Margin leak at one location is annoying. At five it is invisible.
The owner walking the floor catches the wasted product on the counter. The under-stocked retail shelf. The injector who drew up more than they charted. At one location, proximity is the audit.
At five locations, the same leak hides inside a clean-looking P&L for 90 days. The injectable variance gets buried in a quarter-end inventory reconciliation. The refund pattern across sites looks like a normal refund rate. The discount drift compresses margin slowly and shows up on the consolidated P&L three months after it started.
The data exists. The synthesis does not. AI’s job here is synthesis, not reporting. Your manager has the data. Your CFO has the data. Neither has the time to read it weekly the way it needs to be read.
Injectable waste — the largest controllable line in injectables
3 to 7 percent of injectable inventory leaks every month. Overdraw, technique waste, unbilled units, expiration. Quarterly review catches it 90 days late. Monthly review catches it 30 days late. Weekly catches it in time to do something.
The math works because the inputs already exist in your stack. Charted units live in your PMS. Depleted inventory lives in Medvelle. Ticket revenue lives in your point of sale. AI reconciles all three weekly, by provider, and flags the variance the same week it happens.
What gets surfaced: which provider drew more than they billed, which provider’s technique consistently uses more product than the team average, which site has inventory walking out the door. A coachable conversation in week one, not a CFO conversation in month three.
Refund and discount patterns across all locations
Refunds cluster around three things in aesthetics: a specific provider, a mis-sold treatment, or a staff member processing them as a workaround for something else. Your bookkeeper sees the total. The pattern is the cause.
Discount drift is worse because it is slower. Local managers discount to hit numbers. Promotions stack accidentally during overlapping campaigns. Membership pricing drifts. By the time the average ticket compression shows up on the consolidated P&L, the discounting has already been priced into client expectations.
Illume clusters and flags both across every site. Refund spike on Treatment Y during the membership promo. Provider X’s refund rate 4x the team average. Site Z discount rate up six points versus last quarter. The pattern is what tells you what to fix.
LTV that does not depend on front desk skill
Personal-effort retention is the ceiling most practices hit. Your A-team Saturday produces a different LTV trajectory from the same patient than your B-team Tuesday. The follow-up text gets sent on Saturday and forgotten on Tuesday. The loyalty rewards get mentioned by one front desk and skipped by another. The review request happens at the right moment by one staff member and never by the other.
Recura’s AI CRM/EMR layer triggers the right outreach at the right time without front desk intervention. Get Kudos runs loyalty, referrals, memberships, and reviews on autopilot — NPS and CSAT tracked continuously.
The patient gets the same experience whether your A-team is in or not. LTV stops varying by who is on shift.
No-show recovery, scored by lifetime value
Not every no-show is equal. A $2,200 LTV patient missing their botox follow-up is a different problem than a one-time consultation. Most practices treat no-show recovery as a chronological queue: front desk calls them back in the order they showed up on the schedule.
AI ranks recovery effort by LTV and surfaces who to call back first, which patients are pattern (third no-show this quarter), and which are recoverable versus lost. Front desk recovery time moves to the patients worth recovering.
Membership and package liability — the slow leak
Sales pace versus redemption pace. Practices oversell memberships and packages against the rate at which patients actually redeem them. The liability grows for 18 months. Then the redemption wave hits and your CFO calls.
AI runs the liability calculation continuously — unredeemed package value by location, redemption pace vs. sales pace, when the liability will compress margin. The number tells you when to stop selling against tomorrow’s revenue.
The platforms behind the prompts
Four tools doing weekly what your manager and your CFO cannot.
Medvelle AI. Aesthetic-specific inventory intelligence for injectable practices. Waste, expiration, COGS, variance per provider. The data layer that makes the injectable reconciliation possible.
Illume (formerly Illume). Cross-location margin and provider performance. The variance and clustering layer from Carousel 1 of this series, applied here to the margin side of the P&L — refunds, discounts, and pricing drift across every site.
Recura (part of PatientNow). AI CRM/EMR that automates LTV-building outreach without front desk intervention. Same platform from Carousel 2 of this series, doing the retention work, not just the acquisition. This is the case for an all-in-one: one platform doing distinct jobs across the practice. The lead handling from Carousel 2 and the LTV automation here are not separate purchases. They are the same purchase doing more of its job.
Get Kudos. Loyalty, referrals, memberships, and reviews automated. Built specifically for aesthetics. NPS and CSAT tracked continuously. Stops the LTV variance that comes from front desk inconsistency.
Three Claude prompts
Paste-ready. Replace bracketed text with your specifics.
Prompt 1: Injectable reconciliation
You are an aesthetic practice inventory analyst. Below is my charted units per provider for [date range], my depleted inventory per product, and my ticket revenue per provider. Reconcile all three. For each provider, calculate: units charted, units depleted, variance, and dollar impact at our average billing rate per unit. Flag any provider with a variance greater than 5 percent. For each flagged provider, suggest the most likely cause (overdraw, unbilled units, technique waste, charting error) based on the pattern. Do not infer beyond the data. [PASTE CHARTED + DEPLETED + REVENUE DATA]
Prompt 2: Refund and discount clustering
You are a margin analyst for an aesthetic practice. Below is my last 90 days of refunds and discounts across all locations. For refunds, cluster by: provider, treatment, time window (day of week, week of month), and staff member who processed. For discounts, cluster by manager and by location. Surface the three most significant patterns, with the size of each pattern and the most likely operational cause. Output as a one-page summary. [PASTE REFUND + DISCOUNT DATA]
Prompt 3: LTV opportunity analysis
You are a retention analyst for an aesthetic practice. Below is my patient list with last visit date, lifetime spend, last treatment, and any tags. Segment patients into three LTV-recovery buckets: (1) high-LTV lapsed under 6 months (warm), (2) high-LTV lapsed 6 to 18 months (cool), (3) high-LTV lapsed over 18 months (cold). For each bucket, propose the right message angle, the right channel, and the right cadence. Estimate recoverable revenue across the buckets if we ran the campaign. [PASTE PATIENT DATA]
What AI does for your margin
Margin protection at scale is not a discipline problem. It is a synthesis problem. The data exists in your PMS, your inventory system, and your CRM. The synthesis does not. Humans cannot read five P&Ls weekly the way one P&L needs to be read. AI can.
The variance work in Carousel 1 is what AI does for your performance numbers. The lead handling work in Carousel 2 is what AI does for acquisition and retention. The replication work in Carousel 3 is what AI does for your team. This is what AI does for your margin.
Part 8 of The Real Stack covers six dimensions of scaling with AI. This is the fourth.
Read the full series at theaudreyaesthetic.com.