The slow part of a med spa day is not the treatment. It is everything around it, the post-consult summary, the aftercare sheet, and the follow-up check-in, turning the injector's decided plan into clear, friendly patient-facing text. That writing is the kind of work AI is good at. The trap, and it is a serious one here, is that the same tool will happily recommend a treatment the injector never chose, state a dose or product that is not in the chart, or promise a result, and that text goes out to a patient under your clinic's name as if a licensed clinician said it. That is not a typo, it is a clinical and compliance exposure. So the rule for this task is the strictest on this site: AI translates the plan the injector already made, and nothing else.
One ground rule before any of it: these are health records. Keep patient identifiers out of a public chatbot, test with de-identified examples, and for real records use a private tool that keeps the data inside your own systems. More on that at the end.
The honest picture
AI is genuinely useful for turning a plan that is already decided into clean patient communication:
- What AI does well today: take the treatment the injector already chose, the aftercare they already prescribed, and the next steps already documented, and write them up as a clear post-consult summary, a readable aftercare sheet at a fixed reading level, and a warm follow-up message in your clinic's voice. It compresses the writing time around each visit, not the clinical judgment in the room.
- What AI does not do: recommend or suggest a treatment, product, or injectable; state a dose, units, or dilution; make any medical or results claim; promise an outcome; diagnose a skin or health condition; or decide who is a candidate. The licensed injector, and the supervising physician where required, owns every clinical decision and the chart. AI only translates the plan and aftercare they already set into patient-facing text. It must never originate a recommendation or a number.
The right way to think about it: AI is a fast writer working from the injector's decided plan, not a second clinician. The treatment, the dose, the candidacy call, and the chart are theirs. The patient-facing wording is what you hand off.
The line: do not let it recommend, dose, or promise
The specific failure to watch for is helpfulness, and here it is the most dangerous failure on the site. It shows up four ways:
- Recommending. Ask AI to "suggest what this patient should consider next" and it will cheerfully propose a treatment or an upsell the injector never decided. A recommendation that did not come from the licensed clinician, sent to a patient, reads as clinical advice from your clinic. Never ask it to suggest. Ask it to write up what the injector already chose.
- Dosing. Never let it state units, product names, or dilution. If a number or product appears in patient text, it must be copied verbatim from the chart, never generated. A made-up dose is a patient-safety problem.
- Claims and promises. It must make no medical claim and promise no result. No "will eliminate," no "guaranteed," no "you will look years younger." Outcomes are the clinician's to discuss in person, and even then they are not guarantees.
- Diagnosing from a photo. Pointing a vision model at a patient photo and asking "what is this" or "what would help" is the exact line not to cross. A model guessing a condition or a treatment from a picture is both a diagnosis and a recommendation it cannot make.
The fix is one discipline applied without exception: AI never originates clinical content. It uses only the injector's documented plan and prescribed aftercare, copies any product or dose verbatim from the chart, and flags anything it cannot tie to the chart rather than filling it in. A polished follow-up that quietly added a recommendation is worse than a blank one.
The setup that keeps the notes yours
Three habits make AI safe for this task, and the prompts below build them in:
- Give it only the decided plan. Hand it the treatment the injector chose, the aftercare they prescribed, and the documented next steps, with any product or dose written exactly as charted. Tell it to use only these and to copy numbers and product names verbatim.
- Make it flag, not fill. If a piece of patient text would need a recommendation, a dose, or a claim that is not in the plan, it lists that under "needs the injector's input" and the injector decides. It never fills the gap and never softens the rule to be helpful.
- Keep the data yours. De-identified examples for testing in a public tool; real patient records only in a tool that stays inside your own systems.
How to test it on your own work
Do not trust a polished demo, including this one. Pull two or three of your own recent consults, de-identified, with the injector's decided plan and prescribed aftercare, a timer, and the prompts below. Rate each output 1 to 5 on usefulness and accuracy, and compare the time against how you write follow-ups today. Keep what wins.
Paste-ready prompts
Copy these as written. Bracketed text is what you swap per patient.
Test 1: Post-consult summary from the decided plan (text model)
I am giving you the plan a licensed injector already decided for one patient,
plus the aftercare they prescribed. Write a clear post-consult summary for the
patient in our clinic's friendly voice. Rules:
- Write only the treatment and steps the injector already chose. Do not suggest,
recommend, or mention any additional treatment, product, or upsell that is not
in the plan I gave you, even if it seems like a natural next step.
- Copy any product name, units, or dose exactly as written in the plan. Do not
add, change, or "round" any number. If something looks incomplete, do not fill
it: list it under "Needs the injector's input."
- Make no medical claim and promise no result. Do not say a treatment "will"
fix, eliminate, or guarantee anything.
Decided plan and prescribed aftercare: [paste]
Watch for: did it add a treatment, a number, or a "you'll love the results" line that was not in the plan? Anything it originated is content a licensed clinician did not approve. Delete it.
Test 2: Aftercare instructions from the prescribed aftercare only (text model)
Here is the aftercare the injector prescribed for this patient: [paste]. Rewrite
it as a clear aftercare sheet a patient with no medical background can follow,
at about an 8th-grade reading level. Rules:
- Use only the instructions I gave you. Add no new step, product, timeline,
warning, or "while you're at it" suggestion that is not in the prescribed
aftercare.
- Copy any product name or amount exactly as written. Do not invent dosing or
frequency.
- If a step is unclear, do not guess: list it under "Needs the injector's input"
so they can clarify. Tell the patient to call our office with questions and to
contact the clinic about anything that concerns them.
Watch for: did it keep every instruction intact while making it readable, or did it add a tip or a product the injector never prescribed?
Test 3: Follow-up check-in from documented next steps (text model)
Write a short, warm follow-up check-in message to a patient after their visit.
Use only the documented next steps I provide: [paste]. Rules:
- Ask how they are doing and remind them only of the next step the injector
already documented (for example a scheduled follow-up or a charted check-in
window). Do not recommend any new or additional treatment, product, or visit
that is not in the documented next steps.
- Make no claim about results and promise no outcome. Do not imply the treatment
worked or will work.
- Invite them to reply or call our office with any questions or concerns.
Documented next steps: [paste]
Watch for: does it stay a friendly check-in, or does it slip into selling a next treatment or promising how they'll look?
Test 4: Consistency and compliance audit (text model)
Compare this draft patient message against the injector's decided plan and
prescribed aftercare that I provide. For each problem, quote the exact line.
1. Any treatment, product, or upsell in the draft that is not in the decided
plan.
2. Any dose, units, dilution, or product name in the draft that does not match
the plan verbatim, or any number that was added.
3. Any medical claim, diagnosis, or results promise ("will," "eliminate,"
"guaranteed," and similar).
4. Any next step or recommendation not in the documented plan.
Do not fix anything. Only flag. Treat any originated recommendation, number,
claim, or promise as a defect.
Decided plan and prescribed aftercare: [paste]
Draft message: [paste]
Watch for: does it catch the line that quietly recommended a treatment or promised a result? Run it on a message you already sent.
What success looks like, and where it could go
If your own testing shows real time savings, the next step is a small pilot: run a week of consults through the prompts and measure the minutes saved per patient and how clean the drafts come back. If that holds up, the natural next step is a simple agent, running on your clinic's own cloud, that you use in plain language. The most useful version pulls only the injector's decided plan and prescribed aftercare, writes the summary, aftercare sheet, and follow-up at a fixed reading level, copies every product and dose verbatim, refuses to originate any recommendation or claim, and cites the source line in the chart behind every statement, so the injector reviews and approves in minutes instead of writing from scratch. Because it runs in your own systems, patient data never leaves for a public chatbot.
The principle holds the whole way through: AI translates the plan the injector already made into clear patient text. It does not recommend a treatment, state a dose, diagnose, or promise a result. The licensed injector, and the supervising physician where required, owns every clinical decision and the chart. Used this way it speeds the writing without ever putting clinical content in front of a patient that a licensed clinician did not decide.
This is general information about workflow tools, not clinical or compliance advice.
Want a straight answer for your med spa?
I build practical AI and custom software for businesses, on Google Cloud. If you want a second set of eyes on how AI could fit your consult and follow-up workflow, or on a tool you are considering buying, tell me what you are working with. No pitch, just a straight answer.