Shoppers eyeing an LED mask or a microcurrent device want to know if it's safe for their skin, how long results take and whether it beats a course of facials. When one of those checkouts is abandoned, a Loop Voice agent calls and answers from your manual, labeling and approved claims.
The short answer
For a beauty device brand, an AI voice agent calls shoppers who left a device in checkout and answers safety, results and warranty questions from your manual, labeling and substantiated claims. It states FDA status exactly as printed and sends questions about a skin condition or medication to the shopper's doctor or dermatologist. Loop Voice results for beauty devices aren't published yet; until they are, the call notes show whether safety or price is what holds orders back.
Sensitive skin, rosacea, melasma, a pacemaker, a pregnancy, a prescription retinoid: shoppers want to know if the device is safe for them before they ask what it does. The answers sit in your manual's warnings, which a shopper on a product page rarely reaches.
FDA's own consumer page notes that some marketers say their products are "FDA approved", and that FDA's logo is for official government use only. Clearance through a 510(k) isn't approval, and a shopper who asks deserves to hear exactly which status your device has.
LED and microcurrent routines are measured in weeks of regular sessions, so shoppers ask whether they'll see a change before the return window closes. The safe answer is your published routine and timeline.
Shoppers weigh the device against a series of LED or microcurrent facials. They want the cost and convenience gap explained, and anything said about results must stay inside what your evidence supports.
Add a cart value branch to the abandoned checkout flow, so a mask or microcurrent checkout gets the call and a cleanser-only cart gets texts alone. On the call, brief the agent to take the safety question first, then results, then the price against a course of facials.
See the use caseStart the flow from order fulfilled and wait about two weeks plus shipping, so the call lands while the return window is still open. The agent asks whether the device charges, whether the customer follows the manual's routine and whether anything irritated their skin. Irritation becomes a ticket, never a sales opening.
See the use caseMicrocurrent devices run on conductive gel and sonic brushes take replacement heads. A campaign, still in beta, can call device buyers in a Shopify or Klaviyo segment who haven't reordered, and text the refill link on a yes.
See the use caseAn inbound agent answers "it won't turn on" and "how often should I use it?" from your manuals at any hour. Warranty claims become tickets with the call summary, order and recording, and with Gorgias connected, every agent escalation from calls and texts creates a Gorgias ticket.
See the use caseAI agent
Hi Monica, it's Sam, an AI assistant calling from [Store]. If you'd rather not get these calls, just say stop. I saw the LED mask is still in your checkout. Can I answer anything?
Shopper
I have rosacea, and I'm nervous about lights on my face. Is it even FDA approved?
AI agent
It's FDA-cleared, which is what our box says, and cleared isn't the same as approved. On rosacea, our manual asks anyone with a skin condition to check with their dermatologist before starting.
Shopper
Fair. How long before I'd see anything?
AI agent
Our product page sets the routine at ten minutes a session, five days a week, and says results vary. I can't promise what you'll see.
Shopper
Okay. I'll ask my dermatologist first.
AI agent
Good idea. The safety section of the manual is on the product page to show them, and the text after this call has a link back to your checkout.
Shopper
Thanks, that helps.
What shoppers say
“I have sensitive skin. Will this irritate it?”
How the agent answers
It reads the warnings and any patch-test advice in your manual and doesn't predict how anyone's skin will react. A named condition such as rosacea or eczema gets a referral to the shopper's dermatologist.
What shoppers say
“Is it FDA-approved?”
How the agent answers
It reads the status on your labeling word for word. If the shopper asks whether cleared is basically approved, it explains that clearance means FDA found the device substantially equivalent to one already legally on the market, and that approval is a different process.
What shoppers say
“How soon will I see results?”
How the agent answers
It quotes your published routine and the timeline your claims support, then says results vary. It never turns a study average into a personal promise.
What shoppers say
“Why not just book facials?”
How the agent answers
It compares what you publish: the device's price against a series of treatments, and sessions at home against appointments. It claims clinic-level results only if your evidence supports that.
What shoppers say
“Can I use it after fillers, or with my retinol?”
How the agent answers
Only if your manual covers that exact case. Otherwise the agent says it can't advise, suggests asking the provider who treats them and offers a ticket for your team.
Build the knowledge base first: each device's manual and labeling text with its exact FDA status, your warranty terms, the claims your team has approved, and the product and FAQ pages. The readiness score shows what's still missing.
In the AI prompt assistant, list under things to avoid any wording your labeling doesn't support, such as "FDA-approved" for a cleared device, "cures" or a promised results date.
Pick the strict guardrail template, check that never hallucinate and respect refusals are on, and point the agent's Escalation Path at the queue that handles safety reports and warranty claims.
Connect your store in Settings, Integrations, then edit the abandoned checkout flow: set the cart value branch, decide whether the second call carries one of the codes you allow, and put your safety page link in the SMS step.
Run up to ten simulated conversations at once with the skeptical shopper and suspicious of AI personas, plus a customer you write: someone with rosacea asking if the mask is FDA-approved. Check each review's compliance score.
Before the first consumables campaign, confirm that everyone in the segment agreed to calls. Proving consent stays with your store, and the calling rules guide covers what counts.
It can match a concern the shopper describes, such as dryness or breakouts, to the device your buying guide recommends, and read that device's warnings. It doesn't diagnose, and a named condition or medication sends the shopper to their doctor or dermatologist first.
Keep to ones that reflect what users can generally expect. FTC guidance says testimonials reporting more dramatic results are likely to be deceptive, and a "results not typical" line doesn't cure that; it calls for a clear statement of what a typical user can expect. Load only testimonials your team has cleared.
Build a segment of device buyers who haven't reordered gel, serum or brush heads within your replacement interval, then run it as an outbound campaign with its own calling window and pace. Customers who say they stopped using it tell you more than a reorder would; send those to your team.
What you write into a custom guardrail, so write one before launch. A sound one drops any offer, repeats what your manual tells users to do and escalates the call, which opens a ticket with the AI summary, the order and the recording. Test it in the simulator with a customer who reports a burn.
Count how often the QA notes on your calls list safety as the objection, against price. Track irritation tickets per 100 answered calls and orders attributed to called checkouts, and run a holdout test to see what calls add on top of email and SMS.
Install Loop Voice and start generating revenue in under 10 minutes.
