“I’m Still on Doxycycline — Can I Do IPL Thursday?” Is Not a Question Your Front Desk Should Answer From Memory

How many times last week did a 23-year-old at your front desk answer a medical question about a client's skin?

Not a booking question. A medical one. "I'm still on doxycycline for my acne, can I still do my IPL on Thursday?" "I had my teeth cleaned Tuesday, is that fine before lip filler?" "My mum's on a blood thinner, can she get Botox?" "I got a cold sore last month, does that matter?" "My lip feels really tight and it looks kind of white and blotchy, is that normal?"

Every one of those got answered. Some of them got answered correctly because your coordinator has heard the injector say it forty times. Some of them got answered in a friendly, confident, entirely wrong way, and you will never know which ones unless something goes badly.

The questions that arrive before the clinical staff do

The pattern in a medical aesthetics practice or a spa with clinical services is always the same. The people with licences — the RN or NP injector, the laser-certified aesthetician, the medical director who signed your standing orders — are with patients or not in the building. The people who receive the questions are at the desk, on the text line, and in the Instagram inbox, and they are answering at 9:12am when nobody clinical is free.

The high-consequence ones are well known to anyone who has run this business for a year. Photosensitising medications before light-based treatment. Isotretinoin history before resurfacing. Retinoid use before a peel or a wax. Fish oil, ibuprofen and alcohol before injectables. Recent dental work before filler. Cold-sore history before lip work or resurfacing. Fitzpatrick type before choosing a laser. Recent sun exposure or a spray tan before anything with light. Pregnancy and breastfeeding before toxin. And the one that is not a scheduling question at all: blanching, mottling and pain out of proportion after filler, which is a possible vascular occlusion and is measured in hours, not days.

What "answered from memory" costs

Take the mildest possible failure. Your coordinator tells a client that her fish oil is fine, the client bruises across both cheeks before a wedding, and you comp the syringe and a follow-up. That is roughly $600 of product and chair time, plus a client who tells six people.

Take the middle one. Someone books an IPL on a client who has been on a photosensitising antibiotic, she blisters, and now you have a burn, a refund, a course of care you are paying for, and a possible complaint to the state board that will consume your medical director's month.

Take the bad one. A client texts at 8:50pm describing exactly what an occlusion sounds like and gets a reply saying it is normal swelling and to ice it. That is the entire reason your practice carries insurance.

What actually changed in 2026

Grounding means the assistant is only allowed to answer from documents you approved, and it must show which document the answer came from — so a wrong answer is catchable rather than merely plausible. That is the whole development, and it is the thing that was missing when practices tried this in 2024 and got confident nonsense back.

The practical difference is visible in the reply itself. The 2024 version said "yes, doxycycline is fine, just use sunscreen," pulled from nowhere, phrased beautifully. The 2026 version says "our pre-treatment sheet for IPL lists tetracycline antibiotics as a hold — source: Pre-Care, Light-Based Treatments, v4, approved 3 March 2026 — so I am passing this to the RN before we confirm Thursday." You can click the source. Your medical director can read the exact line. If the sheet is wrong, you fix the sheet and every future answer changes.

flowchart TD
  A["Client texts at 9:12am about doxycycline and IPL"] --> B["Assistant searches the approved binder"]
  B --> C{"Is the answer in an approved document?"}
  C -->|No| E["Route to the RN injector queue"]
  C -->|Yes| D{"Does it involve medication or a complication?"}
  D -->|Yes| E
  D -->|No| F["Reply, quoting and naming the source sheet"]
  E --> G["RN answers and the answer is added to the binder"]
  G --> B
  F --> H["Reply logged to the client chart"]

Thursday, 9:12am, run the new way

The text lands. The assistant searches only your approved material: your pre-care and post-care sheets, the standing orders your medical director signed, the manufacturer's instructions for use for each product you carry, your device settings by Fitzpatrick type, and your consent forms. Nothing else. It does not search the internet and it does not search its own general knowledge.

It finds the hold on tetracyclines in your own light-based pre-care sheet. Because the question involves a medication, the rule you set says it does not answer — it drafts the answer, names the source, and puts it in the RN's queue with the client's chart attached. The RN reads two lines at 9:40 between patients, taps approve, and the client gets a reply that is both fast and correct, with a rebooking offer for four weeks out already attached.

The occlusion text at 8:50pm gets a different treatment entirely. Words like blanching, white, mottled, severe pain and vision change sit on a list that never gets an automated answer. That message pages the on-call injector and tells the client, in plain language, to call the number on their post-care card now. That routing rule is the single most valuable thing in the setup and it takes ten minutes to write.

The maths on a three-percent wrong-answer rate

Illustrative figures for a practice with two injectors and one laser room.

The $3,800 is the number you can prove. The unmodelled line is the reason owners actually do this. To measure your own starting rate before you buy anything, have your injector spot-check twenty answered questions from the text log this week and mark how many she would have answered differently. That number will surprise you in one direction or the other, and either way you now know.

Where a human stays, permanently

The assistant does not diagnose, does not clear anyone for treatment, does not tell a client to stop or start a prescribed medication, and does not answer anything about an active complication. Those are scope-of-practice questions in every state, and no amount of grounding changes who is licensed to answer them. Your medical director's standing orders decide what a non-licensed person may say; the assistant simply obeys the same boundary in writing.

Second, the binder is only as good as its last review. A pre-care sheet from 2023 that still reflects an old resurfacing protocol will now be quoted confidently to hundreds of clients instead of a handful. Put a review date on every document, and make the assistant refuse to quote anything past its date.

Third, protected health information. Client names, medications and photos going into any assistant means a signed business associate agreement and a log you can pull. Ask for both before the first message flows through it.

The binder is the project, not the software

Most practices discover something uncomfortable in week one: they do not have an approved binder. They have a Google Doc from an old medical director, three versions of a post-care sheet in the printer drawer, and everything else in one injector's head. Building the real thing — pre-care and post-care per service, hold lists, escalation words, device settings by skin type, all dated and signed — takes a couple of afternoons with your medical director.

Do that first, for one service line, and the grounding part is almost easy. Start with light-based treatments, because that is where the pre-care rules are strictest and the failures are most visible on someone's face.

Frequently asked questions

Can it just answer the medication questions if my medical director approves the list?

It can quote your approved sheet and route to a licensed person, and that is where most practices should stop. Whether a written answer to a medication question constitutes advice from an unlicensed source is a question for your medical director and your malpractice carrier, not for a software vendor. The safe default — quote the sheet, name it, hand the decision to the RN — costs you a 30-minute delay and removes the argument entirely.

How is this different from just putting the FAQ on our website?

A website FAQ answers the question the client thought to look for. This answers the question they typed at 9:12pm in their own words, in the same text thread they book in, and it shows the staff which sheet it used so you can audit it later. The audit trail is the part a website page cannot give you.

What if the answer is in a document but the document is wrong?

Then it is confidently wrong — and that is still better than before, because you can see exactly which sheet produced it and fix one file. That is the trade: you take on responsibility for maintaining your own documents, and in return every answer becomes traceable to one.

Does this work for a hair salon with no medical services?

Yes, and the same logic applies to the patch test. A client asking whether she needs a 48-hour allergy test before colour, or whether a keratin service is safe on chemically relaxed hair, deserves the answer from your own written service policy rather than from whatever the desk half-remembers.

Most of these questions arrive by phone and text outside working hours, which is exactly when nobody clinical is there to take them. CallSphere builds AI voice and chat agents that answer the practice line and web chat around the clock, book and reschedule appointments, and capture the enquiry rather than lose it — and they can be set to answer only from your approved sheets and to hand anything clinical straight to your on-call injector.