A clinic or salon front desk in Singapore is doing three jobs at once. Someone is being checked in, the phone is ringing, and two WhatsApp threads sit half-answered on a phone by the till. The person messaging to ask "do you do wisdom-tooth extraction, and how much" is fourth in a queue they cannot see — and by the time anyone replies, they have already messaged the clinic down the road. The work did not walk in the door; it quietly went next door.
This piece is about running an AI receptionist on WhatsApp: something that answers the routine questions, books and reschedules appointments, and sends reminders, while every clinical question still goes to a person. It is written for GP and dental clinics, TCM and aesthetic practices, and salons and spas, because the front-desk squeeze looks the same across all of them. No hard sell — where an AI receptionist for a Singapore clinic or salon genuinely helps, and where a human must take over, we will say plainly.
Why a missed message is a missed booking
First contact in Singapore is almost never a form or a phone call. It is a WhatsApp message, and it often lands when you are least able to answer it: lunchtime, after closing, or in the middle of a treatment. The customer is not being unreasonable — that is simply when they are free. But an unread message at 9pm is a decision being made without you in the room.
- Enquiries arriving after hours sit unread overnight, and for salons and aesthetic clinics that is the busiest messaging window.
- During peak clinic hours the front desk is with the patient in front of them, so the phone and the chat both lose.
- Every booking is preceded by three or four questions — price, availability, which doctor — and each unanswered question is a chance to lose the person.
- A caller who reaches voicemail rarely leaves one. They just try the next name on the list.
None of this shows up in your accounts as a loss, which is why it runs for years. The gap between "I'm interested" and "it's in the diary" is where the bookings leak out.
What an AI receptionist actually does
Strip away the marketing and the job is narrow and useful. It handles the questions that repeat all day, and it turns a willing customer into a confirmed slot without a human touching the keyboard. A dental patient asks about scaling and polishing on a Sunday night; they get the price, the duration, and three open times, and pick one. A spa client asks whether you have anything after work on Thursday; the answer is real, not "we'll check and revert".
- Answers FAQs from your own information — opening hours, prices, services, parking, which conditions you treat.
- Offers real open slots and writes the confirmed appointment straight into your calendar.
- Reschedules and cancels in the same thread, without the customer chasing anyone.
- Replies in the customer's language — English, Mandarin, Malay — and in seconds, at 2am as readily as 2pm.
- Hands off cleanly to a person the moment a question stops being routine.
That last point matters more than the flashy parts. Good clinic appointment automation is not about sounding human; it is about knowing the edge of its competence and stepping back before it does harm.
Booking across several practitioners
This is where generic booking links fall down for clinics and salons. Availability is not per shop — it is per doctor, per stylist, per therapist. A patient wants Dr Tan specifically; a client wants the senior colourist, not whoever is free. A TCM clinic may have three physicians sharing two treatment rooms, so the real constraint is the room, not just the person. A booking that ignores this double-books a space that physically holds one patient at a time.
The workable pattern is a single shared calendar where each practitioner has their own availability, the right person is assigned automatically, and the system holds a buffer between appointments — the few minutes to reset a chair, sanitise a room, or write up notes. For salon booking on WhatsApp this means a colour that runs three hours blocks the right stylist for three hours, not a generic one-hour slot that leaves the next client waiting.
A booking tool that treats a clinic as one diary will double-book the one room you actually have. The unit that matters is the practitioner and the space, not the shop.
Cutting no-shows without the nagging
No-shows are usually what pushes an owner to start looking. An empty aesthetic slot or a missed dental chair is revenue you cannot get back, and for practitioners on commission it stings twice. There is no clever trick here: a reminder the evening before is late enough that the customer has not forgotten, and early enough that they can still tell you if plans have changed.
Two details decide whether it works. The reminder has to arrive on WhatsApp, where it gets read, rather than an email that does not. And it should be repliable — if a client can answer "can I move to Friday morning" and have it handled in the thread, an empty chair quietly becomes a rescheduled booking instead of a loss.
Where the AI stops and a person takes over
This is the line you must not blur, especially for anything medical or clinical. An AI receptionist should book the consult; it should never play doctor. The safe design routes anything clinical to a human every time, and errs toward escalating rather than guessing.
- Symptoms, diagnoses, or "is this normal after my procedure" — straight to a clinician, not answered by the bot.
- Medication, dosage, and drug-interaction questions — always a person.
- Post-op concerns, allergic reactions, or anything that sounds urgent — flagged and handed off immediately.
- Complaints and pricing disputes — better warm than automated.
The other half of doing this responsibly is data. Clinics and salons handle NRICs, contact details, sometimes health information, and Singapore's PDPA is not optional. The sensible baseline is to redact identifiers like NRIC and card numbers before anything is stored, keep the minimum, and make sure your chat data is never used to train someone else's model.
Where microcrew fits
microcrew is built for exactly this shape of work. It sits on your WhatsApp and Telegram, replies in about 30 seconds around the clock, answers from your own prices, services and policies, and books straight into your Google Calendar inside the conversation — no separate page for the customer to visit. It supports several practitioners on one shared calendar with automatic assignment, holds buffers between appointments, and sends the client a night-before reminder they can reply to. Clinical questions route to you; NRIC and card numbers are redacted before storage, and processing runs on Google Vertex AI in Singapore under a no-training agreement. See how it maps to clinics and salons, or read more on the blog.
It will not replace your front desk, and it is not meant to — it clears the routine questions and the after-hours enquiries so your team can be present with the person in the chair. If that sounds like your Saturdays, the first month is 50% off and you can start a chat on WhatsApp at +65 8466 3236. Access is invite-only for now, so message early if you would like a slot.