Industries: Supplements and wellness

Customer service for Shopify supplement and wellness stores

Supplement stores get the usual order questions plus questions about dosage, ingredients and whether a product is safe with a medicine. AI and support agents can answer the first kind and quote the label for the second, but they should not give medical advice or make health claims. Fyncall lets you block topics, set rules and send conversations to a person, and you decide where that line sits.

Last updated: 19 November 2026 · By

Fyncall wrote this page. Regulatory points link to the FTC and FDA pages they come from, which were checked 19 November 2026. These are US sources, and this is not legal advice. If you sell supplements, ask a lawyer who knows advertising and food and drug law in the countries you sell to.

What do customers ask supplement stores?

Questions tend to fall into five groups: order status, subscriptions, ingredients, dosage, and whether a product is safe to take with something else. The first three are store questions your team can answer. The last two edge towards health advice, and that changes how you should reply.

Question typeExampleA safe way to answer
Order status “Where is order [order number]?” Look up the order and send the tracking link. No health issues involved.
Subscriptions “Can I skip next month’s delivery?” Explain your subscription policy, then make the change in your subscription tool or pass it to the person who can.
Ingredients “Does this contain gelatin?” Quote the product label or ingredient list. If the label does not say, say so rather than guess.
Dosage “How many should I take?” Repeat the directions printed on the label. For anything beyond the label, suggest the customer asks a health care professional.
Interactions and conditions “Is this OK with my blood pressure tablets?” or “Will this help my anxiety?” Do not answer the health question. Point the customer to a health care professional and hand the conversation to a person.

For ready-made order wording, see the order status reply templates and the WISMO entry.

Why should support agents not give medical advice?

Because a support reply can easily turn into a health claim or a medical recommendation, and neither your team nor an AI is in a position to give one. In the US, both the FTC and the FDA publish guidance on what supplement sellers may claim.

The FDA’s questions and answers on dietary supplements say that the FDA does not have the authority to approve dietary supplements before they are marketed. The FDA’s dietary supplements page says firms are responsible for evaluating the safety and labelling of their products before marketing, and that the FDA can act against adulterated or misbranded products after they reach the market.

The same Q&A page says a product marketed to treat, prevent or cure a disease meets the definition of a drug. It also says some supplements might interact with medicines or other supplements, and it advises consumers to talk to their doctor, pharmacist or other health care professional before using one. That is the reply a customer asking about interactions should get from you as well.

On labels, the FDA’s page on structure/function claims explains that these claims describe the role of a nutrient or ingredient in the structure or function of the body. They must not claim to diagnose, mitigate, treat, cure or prevent a specific disease or class of diseases, and they must carry a disclaimer stating that the claim has not been evaluated by the FDA and that “This product is not intended to diagnose, treat, cure, or prevent any disease.”

What does the FTC say about health claims?

The FTC expects health claims about supplements to be backed by competent and reliable scientific evidence, and it holds marketers responsible for implied claims as well as what they say outright. This comes from the FTC’s Health Products Compliance Guidance, dated December 2022.

Points from that guidance that matter for anyone writing to customers:

  • It judges a claim by what consumers take from it, not by what the marketer meant.
  • It treats advertising broadly, including packaging, labelling, social media and influencer marketing, and it asks marketers to consider the product name along with the words and images.
  • It says anecdotal evidence about individual consumers’ experiences is never sufficient to support a claim about the effects of a health product.
  • It warns that pro forma disclaimers may not cure otherwise deceptive messages, and that the standard supplement label disclaimer will not cure an otherwise deceptive ad.
  • It says it does not have the force or effect of law and does not provide a safe harbour from liability.

The guidance is about advertising, and it does not say how support replies are treated. Still, the cautious reading is to hold support replies to the same standard. A chat reply that says “lots of customers find this helps with sleep” makes the kind of claim, based on customer experience, that the guidance says cannot support itself. Again, this is not legal advice.

How do you answer dosage and ingredient questions safely?

Answer from the label, word for word where you can, and send anything about a person’s health to a health care professional. The label is information you already publish. Personal advice is not.

  • Keep one source of truth. Put label text, directions and ingredient lists in a document your team and the AI both answer from.
  • Separate facts from marketing. If your product pages carry benefit copy, an AI that reads them may repeat it. Keep the documents you give the AI to factual label content and policies.
  • Use a fixed referral line. For example: “I can’t give medical advice. Please check with your doctor or pharmacist before taking [product name] with other medicines.”
  • Do not guess. If the label does not answer it, say you will check, and pass it to the person who owns product information.
  • Report safety concerns. If a customer describes a bad reaction, hand it to a person straight away. The FDA Q&A tells consumers to contact their health care provider and explains how to report a problem through its Safety Reporting Portal, so have a written process for these messages.

Can AI answer customer questions for a supplement store?

Yes, for order, subscription policy and label questions, as long as you limit what it may discuss and send health questions to a person. Fyncall’s AI agents are built around that kind of control. Here is what the documentation describes.

  • Never Discuss. In guardrails you can add blocked topics, with extra topic instructions if you need them, and the agent politely redirects the customer to supported topics when one comes up. A supplement store might block medical advice and questions about specific conditions.
  • Must Follow. Custom rules are added to the agent’s instructions in every conversation, for example “Only quote dosage directions from the product label” or “Recommend a health care professional for questions about medicines or conditions”.
  • Never Say. Words or phrases you list are removed from replies, such as “clinically proven” or “miracle”.
  • Tool access. Each tool can be switched on or off per agent, so an agent can look up orders without being able to cancel them.
  • Knowledge base. The agent answers from documents you link to it, and the knowledge base docs say it can only answer from documents that have finished indexing. PDF, Word, text and Markdown files up to 10 MB are accepted.

Guardrails reduce the chance of a bad reply. They do not make a store compliant, and a filtered word list cannot catch every implied claim. Test before you go live and keep reviewing real conversations.

When should a supplement conversation go to a person?

Whenever a customer mentions a medicine, a condition, pregnancy, a reaction or a complaint, and whenever they ask for a person. Fyncall’s escalation settings cover several of these.

  • Phrases such as “talk to a human” trigger an immediate handover.
  • Sentiment triggers escalation: frustrated and angry customers at set thresholds, threatening messages immediately.
  • Escalation also happens when the AI’s confidence falls below 70%, or when the same question is asked three times.
  • Billing disputes, legal questions, complaints and large orders always escalate.
  • Settings can be overridden per agent. The docs’ example has one agent set to “Always escalate pricing questions”.

Health questions are not among the topics the docs list as always escalating, so do not assume they will reach a person. A blocked topic is redirected rather than handed over. The docs do not say exactly how a per-agent override for health questions would behave alongside a blocked topic, so set up what you need and test it in the sandbox with real interaction questions before customers see it. Customers can always ask for a person.

On handover, the docs say the AI prepares a summary, the customer’s details and what was already tried, the team is notified, and the AI pauses its responses once a person takes over. More on this in human handoff and how AI handoff works on Shopify.

How do you handle subscription questions?

Answer the policy questions automatically, and leave the account changes to your team. Fyncall does not connect to subscription apps, so it cannot skip, pause or cancel a subscription.

Upload your subscription terms (billing dates, how to skip, how to cancel) to the knowledge base, so the AI can explain them. When a customer wants a change made, it should hand over, and a person makes the change in your subscription tool. For one-off orders, Fyncall’s AI support agent can look up an order and give a tracking link, update the shipping address or cancel an order that has not shipped, and it asks the customer to confirm before changing an order.

How do you roll out AI on a supplement store?

Start with your team approving every AI reply, test it on your hardest health questions, and only then let it answer on its own.

  1. Upload label text, ingredient lists, directions and your shipping, returns and subscription policies.
  2. Add blocked topics, Must Follow rules and Never Say words for health claims and medical advice.
  3. Configure escalation for health and interaction questions, and check in testing that they reach a person.
  4. Use the test sandbox described in the AI agents overview to try real questions, including ones it should refuse. Testing does not reach customers.
  5. Turn the agent on in Draft & Review, where your team approves each reply. Move a channel to Auto-Reply only when you trust what you see.

You can try this on the Free plan, which includes 25 AI messages a month and up to 2 team members. Pro is $34 a month with 500 AI messages and unlimited team members. See pricing for details.

Common questions

Can my support team tell customers how much of a supplement to take?

They can repeat the directions on your label. Advice for a particular person, or about medicines and conditions, belongs with a health care professional, which is what the FDA advises consumers to consult.

Does adding a disclaimer to replies make health claims safe?

Not on its own. The FTC guidance says a pro forma disclaimer may not cure a message that is otherwise misleading. It is safer not to make the claim. This is not legal advice.

Will Fyncall’s AI hand over health questions automatically?

Not by default. Health questions are not on the documented list of topics that always escalate, and a blocked topic is politely redirected. Configure escalation for them, test it, and remember that a customer asking for a person triggers a handover.

Can Fyncall manage subscriptions?

No. Fyncall connects to Shopify, WhatsApp, Instagram, email and its web chat widget, not to subscription apps. The AI can explain your subscription policy and hand changes to your team.

Which channels does Fyncall cover?

WhatsApp, Instagram, email and web chat in one inbox. Each channel can have its own agent with its own guardrails. See the WhatsApp and Instagram pages, or the beauty page for a related category.

Ready to try Fyncall?

Install it tonight, let the AI draft your first replies, and switch it on one channel at a time as you start to trust it.

  • Free plan, no credit card
  • 14-day free trial on Pro
  • Unlimited human replies on every plan