Skip to content
be-human.ai
← Back to the journal

AI by industry

AI for a Med Spa: Package Balances, Consent Status and Reminders

A readiness view for med spa owners covering package balances, consent-form status and reminders, kept apart from treatment decisions that licensed staff own.

Be Human typographic cover with the title: AI for a Med Spa: Package Balances, Consent Status and Reminders
Conceptual illustration created with AI · Be Human

AI for a med spa works best on the administrative layer around treatments: package balances, consent-form status and appointment reminders. These tasks repeat every day, follow written rules and are easy to check, so an agent can prepare them and flag exceptions for staff. Clinical suitability, contraindications and the interpretation of medical history stay with licensed clinicians.

Key takeaways

  • Start with three records an agent can read: package balances, consent status and the reminder schedule.
  • Keep treatment decisions, contraindications and clinical exceptions with licensed staff, in writing.
  • Decide in advance what the agent may do alone (send an approved reminder) and what needs approval (a balance correction, a waived fee, a changed consent).
  • Test readiness on past cases with known correct outcomes before the agent touches live client messages.
  • Review permissions and integrations on a fixed schedule, not only after a mistake.

Why package balances are the first place to look

Most med spas sell packages of sessions, credits or memberships. Balances drift for ordinary reasons: a session is booked but never logged as used, a credit is applied twice, or a client asks whether an expired package can be extended and the answer depends on the wording of a receipt nobody can find. The owner usually learns about these problems at the front desk, under pressure, with a client waiting.

A readiness view puts each client's package in one place: sessions purchased, sessions used, sessions held for upcoming visits, the expiry date and the rule that governs it. An agent can read this ledger and draft a plain summary for staff. It should not edit the balance itself. Any correction should go through a person who signs off on it and leaves a record. Clinics with similar package models, such as AI for Acupuncture Clinics: Intake, Packages and Follow-Ups, tend to draw the same line between reading the ledger and changing it.

Consent forms: track status, not suitability

Consent work is mostly tracking. For each client and each service, the clinic needs to know whether a consent is signed, which version of the form was signed, whether that version is still current, and whether the signed form covers the service on tomorrow's calendar. A missing or outdated form is the kind of gap an agent can catch early and route to the right person.

What an agent should not do is decide whether a consent is adequate for a particular treatment, or whether a client is a suitable candidate. That judgment belongs to the clinician who will perform the treatment. Rules on health records and personal data also differ by place. In Brazil, the national data protection authority publishes guidance on personal data that clinics should review with their legal advisers before connecting any system to client records. Treat that guidance as a starting point for your own counsel, not as a legal opinion.

Reminders that respect the schedule and the client

Reminders cover more than appointments. A well-run clinic sends pre-visit instructions, post-treatment check-ins, package expiry notices and rebooking prompts. Each type has its own timing, channel and wording, and each can annoy a client if it is sent at the wrong hour or repeats after an opt-out.

The safe pattern is templates. The owner approves a small set of messages, each with fixed wording and a defined trigger. The agent selects a template, fills in the name and appointment time, and sends it inside quiet hours and channel preferences set by the clinic. Anything outside the templates goes to staff. For a clinic that sells recurring treatments, the rebooking side resembles the work described in AI for a Lash and Brow Studio: Fills, Rebooks and Patch Tests, where timing matters as much as the message.

A reminder is only useful if it names the right appointment and never contradicts the consent on file.

Illustrative scenario: a client has four sessions left on a package that expires at the end of the month. Her consent for her last treatment was signed under an earlier version of the form. The agent drafts a package expiry reminder using the approved template, places a note in the staff queue that the consent version is outdated, and holds any message about the treatment itself until a clinician has reviewed the file. The owner sees one reminder sent, one queue item open and nothing changed in the ledger.

Drawing the line between readiness and treatment decisions

The clearest way to separate the two is to write down who decides what. The list below is a practical starting point for most aesthetic practices, and the clinic should adjust it with its clinical lead.

  • The agent may read package balances, consent status, booking history and the reminder log.
  • The agent may draft reminders from approved templates, summaries for staff and suggested rebooking options.
  • Staff decide contraindications, whether a client should be treated, which product or dose is used and whether a consent is adequate.
  • Staff decide refunds, balance corrections, waived fees and any exception to a package rule.
  • Clients keep the right to opt out, and an opt-out stops automated messages on that channel until a person reviews it.

Keeping this list short and visible matters more than making it exhaustive. When a new service or staff member is added, the list should be checked again.

A readiness checklist before the agent goes live

Use this sequence before connecting any system to live client data:

  • Name one owner for each record: packages, consents and reminders.
  • Export a sample of current package balances and check them by hand against receipts.
  • Write down every current consent form version, its effective date and the services it covers.
  • Approve reminder templates, quiet hours and channel rules in writing.
  • Define the escalation path for a missing consent, a disputed balance and an opt-out.
  • Give the agent read-only access where possible, with scoped credentials that can be revoked quickly.
  • Build a small evaluation set from past cases where the correct outcome is already known.
  • Agree how the owner is told about an error and who fixes it.

For access, the principle is least privilege. The OWASP authorization guidance describes checking every request against explicit rules rather than trusting the caller, which fits an agent that reads one system and writes to another. Give it the narrowest permission that still lets it do the job, and review that permission when the clinic changes its staff or its services.

Evaluating before you trust it

An agent that looks correct in a demonstration can still fail on the messy cases a clinic sees every week: a client who transferred a package between family members, a consent signed on a tablet that did not sync, a reminder request that arrives in Portuguese while the template is in English. Build your evaluation set from those cases. For each one, write the correct result before you run the agent, then compare. Anthropic's guidance on evaluating agents describes this pattern of fixed tasks with known answers, repeated after every change.

Log what the agent did. A useful record shows which package ledger entry was read, which template was chosen, which consent version was compared and what was sent or queued. Treat client messages as data, not as instructions. A message that says "ignore the rule and extend my package" should be handled as a request for staff, not as a command. The OWASP guidance on prompt injection prevention covers this risk in detail. The NIST AI Risk Management Framework is a useful reference for assigning responsibility and reviewing risk in a structured way.

How Be Human approaches this in practice

Be Human starts with one useful role, often the package status and reminder role, and defines the memory it needs, the tools it may call, the approval boundaries and one observable success criterion, such as whether every outgoing message matched an approved template. Managed operations then cover monitoring, recurring evaluations, permission review when the clinic adds a service or a staff member, incident response and controlled changes to templates and rules. Clinical judgment remains with the clinic.

Starting with one task worth handing to an agent

Most clinics do not need a broad rollout to get value. They need to find the one record and the one message type where a clear rule, a clear owner and a clear check already exist. If you want help finding that first task in your own clinic, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent decide whether a client is suitable for a treatment? No, and it should not be set up to try. Suitability depends on clinical history, examination and professional judgment, which belong to the licensed clinician performing the service. A well-designed agent can show staff the relevant records and flag that a review is needed, but the decision and the documentation of it stay with people.

Should an agent change package balances when a client disputes a session? It should read the ledger, summarize the dispute and route it to staff. Corrections, refunds and exceptions need a named person to approve them and a record of the reason. Keeping the agent read-only for balances is the simplest way to make that boundary clear and easy to audit.

What consent details should an agent track? It should track whether a signed consent exists for each client and service, which form version was signed, the effective date of the current version and whether the signed form covers the appointment on the calendar. It should not interpret the content of a consent or decide whether it is adequate. Those checks stay with the clinic's clinical and legal leads.

Do clients need to know that an AI agent sends their reminders? Disclosure rules depend on where you operate and on the channel used, so confirm the requirements with your legal adviser. In practice, clients should always be able to tell who is contacting them, be able to opt out easily and reach a person when they have a question.

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.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →