Superintelligence Prep for Med Spas: Consults, Packages and Consent Forms
AI agents for med spas need accurate consult notes, package balances and consent status. Organize those records first, then set approval limits.

AI agents for med spas can answer booking questions, follow up after consults and flag package balances, but only when the records behind those answers are complete and specific. Before an agent touches the front office, a med spa needs three things in order: consult notes written in a consistent structure, one ledger for prepaid packages, and a consent-form status that shows what is signed, what is missing and who may decide. This guide explains how to organize those records so that any future increase in automation works from facts rather than from one employee's memory.
Key takeaways
- Consult notes should separate what the client reported, what the clinician observed and what was recommended.
- Package balances belong in one ledger that every channel reads from, not in a spreadsheet, a point-of-sale memo and someone's memory.
- Consent status should be tracked per treatment, with five defined states, not as a single yes or no on the client profile.
- An agent should read these records, answer within them and route exceptions, while clinical judgment and consent decisions stay with licensed staff.
- Write the approval boundary before writing any prompt, and test the agent against past conversations before it goes live.
Why a med spa front office runs on three records
A med spa front office hears the same questions every day. Can I move my appointment? How many sessions are left on my package? Did I sign the form for the new treatment? What did the injector recommend last time? Each answer lives in a different place: the clinical note, the sales ledger, the consent folder, or a staff member's recollection of a phone call. An agent that answers from only one of those sources will sound confident and will sometimes be wrong.
The fix is not a more capable model. It is a set of records that agree with each other and that the agent is permitted to read, with the limits of each action written down beside them. Be Human starts every engagement from that ordering, because the quality of an agent's answers depends mostly on what it is allowed to see and what it is allowed to change.
Consult notes an agent can actually read
A consult note written only for the clinician's own memory is usually too loose for an agent to use safely. A structure that works for both people and software has five parts:
- Client-reported goals, written in the client's words and dated to the conversation.
- Clinician observations, kept separate from anything the client said.
- Contraindications and open questions, with unanswered items explicitly marked as open.
- The recommended treatment, named the way your clinic names it in its protocols.
- The next step and its owner: clinician, front desk or client.
Keep free text for nuance, but place open items in their own field. That lets an agent tell a client honestly that a follow-up question is still pending, instead of guessing whether the consult is complete. Decide in advance who may edit a note after the visit. In most clinics the agent should be able to append a message to the client thread and to flag a note for review, but not to rewrite the clinical record itself.
Package balances: one ledger that every channel reads
Package questions cause more friction than almost any other request, because a client's memory of a package and the clinic's record rarely match exactly. The goal is a ledger that a clinician, a front desk employee and an agent all read the same way.
1. List every package you sell, with the treatment units it includes and any expiry rule, written in plain language. 2. Record each purchase as a ledger entry with the client identifier, the package type, the units purchased and the payment reference. 3. Debit a session at the time of treatment, by the clinician or the front desk, not at the end of the day. 4. Record every adjustment (transfer, comped session, refund, extension) as a separate entry with a reason and the name of the approver. 5. Let the agent read the current balance and the last few movements, and have it quote only that recorded balance, never a balance it calculated itself. 6. Route any disagreement between the ledger and the client's account to a named staff member, with both facts attached.
If your clinic also runs intake and follow-up workflows for packaged care, the same ledger logic applies. The acupuncture case in AI for Acupuncture Clinics: Intake, Packages and Follow-Ups shows how that structure supports follow-up messages without rewriting the balance.
Consent-form status: five states and one decision rule
Consent is not a checkbox on a client profile. For each treatment, track status separately, because a client may have a current form for one procedure and a pending form for the next. Use five states:
- Not required: the treatment has no consent form in your protocol.
- Requested, not sent: staff has identified the form but has not issued it.
- Sent, not signed: the form has been issued and is awaiting the client's signature.
- Signed and current: the form is signed and matches the version your clinicians use today.
- Expired or superseded: a form was signed earlier, but the protocol changed or your rules require renewal.
Then give the agent a simple decision rule. If the status is signed and current, the agent may confirm the appointment and send a copy on request. If it is sent but not signed, the agent may send the signing link and remind the client that the booking depends on it. If it is requested but not sent, or expired, the agent routes the conversation to clinical staff and does not confirm the treatment slot. Whether a particular consent is sufficient for a particular procedure is a question for your medical director and your local regulations, not for the agent. The same discipline applies to intake for other clinical services, as described in AI Agent for a Chiropractic Practice: Intake and Visit Recalls.
Illustrative scenario: a package question on a Tuesday
Illustrative scenario, not a client case: A client writes to ask whether she has enough sessions left for a treatment next week. The agent reads the ledger, which shows the balance after her last visit and one comped session recorded with the approver's name. It replies with that balance and the expiry date from the package rule. It then checks consent status for the requested treatment, finds it sent but not signed, and sends the signing link with a note that the booking can be confirmed once the form is complete. It does not reinterpret the balance or promise the session. The front desk sees the full thread, with the ledger entry and the consent status attached to the conversation.
The scenario works because each answer came from a record with an owner, not from a guess. Without the ledger and the consent states, the same message would require a staff member to search three systems before replying.
How Be Human runs this after launch
Be Human designs the memory and tool access for these records as part of a first engagement, then defines the approval boundary for each action. A typical boundary allows the agent to quote a recorded balance, send an existing form link and propose a reschedule, while a clinical note change, a fee waiver or a consent exception always goes to a named person. Managed operations then keep the arrangement accurate over time: recurring checks that the ledger and the agent still agree, permission reviews when staff roles change, and controlled updates when a treatment protocol or a form version changes. Test cases drawn from real past conversations form a short set that must pass before any change is released. The approach follows the risk habits described in the NIST AI Risk Management Framework, which emphasize mapping risks and assigning owners before deployment.
For the rebooking side of the same front desk, including how fills and patch tests are handled, see AI for a Lash and Brow Studio: Fills, Rebooks and Patch Tests. The sequence there mirrors the one described above: records first, then permissions, then the agent.
Take the free Be Human AI diagnostic to find the first task in your med spa worth handing to an agent.
Frequently asked questions
Do we need a new practice management system before an agent can help? Not necessarily. Many clinics can start by defining the consult structure, the package ledger rules and the five consent states in the systems they already use. The first step is to make the existing records consistent and readable, because an agent can only be as accurate as the data it reads.
Can an AI agent tell a client how many package sessions remain? It can, provided it quotes the balance from the ledger rather than calculating its own. The agent should also show the expiry rule and the most recent movements, and it should route any disagreement to a named staff member. Adjustments such as comped sessions should already be recorded with an approver before the agent reads them.
Should the agent decide whether a client can be treated when consent is incomplete? No. The agent can confirm the status, send the signing link and explain that the booking depends on the form. Whether consent is sufficient for a specific procedure belongs to your medical director, your clinical staff and your local regulations, and the agent should route those decisions to them.
How often should these records be checked after launch? Check the agreement between the ledger, the consent states and the agent on a recurring schedule, and review permissions whenever staff roles or treatment protocols change. Keep a short set of past conversations as tests, and run them again after every change to the agent, the records or the form versions.
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.


