever you decide, configure the assistant to collect as little as it needs. For most bookings that's a name, a contact number and the appointment type. It doesn't need a date of birth, an NHS number, an insurance policy number or a description of symptoms to book a routine appointment. When a reason for the visit genuinely helps your team, ask for it in general terms ("new problem or follow-up?") rather than in clinical detail.
Tell patients they're talking to an AI
Klugon's default greeting doesn't say the caller is speaking to an AI, so write your own greeting that does. For a clinic this isn't optional good manners. Patients have a right to understand how their information is handled, and they're more likely to keep to admin when they know there's no clinician on the line.
CQC and record-keeping
If you're registered with the Care Quality Commission in England, your governance has to cover how you manage records and information. CQC's guidance on Regulation 17, good governance covers keeping accurate, complete records and keeping them secure. In practice that means two things for an AI receptionist. Any clinically relevant information that comes in through it (a patient calling to report a problem, a message for the clinician) needs to make it into your clinical record by your normal process, because Klugon's inbox is not a clinical record system. And your policies should say who checks the inbox, how often, and what happens to handed-over conversations. An inspector will be interested in how you know a message about a deteriorating patient didn't sit unread over a weekend.
Dental practices answer to the GDC for professional standards, and vets to the RCVS. The same principle applies to both: the AI is part of your admin process, and your professional obligations don't move to it.
For US practices: HIPAA
If you're a covered entity in the US, HIPAA requires a business associate agreement with any vendor that handles protected health information on your behalf. HHS publishes sample business associate agreement provisions that show what's involved. Klugon doesn't sign BAAs and isn't HIPAA compliant, so don't use it to handle PHI. The cautious set-up above, with general questions only and no patient details, is the most a US covered entity should consider, and you should check even that with your compliance adviser.
The practice types under this hub
Clinics have a lot in common, but the calls, the rules and the risks differ by type. Each of these pages goes into one kind of practice in detail.
Medical practices and private GP clinics
Private GP clinics, health screening and walk-in style services, where most of the phone traffic is booking and pricing, and the clinical boundary is the sharpest. Our page on the AI receptionist for medical practices covers consultation types, the questions patients ask about private fees, and how to word the emergency route.
Dental practices
Examinations, hygiene, emergencies and treatment consultations, plus NHS and private patients who ask very different questions. The dental AI receptionist page goes into appointment lengths, dental pain calls and what a dental practice should keep away from the assistant.
Opticians
Sight tests, contact lens checks, collections and repairs, with a predictable back-to-school rush. The page on optician appointment booking with AI covers those, and the vision symptoms that must always go to a person.
Physiotherapists and therapists
Physiotherapy, osteopathy, sports massage, counselling and other talking therapies, where initial assessments and follow-ups make up most bookings. The therapist booking assistant page covers session types, packages, and the extra care needed when a caller to a mental health service is in distress.
Veterinary practices
Vaccinations, neutering, post-op checks and a steady stream of worried owners, plus the out-of-hours emergency question that every vet faces. The veterinary AI receptionist page covers toxins, emergency vet referrals and what the assistant should say when an owner is frightened.
Booking: Klugon's calendar and your practice system
Here's the honest version, because it's the question practices ask first.
Klugon books into its own built-in calendar. That calendar knows your opening hours, your timezone, each service's length, which clinicians offer which services, buffers between appointments, minimum notice, how far ahead patients can book and closed dates. Cancelled appointments free their slot. Your team can edit, cancel or delete bookings in the dashboard, with a conflict check.
What it doesn't do is connect to clinical or practice management systems. If your appointment book lives in EMIS, SystmOne, Dentally, Cliniko, Jane, Pabau or a veterinary system, Klugon won't read it or write to it. On Business plans and above, the assistant can check Google, Outlook or Apple calendars and send Calendly links after setup, but that isn't reliable two-way sync.
That leaves three workable patterns:
- Klugon's calendar is the book. Fine for a small clinic that doesn't use a clinical system for appointments: a sole practitioner physio, a counsellor, a small private clinic.
- Klugon holds a block of slots. You set aside some slots, such as new-patient assessments or after-hours requests for the next morning, which the assistant books in Klugon's calendar. Reception copies them into the main system each morning. It's double entry, but on a known, small number of appointments.
- The assistant takes requests, not bookings. It collects the patient's details and preferred times as a lead and hands them to the team, who book in the main system and ring back. You lose instant booking, but nothing gets double-booked, and you still catch the evening caller who would otherwise have gone elsewhere.
Remember there's no automatic confirmation email or text. The patient gets a reference in the conversation. If your practice sends confirmations from its own system, pattern 2 or 3 keeps that working.
Setup checklist for a clinic
Work through this in order. Most of it is writing things down that your team already knows.
Before you start
- Complete a DPIA, or at least a documented risk assessment, covering the assistant (see the data section above).
- Decide which booking pattern you'll use: Klugon's calendar, a block of slots, or requests only.
- Decide which calls go to the assistant: after hours only, overflow, or all calls.
- Name the person who checks the inbox for handovers, and how often, including weekends.
Write the assistant's rules
- The emergency instruction: anything that sounds urgent gets 999, or 911 in the US, immediately, with no further questions.
- Your urgent-but-not-emergency route: 111, your out-of-hours provider, or your emergency vet.
- The list of clinical questions it must refuse and hand over.
- What it may collect, and what it must never ask for.
- A greeting that says the caller is speaking to the practice's AI assistant.
Add your practice information
- Services with durations and which clinicians offer them.
- Opening hours, closed dates and staff working patterns.
- Fee list, payment methods, and accepted insurers or plans.
- New-patient process, cancellation policy, what to bring, parking and access.
- Any patient leaflets you're happy for it to quote word for word, uploaded as PDF or Word files. Scanned PDFs aren't read, so use text-based files.
Connect and test
- Start the 7-day trial (a card is needed, and you're charged when the trial ends unless you cancel). During the trial, test calls in the browser.
- Run the break-it tests from the triage section: chest pain, medication questions, results, a frightened pet owner.
- Add the website chat with one script tag and test it the same way.
- After the trial, get a Klugon number and set your phone provider's call forwarding to it. You can't port your existing number, but forwarding keeps it as the one patients dial.
- If you use Slack, connect it for handover alerts (Business plan or above).
First month
- Read every handed-over conversation in the first week.
- Check the insights page for questions the assistant couldn't answer well, and add the missing information.
- Repeat the safety tests after every change to the instructions.
What it costs: worked examples
Klugon's prices are in US dollars. The voice product is $99 a month on Starter (1,000 minutes), $299 on Business (3,600 minutes) and $499 on Scale (6,000 minutes), billed monthly, or 20% less on yearly billing. Each phone number is $9 a month. Extra minutes are $15 per started 100. The website chat is a separate product from $49 a month. Full details are on the pricing page.
The examples below use stated assumptions. Change them to your own numbers.
Example 1: a small physio clinic, after hours and overflow
Assumptions:
- The assistant handles calls after 18:00, at weekends, and when both lines are busy.
- That's about 12 calls a day reaching the assistant, averaging 3 minutes each.
- The clinic is open 26 days a month, and the after-hours calls spread across the month.
Minutes: 12 × 3 × 26 = 936 minutes a month, inside Starter's 1,000.
Cost: $99 + $9 for the number = $108 a month, or $88 a month on yearly billing.
Starter doesn't include Slack or Google Calendar connections, so handovers show only in the Klugon inbox and insights page. For a clinic of this size that may be enough, provided someone checks it every morning.
The comparison: covering the same evening and lunchtime hours with a person. Say that's 15 hours a week. At the UK National Living Wage of £12.71 an hour from April 2026, that's about £191 a week before employer's National Insurance, pension and holiday pay, or roughly £825 a month. Few clinics would staff those hours at all, which is the point: the realistic alternative is voicemail.
Example 2: a dental practice with heavier traffic
Assumptions:
- The assistant takes overflow during the day and everything after hours.
- About 40 calls a day reach it, averaging 3.5 minutes.
- 22 working days, plus weekend calls bringing the total to roughly 25 days' worth.
Minutes: 40 × 3.5 × 25 = 3,500 minutes, inside Business's 3,600.
Cost: $299 + $9 = $308 a month, or $248 a month on yearly billing. Business includes Slack and Zapier alerts and calendar connections, which a practice of this size will want for handovers.
If a busy month runs to 4,000 minutes, the extra 400 minutes cost 4 × $15 = $60.
Example 3: what a missed call is worth
Assumptions, which you should replace with your own:
- 8 calls a week currently go unanswered after hours or at peak times.
- 1 in 4 of those callers would have booked.
- The average first appointment is worth £60.
That's 2 bookings a week, or about £520 a month of appointments that are currently going elsewhere, and that's before counting any follow-up appointments. If that's anywhere near your real numbers, Example 1 pays for itself with two bookings a month. If your phones are already answered well, it may not, and that's worth knowing. Our comparison of an AI receptionist vs a human receptionist goes through the trade-offs in more detail.
How to tell if it's working
Give it a month, then look at four things.
Bookings made by the assistant. The analytics page shows bookings, calls and minutes, split by channel, by day and by hour. Look at the hours: if most assistant bookings land in the evening, that's revenue your phones weren't catching before.
Handovers. The "handled by the assistant" figures show how many conversations went to a person. A high handover rate in the first weeks usually means missing information. Read the handed-over conversations and add the answers to its knowledge. A handover rate of zero is a warning sign too. It may mean the assistant is answering things it shouldn't, so read a sample of conversations to check.
Questions without a good answer. The insights page lists questions the assistant struggled with and the most asked questions over the last 30 days. Both are a to-do list for your knowledge base.
Your team's view. Ask reception whether the Monday morning queue feels different, and whether the messages they pick up from the inbox are clear enough to act on. If they're not, change what the assistant collects.
And keep testing the safety rules. A monthly five-minute run through the emergency and clinical scenarios is cheap insurance.
Frequently asked questions
What is an AI medical receptionist?
It's software that answers a practice's phone and messages, books and cancels appointments, answers admin questions from your own information, and hands anything clinical to your staff. Klugon's assistant does the administrative side only. It never triages and never gives clinical advice.
Can an AI receptionist triage patients?
No, and it shouldn't be set up to try. Klugon's assistant is configured by you to send anything urgent to 999, 111 or 911 straight away, and to hand clinical questions to your team without answering them. Test those instructions before you go live, and after every change.
Can it put a caller through to a nurse or the practice manager?
No. Klugon has no live call transfer. The assistant flags the conversation for your team, it appears in your inbox marked as needing a person, and on Business plans it can alert Slack or Zapier. Your team then calls the patient back.
Is Klugon HIPAA compliant, or will it sign a BAA?
No to both. Klugon is not HIPAA compliant and doesn't sign Business Associate Agreements, so US covered entities shouldn't use it to handle protected health information. It's also not certified for SOC 2 or ISO 27001.
Where is patient data processed?
Klugon is operated by a German company under the GDPR. Its sub-processors include Google Firebase, OpenAI and Telnyx, which the privacy policy lists as US-based with transfers covered by Standard Contractual Clauses. We don't claim EU or UK data residency. Read the privacy policy and do a DPIA before routing patient calls to it.
Does it work with EMIS, SystmOne, Dentally or Cliniko?
No. Klugon books into its own calendar and doesn't connect to clinical or practice management systems. Many practices have the assistant take booking requests, or hold a block of slots, and copy those into their main system.
Will it send appointment reminders or confirmations?
No. Klugon doesn't send reminders, recall messages, texts or confirmation emails. The patient gets a booking reference in the conversation itself. If you need confirmations, keep sending them from your own system.
Will patients know they're talking to an AI?
Only if your greeting says so, because the default greeting doesn't. For a clinic, write a greeting that tells callers they're speaking to the practice's AI assistant.
Can I keep my practice phone number?
Yes, through call forwarding. Klugon can't port numbers, but after your trial you can get a Klugon number and forward your existing line to it, either all calls or just busy, unanswered and after-hours calls.
How long is the free trial?
Seven days, with a card required. You're charged when the trial ends unless you cancel. During the trial you test the assistant in the browser; a phone number can only be added once the trial is over.
Before you decide
An AI medical receptionist is worth having when your problem is admin volume: callers who can't get through to book, questions that tie up reception, and the evening enquiries that go to whichever practice answers first. It isn't worth having if you need it to do anything clinical, or if your data protection assessment says patient details can't go to the processors involved. Do the DPIA, write the safety rules, test them hard, and start with after-hours calls if you want to see how it behaves before it touches your busiest hours.