Yes for routine bookings and calls nobody is free to answer, if the system passes the clinic's own booking and handover tests; no as the voice that answers treatment questions or complaints. An AI receptionist, meaning software that answers the phone, speaks with the caller and can book into the diary, is a way to cover a gap in the phone service, not a replacement for the desk. Before buying one, check that it can complete the clinic's own booking tasks, say plainly what it cannot do, and reach a person when it needs to. Whether it earns a place depends on the calls being missed, the work it removes and the work it creates.
Where AI fits across the rest of the week, from reminders to stock, is covered in where AI earns its place in a clinic's week. The decision here is the phone alone, because a call is where a booking and a client's question can arrive together.
The missed call or the wrong answer
Use the phone log to find the unanswered calls, then check what happened to them before treating them as lost bookings. Review a month of calls alongside messages, completed bookings and the desk's own account of its busiest periods:
- Outside opening hours. Check which callers left a message, got in touch another way or booked later. Mark an unrecognised number as unknown until its purpose is clear, and look for routine booking requests that went unanswered.
- During opening hours. Compare missed calls with the rota and the desk's workload. Ask whether the team was occupied, whether calls reached the right extension and whether anyone owned the callback, and respond to the cause found rather than assuming another person is needed.
- Across both. Mark repeat attempts and short abandoned calls separately, and match them to later contacts, so one enquiry is not counted three times.
A limited trial makes sense where this review finds routine requests that neither the desk nor the existing callback process completes. Which calls need a person is decided by what the call is about, not by whether the number is familiar.
Speed or judgement
Judge speed by the time taken to complete the caller's request correctly, including any handover, not by how quickly the phone is answered.
Use an invented call as a rehearsal: a caller asks to move next week's appointment, then says she was unhappy with how the last treatment went. Ask the team and the supplier to show what happens next. Does the concern reach the right person with the caller's request intact, or does the process stop once the appointment has moved?
Record the booking outcome and the handover outcome separately, and do not let a correct diary entry pass the test if the concern was left unanswered. Service that completes the transaction and misses the person is the subject of the quiet service gap that never shows up as a complaint.
One voice for every call, or a person for some
A person for some, chosen in advance and written down. Reserve treatment questions and complaints for people, along with requests for a named practitioner or simply to speak to someone. For reactions, pain or concerns about a result, use the contact route agreed by the clinical team; the software does not decide what can wait. For routine administrative callbacks, name who receives the message, when it is checked and who covers an absence. Test what happens when the first transfer goes unanswered or fails, and promise only a callback time the team has agreed to cover.
Test the greeting as part of the call: can a new caller tell which clinic has answered, that the receptionist is automated and how to reach the team? Settle those before the caller is asked for any booking details. For a new client, that first exchange sits alongside the other ways a clinic earns trust before any treatment result is visible.
What it costs, or what a lost booking costs
Set the extra contribution from attended appointments against the full cost of the tool. Count staff time taken off answering calls separately from time added for setup, checking, corrections and callbacks, using appointment and cost records alongside the phone log, and state the hourly value used for staff time.
As an illustration only: ten distinct unanswered booking enquiries a week over four weeks gives forty. If one in four becomes an attended first appointment, that is ten appointments. Those are assumptions for the sum, not observed results. Before crediting all ten to the software, check how many would have booked through an existing route anyway. Multiply the extra attended appointments by what remains after the cost of delivering them, then compare that with the subscription, any usage or setup charges, and the staff time.
Keep the service record beside the sum: complaints, correction work and missed callbacks. If a regular stops booking after a difficult call, find out why before putting the loss down to the software.
Switching it on, or testing it first
Test it first. Complete the team rehearsals before it takes a live call, then limit the trial to the booking tasks and hours for which handovers and messages have named cover; out of hours is an option only if those arrangements work out of hours. Set a first review after a month, but do not widen the service because the date has arrived. If important call types have not come up, rehearse them and keep the trial narrow until there is evidence to judge them.
Check every booking during the trial, and review failed transfers, missed callbacks and calls the team had to correct. If a booking error or failed handover is left unresolved, pause that task, fix the cause and repeat its test before switching it back on. How call recordings and client details are stored is a question for a data-protection adviser before anything goes live. Write down, before any supplier demonstration, which phone tasks the trial covers and what return would justify keeping each, and judge the tool against that list rather than against what was promised in the demonstration.
Questions clinic owners ask about AI on the phone
What should a clinic test before an AI receptionist takes live calls?
Have the team call it with the twenty questions the desk hears most, including the awkward ones. Check that it books the right room, practitioner and appointment length, that it refuses slots the diary does not allow, that every handover phrase reaches a person, and that out-of-hours messages arrive with the caller's number and request intact. Then check how it ends a call it cannot help with.
What if the practitioner a caller asks for is unavailable?
Offer the caller a choice between another suitable team member and a callback from the named practitioner, using the arrangements the clinic has agreed. Do not promise that practitioner's availability without checking it. If the caller raises a concern about a treatment, use the clinical team's agreed contact route rather than putting the request into the ordinary booking callback queue.
What should an AI receptionist's opening greeting say?
Give the clinic's name, then something like: "You are speaking to the clinic's automated receptionist. It can help with routine bookings, and you can ask to speak to the team at any point." Adapt the last part to the cover actually available: promise a transfer only when someone can take it, and otherwise explain the real callback arrangement. Then test that the choice offered works.
What should a clinic measure in its AI receptionist's first month?
Record distinct calls by purpose, correct bookings, attended appointments, failed transfers, callbacks completed on time and staff minutes spent checking or correcting its work, and include caller feedback. Compare equivalent opening hours, noting changes in demand or staffing, before putting a difference down to the tool. Use the results to keep, change or stop each tested task, and rehearse any new call type before widening coverage.
The first task to hand over
Choose the first phone task by the gap it covers, the work it saves or creates, and the cost of reliable cover. Then write it down on one page (we can help write it): the booking actions the software may take, the hours it runs, the named person every other call goes to, and the date of the first review. Rehearse it and run it as a limited trial. What the clinic keeps, widens or switches off afterwards follows from what the trial recorded: correct bookings, handovers, caller feedback and the return on the cost.
Gaia Gabiati, Consulting Lead at The Boutique Consultancy. A decade across health clubs, private members' clubs, hospitality, wellness and multi-site aesthetics clinics, from Milan through Harvey Nichols, Virgin Active, Third Space and Soho House, to running the operational side of multi-site luxury aesthetics clinics.