Top 10 AI Agents forMed Spas and Aesthetics 2026

The claim that sells AI into a med spa is that the software is HIPAA compliant and the agent books appointments. At two of the most-recommended platforms in this category, HIPAA coverage is not switched on when you sign up, and the best-documented AI receptionist on the market cannot book an appointment by voice. Both statements come from the vendors' own documentation, not from a competitor, not from a review site, and not from us. They are the two findings that reorganised this entire ranking, and neither appears in any competing list we could find.

Mangomint's published compliance page states plainly that HIPAA compliance is not automatically enabled when you sign up, and that you must request HIPAA enablement by contacting support; the business associate agreement is then provided for you to review and sign. Boulevard's support documentation treats HIPAA Coverage, which it describes as including signing a business associate agreement in your Boulevard dashboard, as a paid upgrade rather than part of the base subscription, and lists its Medspa Add-On at $65. Neither is a scandal. Both are documented in public, which is more than most vendors on this roster manage. But read them together and the operational consequence is unambiguous: a clinic that injects prescription neuromodulators, fires energy devices at people's faces and stores patient photographs can be running live on either platform with no business associate agreement in force, and nothing in the product will tell it.

The second claim under test is about autonomy. Boulevard publishes a frequently-asked-questions page for its AI Receptionist that is, by some distance, the most honest piece of vendor disclosure in this vertical. It states that the product is in beta. It publishes the price: $125 for 200 minutes, $0.60 per additional minute, standard text rates for the booking links. And it states that the product cannot book directly by voice - it sends a booking link and the client completes the booking themselves. Reschedules and cancellations also go through portal links. The same page lists what it cannot do: group appointments, memberships, packages, gift cards, consultation-first workflows, responding to incoming texts, or making outbound calls.

Read that list next to how a med spa actually earns money and stays compliant. Memberships and packages are the med-spa revenue model - an entire category of software exists to sell pre-paid treatment credits. Consultation-first is the workflow that keeps a new patient in front of a licensed prescriber before anything is injected. The one product in this market that publishes its own limits publishes limits that exclude the two things a med spa most needs. That is not a criticism of the disclosure. It is an argument for reading it, and for treating a vendor that publishes limits as more credible than one that publishes a conversion rate.

So this article does something different from the rankings you have already read. It does not score accuracy, conversion or revenue lift, because no independent benchmark of any of these commercial products exists - not one, anywhere in aesthetics. It scores what an owner can verify without leaving the front desk: what price is published, what security attestation sits on a page you can open, whether a business associate agreement is offered at all, whether an API is documented publicly, whether the product is standalone or locked to a suite, and who owns the company today according to public record. Every cell carries a citation or the literal words not publicly disclosed. If you want the cross-industry version of this method, our review of AI agents across ten industries in 2026 applies the same refusal everywhere, and the healthcare-specific version sits in our ranking of AI agents for medical practices, where the binding constraint is a read-only certified API rather than a state medical board.

We scored eight attributes, every one of which you can re-check at a public URL, and we refused to score anything a vendor publishes about its own performance. That refusal is not modesty. In this market every conversion rate, answer rate, growth multiple and revenue-lift figure in circulation was published by the company selling the software or by an SEO page with no source at all.

The conflict of interest that shapes every ranking you have already read

Several of the highest-ranking “best medspa software 2026” pages in search are published by vendors who appear in their own rankings. At least four platforms on our roster publish comparison articles that rank the category. A vendor comparing itself to its competitors is producing marketing, and it is not a source. We cite none of them for anything, including for facts about themselves that appear only in a comparison page rather than in product documentation. If you have been comparing products by reading such pages, you have been reading four different companies' sales decks arranged as a table.

What an independent benchmark would look like, and what exists instead

No independent benchmark of any commercial AI product sold to med spas exists. That is a precise claim rather than a blanket denial, and the distinction matters because two third-party assessments genuinely do exist in aesthetics - they simply are not benchmarks, and neither measures an AI agent at all.

• A customer-satisfaction award: Best in KLAS 2026 for Ambulatory Plastic Surgery, claimed by Symplast on its own EHR page. KLAS is a genuinely third-party research firm, which makes this the closest thing to independent recognition in the vertical. But KLAS ratings are built from satisfaction surveys of paying customers rather than from a capability test, and the full report is paywalled. It tells you that users like a product. It tells you nothing about what the software outputs.
• A single-case review of an imaging system: The only peer-reviewed literature naming any product on or adjacent to this roster is a single-case review of a facial imaging system in Aesthetic Surgery Journal Open Forum. A single-case review has an n of one. It is a case report, not an evaluation, and emphatically not a benchmark. That the entire aesthetics technology market's peer-reviewed evidence base is one case report is the most honest sentence available about the state of evidence here.
• What does not exist: No head-to-head comparison of these products by any party that does not sell one of them. No audited accuracy figure. No third-party measurement of booking conversion, call handling or charting quality for any vendor in the table. When a vendor quotes you a number in this category, the correct follow-up question is not whether the number is high but who measured it, against what denominator, and whether you may see the method.

What happened the one time a regulator audited a vendor's AI metrics

There is a reason to hold vendor performance numbers at arm's length that is stronger than professional caution. The one time a US federal regulator audited a public company's published AI performance metrics, it found them materially misleading. In In the Matter of Presto Automation Inc., Securities Act Release No. 11352, Exchange Act Release No. 102177, Admin. Proc. File No. 3-22413, entered 14 January 2025, the SEC addressed a drive-thru voice AI product the company had told investors delivered over 94% accuracy even in noisy environments and 95% to 99% automated order completion.

The product “lacked the capability to take orders on their own and required substantial human involvement,” with “human order takers located abroad (primarily in the Philippines and India), who processed the vast majority of drive-thru orders.”

— SEC, In the Matter of Presto Automation Inc., Release No. 33-11352 (14 January 2025), paragraph 3

Four things must be said precisely, because that order is frequently mis-summarised. Presto consented without admitting or denying the findings - it never admitted them. The remedy was a cease-and-desist order with no civil penalty; the company was not fined. The findings are against Presto only, and the order's “Supplier A” is Hi Auto, against whom the SEC made no findings whatsoever. And Presto Automation Inc. is not Presto Phoenix Inc., which acquired assets in December 2024. Held precisely, the order is still the strongest available argument for this article's method: it is what happened the one time anyone with subpoena power checked a published AI performance number.

The aesthetics equivalents are already in circulation. Searching for med-spa voice agents surfaces a dense layer of pages publishing precise-looking figures - answer rates rising from 60% to 99%, booking rates from 36% to 52%, a net revenue lift of $25,000 to $60,000 for a single location - with no source of any kind, several of them published by vendors ranking themselves and one of them a paid press-release wire item. A paid wire is not a publisher. We cite none of them, for anything, and neither should your vendor.

Every cell below was read off the vendor's own page on 23 August 2026, or says “not publicly disclosed.” Nothing here is copied from a review site, an aggregator or a competitor's comparison article. Where a figure exists only in a data aggregator, we say so rather than presenting it as a filing.

Ownership, lock-in and documented integrations

Price opacity in this market tracks ownership structure, not product quality. The aesthetics-native EMR and the self-serve tools publish prices; the private-equity-owned enterprise platforms do not. That is not an accusation - enterprise software is often genuinely priced by footprint - but it is a pattern a buyer should recognise before concluding that the expensive-feeling product is the serious one.

One methodological note on the API column. A 404 or a non-resolving developer subdomain is notproof that no API exists; several vendors gate documentation behind a partner portal. The defensible claim, and the one we make, is that no publicly documented API was found at the standard developer paths on 23 August 2026. On that criterion Zenoti is the clear leader, and there is an irony worth one sentence: the capabilities Zenoti documents publicly - memberships, packages, gift cards - are precisely the ones Boulevard's AI Receptionist FAQ says it cannot handle.

What we refused to print, and why

A ranking that refuses vendor claims and then quietly repeats an industry statistic has learned nothing. Here is every figure we encountered, chased and declined to print.

These five are the systems that hold the chart, the schedule and the money. Whatever agent you buy, it will either live inside one of these or have to integrate with one. That is why the compliance and API columns matter more here than any feature list: a platform you cannot get data out of, and cannot get a business associate agreement from, is a platform you cannot safely put an agent on top of.

These five sit around the record rather than holding it - marketing, commerce, imaging and specialty EHR. Three of them are owned by private equity, and in two cases the product an operator originally bought as a standalone tool is now a module of a suite. That pattern is the most concrete, dated, checkable argument in this article about why ownership belongs in a buying decision.

The exclusions are more useful to a buyer than three more table rows, because most of them appear as independent vendors in rankings you have already read. Each one below is named, with the reason and the date.

• TouchMD - acquired by Nextech, announced 11 October 2022: Not an independent vendor and has not been one for nearly four years. It is Nextech's imaging and consultation module, with a product page on Nextech's own site. It still markets under its own name and holds its own domain, which is exactly why it survives in so many rosters. Ranking TouchMD and Nextech separately is the single clearest signal that a list was copied rather than checked, and it is the error most likely to get an article dismissed by an operator who already owns both.
• RxPhoto - now a PatientNow module: The clinical before-and-after imaging product was acquired and folded into PatientNow. Both of the acquisitions on this roster did the same thing: converted a standalone imaging tool into a module of a private-equity-owned suite. Imaging is where this market consolidates. An operator who bought either as a point solution now depends on a platform they did not evaluate.
• Pabau - trading, but ownership could not be confirmed from public record: A UK-origin clinic-management platform, actively publishing in 2026, that we could not place on the table because no funding, ownership or corporate-record confirmation was located within this review. Ownership from public record is one of our eight scored attributes and the corporate record for a UK entity is public, so the honest position is that we did not complete the check rather than that the company failed it. It also publishes its own competitor comparison articles, which we do not use as a source for anything.
• Moxie - freshly funded, but not software you buy as an agent: Founded 2022, it raised a $25 million Series C led by Viewpoint Ventures in March 2026, taking total capital to roughly $51 million, and has launched products named Scale Suite and Compliance Defender. It is excluded because it is a medspa-launch and business-services model aimed at nurses opening practices rather than a product an operating clinic buys as an agent - it sells the management-services scaffolding itself. Its claim that supported practices grow 60% faster than the industry average with 10% higher margins is refused outright: no method, and no denominator exists for the industry average it claims to beat.
• Podium - horizontal lead-response automation, not aesthetics software: Lehi, Utah, private, marketing an AI Employee for lead response and messaging across auto, home services, retail and healthcare. Med spa is one segment among many. Every figure available about it - valuation, capital raised, employees, customer count - comes from data aggregators or the company itself, with no filing behind any of them. There is also a name-collision trap: at least two unrelated companies with similar names have separate investor-database profiles, and a writer grabbing the wrong one would print a false funding fact.
• Cherry and PatientFi - financing, not agents: Both are real and trading in 2026, and both are patient-financing products rather than AI agents, so neither belongs in an agent ranking. Their published conversion and approval figures are vendor claims and we refuse them. One adjacent point is worth an operator's attention: consumer-financing disclosure sits close to the advertising rules discussed below, so if you promote financing on your site, the promoted terms and the typical terms should match. We deliberately do not print the interest-rate figures circulating in this space, because the only place we found them was a direct competitor's blog rather than the rating agency report they were attributed to.
• Canfield Scientific - imaging hardware and software, and the sole peer-reviewed citation: A facial imaging platform, still trading, whose ownership reportedly changed in 2022 - we did not confirm that from a primary source, and ownership is a required field, so we do not rank it on an unverified cell. It is worth naming for one reason: it is the only vendor anywhere near this roster with peer-reviewed literature naming its product, and that literature is a single-case review. An n of one. That the entire aesthetics-technology market's peer-reviewed evidence base amounts to one case report is the most useful thing an operator can know about how much evidence to expect from anyone here.

A med spa is a medical practice, not a beauty business, and that single fact decides whether any of this software is safe to deploy. Botulinum toxin, dermal fillers, laser and energy devices and IV therapy are the practice of medicine. In corporate-practice-of-medicine states the clinic must be owned by a licensed physician - typically through a professional corporation - with any non-clinical investor confined to a management services organisation that sells administration and never touches clinical judgment. Every patient must receive an appropriate examination by a licensed prescriber before a prescription treatment. An AI agent that recommends a treatment, tells a caller whether they are a candidate, or books a specific procedure is operating at the edge of that line - and the line is drawn by state medical boards and state statute, not by the software vendor.

The California statutes, quoted properly

Three sections of California law do most of the work, and secondary sources routinely misquote all three. Business and Professions Code section 2400 is the corporate-practice bar itself.

“Corporations and other artificial legal entities shall have no professional rights, privileges, or powers.”

— California Business and Professions Code section 2400

Note the exact wording - artificial legal entities. A great many secondary sources drop the word “legal.” If you quote it, quote it right. Section 2052(a) is the provision that makes cosmetic work medicine, and it does so through a single word that most articles never reach: it applies to anyone who, without a valid certificate, diagnoses, treats, operates for, or prescribes for any ailment, blemish, deformity, disease, disfigurement, disorder, injury, or other physical or mental condition. Blemish is the statutory hook, and it is far more precise than the usual hand-wave that Botox is a medical procedure. The same section also reaches anyone who advertises or holds themselves out as practising - which is directly relevant to an agent making clinical-sounding claims on your behalf.

The ownership arithmetic is also commonly stated backwards. Corporations Code section 13401.5 does not say physicians must own 51%. It provides that certain other licensed professionals may be shareholders, officers, directors or professional employees of a professional corporation so long as the sum of all shares owned by those licensed persons does not exceed 49 percentof the total shares, and so long as their number does not exceed the number of persons licensed in the primary profession. Same arithmetic, materially different statement. Write it the statute's way.

What it actually does matters more to an agent buyer than the invented version would. Section 1191(a)(1) prohibits a private equity group or hedge fund involved with a physician or dental practice from interfering with professional judgment, including determining what diagnostic tests are appropriate, the need for referrals, responsibility for the ultimate overall care of the patient, and determining how many patients a physician or dentist shall see in a given period of time or how many hours a physician or dentist shall work. Section 1191(a)(2) bars exercising control over, or being delegated, owning or determining the content of patient medical records, hiring and firing clinicians based on clinical competency, setting payer-contracting parameters, making decisions regarding the coding and billing of procedures, and approving the selection of medical equipment and supplies.

Three further provisions close the loop. Section 1191(b) provides that the corporate form - sole proprietorship, partnership, foundation or a corporate entity of any kind - does not affect applicability, so the management-services wrapper is not a shield. Section 1191(c)(2) makes any contract provision violating subsection (a) void, unenforceable and against public policy. Section 1191(e) gives the Attorney General injunctive relief, equitable remedies, and attorney's fees and costs. And section 1191(g) settles the relationship to the older doctrine: the section does not narrow, abrogate, or otherwise lower the bar on the corporate practice of medicine or dentistry as set forth in the Business and Professions Code.

Scope precision, because overstating this is its own error.SB 351 binds private equity groups and hedge funds as defined in section 1190. It is not a general prohibition applying to every management services organisation or every software vendor, and the definition of hedge fund expressly excludes passive contributors, pure debt financiers such as banks and credit unions, and hospitals and hospital systems. Say “private equity or hedge fund,” not “management companies” generally.

The good faith exam - and the claim that is simply backwards

A claim circulating widely in 2026 med-spa vendor content says California's medical board has moved to disallow static forms and text-only chat for good faith exams. The statute says close to the opposite. Business and Professions Code section 2242(a) provides that prescribing, dispensing or furnishing dangerous drugs without an appropriate prior examination and a medical indication constitutes unprofessional conduct, and then continues:

“An appropriate prior examination does not require a synchronous interaction between the patient and the licensee and can be achieved through the use of telehealth, including, but not limited to, a self-screening tool or a questionnaire, provided that the licensee complies with the appropriate standard of care.”

— California Business and Professions Code section 2242(a)

The asynchronous form is expressly permitted, the statute names a self-screening tool and a questionnaire by category, and the only condition it attaches is compliance with the standard of care. Present both halves in that order, because they cut in opposite directions: a practice told falsely that asynchronous intake is banned will buy a product it does not need, and a practice told the truth still has to meet the standard of care, which is where the real exposure lives.

Two boundaries must be stated or this becomes its own overreach. Section 2242 is about prescribing dangerous drugs without a prior exam. It is the right hook for neuromodulators and fillers, which are prescription products. It is not a general good-faith-exam statute covering every laser or facial, and where “good faith exam” is used as a term of art it is largely a board-and-enforcement construct plus scope-of-practice rules rather than a single uniform statutory phrase - so no national standard exists to cite. Second, and worth knowing before you read another comparison table: every multi-state good-faith-exam source we located was operated by a company selling good-faith-exam or medical-director services. Each profits from the strictest possible reading, none publishes a source for its state-by-state tables, and that genre is the likely origin of both the false SB 351 claim and the forms-banned claim. We cite the statute instead.

Two states that changed the rules, read from the bills

Indiana enacted the first comprehensive state med-spa registration regime.Senate Bill 282 (2026) adds a Medical Spas chapter to the Indiana Code, marked effective 1 July 2026, with the registration duty itself beginning 1 January 2027 and the licensing board required to establish the registration procedure no later than 1 October 2026. Its statutory definition is the best one available anywhere: a facility that offers medical health care services, prepares, administers or dispenses prescription drugs or uses them for intravenous, intramuscular or subcutaneous delivery, and holds itself out as focused on cosmetic or lifestyle treatments - expressly including weight loss, wellness and longevity, botulinum toxin injections and dermal fillers, hair loss, hormone therapies, parenteral nutrient therapies, and the nonsurgical use of a laser or other energy device for cosmetic purposes including rejuvenation, anti-aging or hair removal. Note the structural choice: a physician's own office is excluded, so the regime targets the non-physician-fronted med spa.

The operating requirements are the part that touches software. A registrant needs a responsible practitioner with prescriptive authority and training in the services performed, who must be physically present at the location for a sufficient amount of time to ensure compliance. Serious adverse events must be reported within 15 days, with the notice including the patient name, the medication or treatment, the date, the nature and location, and the patient's medical records. Services may not be delivered anywhere other than the registered office except for education or training, which ends the pop-up injectable party. And the registrant must comply with the state's advertising requirements, with suspension of the registration available as a penalty. That is directly on point for an AI marketing agent: from 2027 in Indiana, copy generated by an automated system is the registrant's regulatory exposure, and the penalty reaches the registration that permits the business to operate at all. One caveat we will not hide - we read the engrossed text, and the enrolled act carries a public law number; confirm section numbers against the enrolled version before relying on them.

Texas HB 3749, known as Jenifer's Law, is narrower than every summary claims. Secondary sources list it among general med-spa statutes. It is not one. Its caption concerns the regulation of the provision of elective intravenous therapy, and it adds a single chapter to the Occupations Code. It defines elective intravenous therapy as administering fluids, nutrients, medications or blood directly into the bloodstream through a vein, sought by the patient to alleviate temporary discomfort or improve temporary wellness, and not administered in a physician's office or a licensed facility - so it aims squarely at IV bars and med spas and, like Indiana, exempts the physician's office. Its operative rule is an asymmetry that an AI booking agent can violate in one click: a physician may delegate prescribing or ordering elective IV therapy only to a physician assistant or an advanced practice registered nurse, while administering it may also be delegated to a registered nurse, each under adequate physician supervision. The registered nurse may administer but may not order. The Act took effect 1 September 2025 and applies to acts performed on or after that date regardless of when the delegation was made, so pre-existing standing orders do not grandfather.

Where the AI agent line actually falls

These are the boundary statements we would defend to a board investigator, drawn from the primary sources above rather than from a vendor's compliance blog.

• An agent may answer, schedule, take a deposit, send intake forms and route a message: A licensed prescriber must perform the examination and issue the treatment order. Everything on the agent's side of that sentence is administrative work that has never been reserved to licensees, and it is where essentially all of the available labour saving lives.
• An agent that tells a caller they are a good candidate is diagnosing: Recommending treatment for a blemish is conduct section 2052 reserves to licensees, and the same section reaches holding oneself out as practising. The safe design is that the agent books a consultation, never a procedure, for any new patient - and that restriction should be enforced in the platform's bookable appointment types, not in a prompt.
• An agent must not be the thing that decides how many patients a provider sees: Under California Health and Safety Code section 1191(a)(1)(D) that is a decision a private-equity or hedge-fund-backed entity may not control. A throughput-optimising scheduler configured by a management company rather than by the practice is exactly that decision wearing software.
• An agent must not own or determine the content of the medical record: Section 1191(a)(2)(A) targets precisely that structure. Ambient charting is fine and often genuinely useful; ambient charting that writes to the chart under management-company control rather than under the physician practice's control is the arrangement the statute names.
• The management-services wrapper is not a shield: Section 1191(b) states that the corporate form does not affect the applicability of the section. If the analysis of your agent deployment depends on which entity nominally holds the software contract, the analysis is already wrong.
• An agent must never book past a scope-of-practice boundary: In Texas, a registered nurse may administer elective IV therapy but may not order it. An agent that books a patient straight into an RN-staffed IV appointment with no prescriber order anywhere in the chain has produced a scope violation through a calendar entry. Encode the boundary in the bookable appointment types.

The most legally exposed agent in a med spa is not the one that books appointments. It is the one that does marketing. Med spas run on social before-and-afters, referral incentives, review requests and reactivation campaigns - and the federal rules that govern all four are in force today, carry a per-violation civil penalty, and are violated at machine speed by a system generating copy at scale.

The FTC reviews rule - what it does and does not say

The Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465, took effect 21 October 2024 under authority of 15 U.S.C. 57a, sourced at 89 FR 68077. Start with the correction, because the mis-statement is everywhere: section 465.3 is codified as [Reserved]. The proposed provision on review hijacking was never finalised, so any article telling you the rule bans review hijacking is wrong. What the rule does prohibit is substantial:

• Section 465.2 - fake or false reviews and testimonials: Unlawful to write, create or sell a consumer review, consumer testimonial or celebrity testimonial that materially misrepresents that the reviewer or testimonialist exists, that they used or had experience with the product, service or business, or the reviewer's or testimonialist's experience.
• Section 465.4 - buying positive or negative reviews: Compensation or other incentives conditioned on a review expressing a particular sentiment, whether positive or negative. This is the provision that catches the friendly-looking review request offering a discount, and an agent can send thousands of those before anyone reads one.
• Section 465.5 - insider reviews and testimonials: An officer or manager writing a review without a clear and conspicuous disclosure of the material relationship. An agent drafting a review in a staff member's voice produces exactly this violation, and produces it faster than a compliance review can catch it.
• Sections 465.6 to 465.8 - controlled review sites, suppression, fake influence: Company-controlled review websites; review suppression, which the rule describes as including an unfounded or groundless legal threat, a physical threat, intimidation, or a public false accusation; and misuse of fake indicators of social media influence.
• Section 465.1(c)(4) - the disclosure standard that catches AI-generated content: In an interactive electronic medium the disclosure must be unavoidable, and the rule states that a disclosure is not clear and conspicuous if a consumer must take any action, such as clicking on a hyperlink or hovering over an icon, to see it. A disclosure behind a see-more link is not a disclosure.

The maximum civil penalty per violation under the FTC Act provision behind that rule is $53,088, per 16 CFR 1.98(d), as adjusted at 90 FR 5581 on 17 January 2025 and applying to penalties assessed after that date. There is no 2026 adjustment; do not let anyone quote you a higher figure. One further precision the rule's critics get wrong: the operative text of Part 465 never uses the word “AI.” Coverage of AI-generated content rests on a footnote in the rule's Federal Register preamble. Cite it that way or not at all.

FDA in 2026 - two real, dated actions that changed the risk picture

On 3 March 2026 the FDA announced 30 warning letters to telehealth companies for false or misleading claims about compounded GLP-1 products offered on their websites. The agency named two primary violation types: claims implying sameness with FDA-approved products, and obscuring product sourcing by advertising drug products branded with the telehealth company's own name. The release states that over the preceding six months the agency had sent thousands of letters warning pharmaceutical and telehealth firms to remove misleading ads - more than in the entire preceding decade.

“It's a new era. We are paying close attention to misleading claims being made by telehealth and pharma companies across all media platforms—and taking swift action. Compounded drugs can be important for overcoming shortages or meeting unique patient needs—but compounders should not try to compound drugs in a way that circumvents FDA's approval process.”

— FDA Commissioner Marty Makary, M.D., M.P.H., FDA news release, 3 March 2026

Med spas now sell medical weight management at scale - the plastic surgeons' own 2024 survey counted 837,485 prescribed weight-loss medication procedures in its first year of collecting that category - so this enforcement posture reaches a service line most clinics added recently and market aggressively. An AI marketing agent generating GLP-1 copy is generating exactly the claims in those warning letters, at volume, without a compliance reviewer in the path.

The second action matters even more for what a med spa should automate first. In April 2026 the FDA is reported to have issued its first Drug Supply Chain Security Act warning letter to a medical spa as a dispenser, following a multi-day inspection in December 2025, concerning botulinum neurotoxin products; the reported findings involved discrepancies between units purchased from authorised trading partners and units documented as administered in patient records, and an unlabelled vial lacking the product identifier federal law requires. We did not retrieve the letter itself, so we neither name the business nor quote its findings - but the structural point stands and does not depend on the specifics. DSCSA makes a med spa a dispenser with tracing obligations, and the compliance artefact is reconciliation between purchase records and the chart. That is a genuine, non-hyped automation use case, and notice what it is: an inventory-and-records agent, not a booking bot. It is the best available answer to what a med spa should actually automate first, and it is grounded in an enforcement action rather than a vendor claim.

When the agent picks up the phone or sends the text

Under FCC Declaratory Ruling 24-17, adopted unanimously and released 8 February 2024, AI-generated voices are artificial voices under the Telephone Consumer Protection Act. The ruling expressly refuses any carve out of technologies that purport to provide the equivalent of a live agent. Callers must obtain prior express consent - prior express written consent for marketing - identify themselves and the party responsible for the call, and offer an opt-out.

The trap, and it is the most damaging error available in this article. Insurance Marketing Coalition v. FCC (11th Cir., 24 January 2025) vacated only Part III.Dof the 2023 Order - the one-to-one consent provision - and the court's own footnote states that the 2012 Order is not at issue. The prior express written consent requirement survives, as does the artificial-voice classification. Writing that the TCPA got easier for AI callers inverts the law, and it is advice an operator could act on and be sued for. Separately, the revocation rule at 47 CFR 64.1200(a)(10) is in force: a consumer's plain-language cancel or stop request must be honoured across any channel the caller uses. This bites harder in aesthetics than almost anywhere, because the standard campaigns - reactivation texts, we-miss-you messages, treatment promotions - are marketing, and marketing needs written consent, not merely a phone number sitting in the record. Several states also impose their own consent rules with private rights of action; unless you have read your state's statute, treat that generically and check it before you launch.

If you have EU clients, Article 50 is live law today

The claim that the EU delayed the AI Act to 2027 is wrong, and it is the kind of wrong an operator acts on. Regulation (EU) 2026/1744, the Digital Omnibus, in force 27 July 2026, deferred only the high-risk obligations - Annex III standalone systems to 2 December 2027 and Annex I embedded systems to 2 August 2028. It did not defer Article 50. The Article 50 transparency obligations applied on 2 August 2026, three weeks before this article was published, and they bind any customer-facing agent serving EU users: a person must be told they are interacting with an AI system unless it is obvious. If your clinic serves visiting international clients, runs locations outside the United States, or uses a platform operating across dozens of countries, that disclosure belongs at the top of the conversation - and it is good practice everywhere regardless.

The thing that makes an AI front desk work at a med spa is almost never the agent. It is whether every inbound channel already lands in one record with documented consent attached. Get that wrong and the agent multiplies the mess; get it right and a fairly ordinary agent performs well, because it is finally operating on clean inputs.

That is the problem our iGlow Med Spa engagement was built to solve, before any of this generation of agents existed. iGlow is a Beverly Hills medical aesthetics practice offering injectables, fillers, facials, laser and body contouring. We built a conversion-focused site on a service architecture that matches how clients actually search - concern first, then treatment - and wired the website chat, the contact and consultation forms, and inbound social messages into a single CRM pipeline, so a lead arriving from Instagram and a lead arriving from a service page become the same contact record rather than two conversations nobody joins up. The forms collect name, phone, email, procedure interest and message, and carry SMS consent language that supports automated messaging. That last detail is the one that matters most for anything you might bolt on later: consent is not a checkbox you retrofit after buying a voice agent, it is the artefact that makes the outbound channel legal at all.

An illustrative cost scenario, worked from published prices only

The following is an illustrative scenario, not a client result, and every input is a price read off a vendor's own page on 23 August 2026. Consider a single-location med spa with two injectors, one aesthetician, an owner and two front desk staff - six users - taking roughly 25 inbound calls a day.

• Option A, built on the most granular price card: Mangomint at $120 base plus five additional users at $10 each is $170 a month, plus one phone line at $70 and marketing from $30 gives $270 a month, before card processing at 2.45% plus 15 cents on card-present transactions. HIPAA enablement and the BAA cost nothing extra but must be requested; if nobody requests them, the practice is running without one.
• Option B, built on the platform with an AI receptionist: Boulevard Premier at the $293 list price, plus the $65 Medspa Add-On that carries HIPAA Coverage, plus the AI Receptionist at $125 for 200 minutes, is $483 a month. Twenty-five calls a day at three minutes each is roughly 1,500 minutes a month - so the 200-minute bundle covers about four days of calls, and the remaining 1,300 minutes at $0.60 would add $780. The honest reading is that the bundle is sized for overflow, not for the main line.
• Option C, built on the aesthetics-native EMR with metered AI: Aesthetic Record at the $19 per-user tier for six users is $114 a month, plus ChartSmart AI unlimited at $75 is $189 a month, plus a one-time $399 onboarding and per-prescriber e-prescribing on top. If the practice charts under 100 encounters a month, the $50 tier drops that to $164. Because the AI is metered by chart rather than by seat, the cost scales with clinical volume, which is the honest shape for a clinic with seasonal demand.
• What the arithmetic actually tells you: The spread between these options is a few hundred dollars a month, which is less than one filler syringe at most published list prices. The decision is therefore not a cost decision. It is a compliance and control decision: which vendor will put a business associate agreement and a data export commitment in your contract, and which one has documented what its AI cannot do. That is the calculation, and no vendor ROI model will run it for you.

Note what we did not do. We did not multiply a no-show rate by an invented per-appointment value, because no med-spa no-show cost figure with a published method exists and every one we found traced to a scheduling or messaging vendor's marketing. If you want that arithmetic, run it from your own list prices and your own open slots, and label it as your own. A number you built from your own books is worth more than an industry figure nobody can source - which, in this sector, is all of them. The same logic drives the scheduling and no-show analysis in our guide to patient scheduling agents, where the refusal to price a no-show from vendor content is the starting point rather than a footnote.

In an aesthetics practice, the first thing that breaks is almost never the model. It is a default setting nobody checked, or a boundary nobody encoded. Each row below pairs a failure mode with the signal that detects it and the rollback that contains it. Detection is the part most deployments skip, and it is the only part that turns a bad week into a small one.

This is the table to put in front of a vendor, because it converts the legal analysis above into a configuration you can check on their platform in an afternoon. If a product cannot enforce the middle column at the platform level - rather than in a prompt - it is not ready for a practice that injects prescription drugs.

The pattern across every row is the same: an agent may read, gather, draft, schedule and propose; a licensed or authorised person commits anything that touches the medical record, the clinical plan, a published claim, or an outbound marketing contact. That boundary is not caution for its own sake. It is the only architecture that survives a prompt injection, a board inquiry and a busy Saturday at once - and it happens to be the one that actually ships, because it never depends on the model being right.

For most single-location med spas, the right first move is not a custom build. It is to execute the business associate agreement you may not have, buy one product with a published price, and measure it against your own baseline for a quarter. Custom work earns its place when the workflow is specific to your practice, spans systems that do not integrate, or carries a compliance requirement no off-the-shelf product will contract to.

Senior-led delivery is $150–$225 per hour and retainers run $2,500–$9,500 per month. Every engagement carries a 30-day post-launch warranty, full source-code and IP ownership transfers to you, and you receive a fixed-price phased proposal within 5 business days of the discovery call. Frenchy Digital is a senior-led, Black-owned Los Angeles agency; you can book a discovery call at calendly.com/frenchydigital/discovery-call or call +1 (424) 272-5601. Those bands are ours and are not an industry benchmark - nobody publishes one for this vertical, which is the recurring theme of this article.

Every item below is checkable during a demo, and most are checkable before one. None of them is disqualifying on its own. Two or three together tell you how recently anyone at the company checked anything.

• HIPAA compliant stated as though it were a certification: There is no such thing as HIPAA certification. It is a self-assertion. The questions that matter are whether a business associate agreement is executed on your account, on what date it took effect, and whether it was in force before the first patient photograph was uploaded. Ask for the executed document and the date; if nobody can produce a date, you do not have one.
• A SOC 2 badge with no type, or a certificate that belongs to someone else: SOC 2 Type I assesses design at a point in time; Type II tests operating effectiveness over a period. Do not upgrade an unspecified mark on the vendor's behalf. And check whose certificate it is - vendors in adjacent verticals have been found citing ISO 27001 and SOC 2 marks on their security pages that belong to their cloud hosting provider rather than to their own legal entity. Confirm the certificate names the vendor.
• A conversion, answer or growth rate quoted without a method: Ask who measured it, against what denominator, over what period, and whether you may see the methodology. In this market the honest answer is always that the vendor measured itself. That is not disqualifying. Presenting it as though a third party had measured it is.
• A comparison table published by a company that appears in it: At least four platforms in this category publish articles ranking the category. If your shortlist came from such a page, you built it from sales decks. Rebuild it from pricing pages, trust pages and developer documentation, all of which are primary.
• A dead product name in the vendor's own competitive deck: If a comparison slide still lists TouchMD or RxPhoto as independent competitors, the deck predates acquisitions from 2022 and earlier. It is a small thing that tells you how recently anyone updated anything.
• A promotional price presented as the price: One vendor's published tiers carry a limited-time, new-customers-only disclaimer, and the discounted figures are the ones every listicle copies. Model with the list price, ask what happens at renewal, and get the renewal uplift in writing.
• A compliance claim about state law with no statute cited: If a vendor tells you a state has banned questionnaire-based good faith exams or that a new statute restricts who may perform them, ask for the bill number and the section. Two of the most-repeated claims in this category are contradicted by the enrolled text, and the sources circulating them sell the services those claims require.
• An AI feature that writes to the chart with no clarity about who controls it: Ask, in writing, who configures the agent, whose instance holds the data, and whether a management company can change its behaviour without the practice's approval. Where private-equity or hedge-fund ownership is in the structure, California has made control of medical record content a statutory question rather than a preference.
• No answer on data residency or retention: Zero of the ten vendors here publish one, so absence is the norm rather than a differentiator - which means it has to move into your contract. Ask where patient images are stored, how long they are retained after termination, in what format they are returned, and what deletion is verified against.
• An ROI model that requires the human out of the loop: If the payback only closes once clinician review or front-desk approval is removed, the model is wrong. The commit step takes seconds. Essentially all of the available saving lives in everything before it, and the review is what makes the rest defensible.

Here is what this article does not know, stated plainly, because a ranking that refuses vendor claims and then hides its own gaps has learned nothing.

• Every compliance, pricing and API cell is a timestamp, not a permanent fact: All of them were read off vendor pages or probed at vendor domains on 23 August 2026. Vendors change pages without notice. Re-check at the URLs in the sources list before you sign, rather than trusting a table you found in an article - including this one.
• Absence of published evidence is not evidence of absence: Several vendors publish no compliance certification on any page we could reach, and several may well hold reports behind a sales NDA. What we can report is what is published. A non-resolving developer subdomain likewise proves only that no public documentation sits at the standard path.
• We could not retrieve the Rhode Island statute text: Rhode Island enacted a Medical Spas Safety Act and its health department was directed to promulgate implementing regulations, but the legislature's server did not respond during our research, and every account we have is a law-firm summary. We therefore state no Rhode Island specifics at all. If you operate there, read the statute and check whether the implementing rules have been finalised.
• A Texas Medical Board rule may have been renumbered and we could not confirm it: The long-standing delegation rule for nonsurgical medical cosmetic procedures is reported to have moved during a 2025 restructuring, and our fetch of the current chapter returned a stub rather than rule text. We therefore cite no section number for it. Pull the current rule from the Texas Secretary of State's administrative code viewer before relying on either numbering.
• We did not open the FTC's health products guidance, so we attribute nothing to it: The general substantiation principle we state - that a claim conveyed by implication still requires substantiation - follows from FTC Act section 5 and the substantiation doctrine, not from a specific passage of that guidance. We deliberately do not write that the guidance says anything specific about before-and-after photographs, because we did not read it.
• The covered-entity analysis is stated without a regulatory section citation: We did not complete a fetch of the federal definitional text for covered entity, health care provider and transaction. The framework in the FAQ is a correct statement of how HIPAA scoping works, but confirm the section-level citation with counsel before making a decision on it.
• A DSCSA enforcement action is described but the letter itself was not retrieved: The first warning letter to a med spa as a dispenser is reported by several trade and legal publications. We did not pull the letter from FDA's warning-letter database, so we name no business and quote no findings. The structural point about purchase-to-chart reconciliation does not depend on the specifics.
• Ownership for two companies rests on an aggregator or was not completed: One vendor's private-equity ownership is recorded in a directory rather than in a filing, and one candidate was excluded because its corporate record was not checked within this review. We label both rather than presenting a directory entry as a filing.
• No employment or wage statistic for this sector appears here: The federal labour statistics site refused every automated request during our research and its public data interface was rate-limited, so no figure from it is verified. We would rather publish nothing than publish a labour statistic from memory, and the Census establishment counts we do use come from the underlying data files rather than from a summary.
• No accuracy, conversion or revenue figure appears here as fact: Not because we did not look. The precise claim is that no independent benchmark of any of these commercial products exists; the only third-party assessments in aesthetics are a paywalled customer-satisfaction award and a single-case review of an imaging system, and neither measures an AI agent.

None of this argues against buying. It argues for buying the way you would buy a laser: confirm what you are permitted to do with it, get the paperwork executed before the first patient, measure your own baseline, run one workflow, and keep the licensed human on the commit step. The clinics that get value from this technology are the ones that found out what they were allowed to do before they found out what the model could do.