Superintelligence Prep for Med Spas: Consent and Rebook Records
An AI agent for a med spa can handle rebook reminders and consent record checks, but it must stop before clinical judgment. Here is what to structure now.

An AI agent for a med spa should handle the administrative work around treatments: sending consent forms for signature, checking that each booked treatment has a current and matching consent on file, and following up to rebook at the interval a provider has approved. It should not judge whether a client is a suitable candidate, interpret a reaction, or change a protocol. The practical question for an owner is which consent, treatment-history and rebook records must be structured now so that an agent can act on them without guessing.
Key takeaways
- Treat each consent as a dated, versioned record tied to one treatment, not as a file sitting in a folder.
- Keep treatment history in fields an agent can read, and reserve clinical notes for licensed staff.
- Write rebook rules that the medical director approves, with explicit exclusions written in plain terms.
- Give the agent hard stops for safety questions, reactions, dosing, and any request to bypass consent.
- Judge the setup by one observable criterion: every booked treatment has a valid consent on file before the appointment, and every held message has a logged reason.
What can an agent safely touch in a med spa?
Med spa operations mix administrative work with clinical judgment. The administrative side is large: confirming appointments, sending intake and consent forms, collecting signatures, reminding clients about aftercare timing, and asking whether someone wants to rebook. An agent can run these steps reliably when the rules are written down and the records behind them are clean.
The clinical side belongs to licensed people. Who may treat, what is appropriate for a given client, and what a change in the skin or a side effect means are decisions for the provider. Your medical director and legal advisor set these lines for your state, province, or country, because requirements differ. Write the lines down before you configure anything, and treat the agent's permissions as a copy of those lines rather than a substitute for them. The NIST AI Risk Management Framework is a useful reference for naming these risks and assigning an owner to each one before deployment.
How should consent records be structured before an agent uses them?
An agent cannot work reliably from a consent that exists only as a scanned file with a vague filename. It needs a record it can check: which form version was signed, for which treatment, by whom, and whether anything has changed since. Structure these elements for every consent:
- The client and the treatment category the consent covers, such as neurotoxin injections or chemical peels
- The form version identifier, so a signature on an older template stays tied to that template
- The date and time signed, and whether it was completed in the clinic or remotely
- The staff member who reviewed the consent with the client
- Any revocation, update, or question, logged with its own date
- The renewal rule the practice has set, such as when a consent must be signed again
The agent's job is to read these fields and answer a narrow question: is there a current, matching consent for this appointment? It does not interpret the wording of the risks or debate them with a client. When a form version changes, old signatures remain attached to the old version, and that is what keeps the record honest.
What belongs in treatment history, and what should the agent never write?
Treatment history supports both rebooking and safety checks. The administrative fields are straightforward: date, treatment type, area, product, the batch identifier where your practice records it, the provider who performed the treatment, and the interval the provider recommends before the next visit. These fields let the agent calculate when a rebook message is appropriate.
Clinical observations are different. Notes about how the skin responded, whether a client reported a reaction, or whether a plan was adjusted belong to the licensed provider. The agent should be able to see that such a note exists, because its existence is a reason to hold a message. It should not write, summarize, or reinterpret the note. Give it read access to a flag, not to the clinical narrative, unless your counsel and provider decide otherwise.
Illustrative scenario: a client books a follow-up eight weeks after a treatment. The agent checks three things: a signed consent exists for that treatment category, the provider has entered a recommended interval, and no open flag is attached to the prior visit. The prior visit carries a flag from the provider noting that the client asked about a reaction. The agent does not send the standard rebook offer. It places the case in the staff queue with the flag attached, and the front desk contacts the provider before anything else happens.
Which rebook rules can a medical director approve?
Rebooking is where an agent saves the most administrative time, and where a poorly written rule causes the most damage. Lash and brow studios face a similar pattern, where a patch test plays the role that a consent check plays in a med spa. A good rebook rule is simple enough to read aloud in a staff meeting. Build each rule with these steps:
- Write one rule per treatment category, naming the interval, the approver, and the date the rule was last reviewed.
- Define exclusions in plain terms: an open adverse event flag, a pending provider review, an expired or missing consent, a client request to stop contact, or a treatment plan still in progress.
- Write message templates in the practice's voice, and have the medical director approve each one before it is used.
- Limit channels and contact hours to what the client has agreed to, and record that agreement in the client record.
- Require a person to approve any message that mentions results, symptoms, pricing exceptions, or a change of treatment.
- Log every message sent and every message held, with the rule that produced the decision.
Keep the rule set small. A long list of exceptions is hard to maintain, and the agent will be only as careful as the least clear rule it has been given.
Where must the agent stop and hand the case to a person?
Stop points should be specific enough that the agent can recognize them without exercising judgment. The agent should not decide that a message is probably fine. If a client message touches a listed category, the agent acknowledges receipt, does not answer the clinical part, and routes the case to licensed staff with the conversation attached. Route these categories to a person:
- Questions about whether a treatment is safe given a medication, a condition, pregnancy, or a recent procedure
- Descriptions of redness, swelling, pain, bruising, or any other reaction after a visit
- Requests to change an interval, product, dose, or treated area
- Requests to skip, shorten, or sign any consent step on the client's behalf
- Questions that compare the client's result with another person's or promise a particular outcome
Client messages can also contain instructions that try to override the rules, such as a note telling the agent to ignore the consent requirement because the client is in a hurry. The OWASP Top 10 for LLM Applications lists prompt injection as a primary risk for exactly this reason. Keep the rules in configuration that the conversation cannot edit, and make the agent's permissions narrower than its helpfulness. OWASP's authorization guidance supports the same principle: access should be granted only as far as each task requires.
How does Be Human handle this in managed operations?
Be Human starts with one defined role for the practice, rather than a general assistant that can do everything at once. The first engagement maps which consent and treatment records the agent may read, which fields it may write, where medical staff must approve before anything happens, and one observable success criterion. Managed operations then keep the setup accurate over time: monitoring of sent and held messages, recurring reviews of exclusion rules with the medical director, permission reviews when staff change, and incident response if a message goes out that should have been held. The service is remote and available in English and Portuguese.
Where should a med spa start this month?
Pick one treatment category that has a clear interval and a consent form already in use. Then work through the records for that category before changing anything else:
- Confirm the current consent form version for that category and where signatures are captured
- Check whether each signed consent links to its form version and its date
- List every rebook message sent last month and the rule that produced it
- Identify which clinical flags currently stop a rebook message, and who clears them
- Ask the medical director to approve the exclusion list in writing before any agent uses it
If you want a starting point, take the free Be Human AI diagnostic to find the first task worth handing to an agent.
Frequently asked questions
Can an AI agent answer client questions about aftercare? It can share the approved aftercare sheet for the treatment category and explain when the provider's team will follow up. It should not offer its own medical advice about swelling, bruising, or timing, and any question about symptoms should go to licensed staff with the conversation attached.
Does consent have to be digital for an agent to work with it? Not necessarily. Your counsel and medical director decide which consent formats are valid where you operate. What the agent needs is a record it can read, showing the form version, the treatment, the date, the staff member involved, and whether the consent is still current, regardless of whether it began on paper or in a digital form.
How much access should the agent have to treatment records? Give it the narrowest access that lets it do its job: read the administrative fields and the presence of clinical flags, write to scheduling and message logs, and leave clinical notes to licensed staff. Review those permissions whenever someone joins or leaves the team, and record each change.
What if a client tries to talk the agent out of a consent step? The agent should have no authority to waive a consent, so the request has no effect on the record. It should acknowledge the message, explain that the team will handle the request, and route the case to staff with a note. Rules stored in configuration, outside the chat, are what keep this boundary firm.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


