Most small-store inboxes are 60-70% repetitive ticket types, but the remaining 30% carries most of the refund and churn risk. Shopify support automation works when you automate the repeatable majority, route risky tickets through approval, and keep a short list human-only. This guide gives you the rules, a decision table, and a rollout plan.
Key takeaways
- 60-70% of tickets (WISMO, FAQ, early address changes) are safe to auto-resolve.
- Refunds, out-of-policy returns and damaged items fit draft-and-approve.
- Chargebacks, legal threats and VIP complaints stay human-only.
- Set approval thresholds by order value, then audit weekly for four weeks.
Which support tickets should you automate first?
Automate tickets where the answer is factual, the data lives in Shopify, and a wrong reply costs little. Typically that means WISMO (where is my order), policy FAQs, and address changes before fulfillment. Together these often make up 60-70% of a small store's inbox, so they deliver most of the time savings.
The test is simple. Ask three questions of any ticket type:
- Is the correct answer in your order data or knowledge base?
- Can a wrong answer be fixed with one follow-up email?
- Does it involve money leaving your account?
If the answers are yes, yes, no, automate it. This is the core of any sound set of customer support automation rules: risk, not volume, decides the mode.
Which ticket types are safe to fully automate?
Three ticket types are safe to auto-resolve: WISMO, FAQ, and address changes made before fulfillment. Each has a verifiable answer in Shopify, and a mistake is cheap to correct. A WISMO reply that quotes the wrong carrier status costs one extra email, not a refund.
- WISMO: look up order status and tracking, reply with the current state and ETA.
- FAQ: shipping times, sizing, care instructions, and return windows, answered from the knowledge base.
- Address change before fulfillment: check the order is unfulfilled, update the address, confirm.
The guardrail is the fulfillment check. Once an order has shipped, an address change becomes a carrier problem, so it should drop to draft-and-approve.
With Heylio, each of these maps to a playbook (FAQ, WISMO) that uses the Shopify order lookup and address change tools. Auto-resolved tickets cost €0.49 each, or €0.39 with a monthly pack.
Which tickets should be AI-drafted with human approval?
Use draft-and-approve for any ticket that moves money or bends policy: refunds, out-of-policy returns, and damaged items. The AI does the reading, the order lookup, and the writing. A human spends about ten seconds approving. You keep speed without handing over the checkbook.
Typical cases:
- Standard refunds: the AI checks the order, drafts the reply, and proposes the refund.
- Out-of-policy returns: the request is 35 days in on a 30-day window. A human decides whether goodwill is worth it.
- Damaged items: the AI asks for photos, drafts a replacement or refund offer, and waits.
This is a shopify ai support approval workflow in practice. In Heylio, an approved draft costs €0.10, which is less than the cost of writing the reply yourself. Edits you make are tracked, and Heylio suggests knowledge base and playbook updates from those corrections.
Which tickets should stay human-only?
Keep chargebacks, legal threats, and VIP complaints human-only. These tickets are rare, often under 5% of volume, but a single mishandled one can cost an order value, a customer for life, or a legal bill. Tone and judgment matter more than speed here.
- Chargebacks: evidence submission is time-sensitive and account-affecting.
- Legal threats: any mention of lawyers, consumer authorities, or regulators.
- VIP complaints: repeat buyers, wholesale accounts, and press or influencer contacts.
Good ai customer support human handoff means the AI recognizes these, stops, and passes the full thread with context. In Heylio, handing over to a human is never billed, so there is no financial reason to let the AI push through a ticket it should not touch.
What is the decision table for each ticket type?
The table below assigns a risk level, a recommended mode, and the Shopify action involved to each common ticket type. Use it as a starting default, then adjust it to your margins and policies.
| Ticket type | Risk level | Recommended mode | Shopify action involved |
|---|---|---|---|
| WISMO | Low | Auto-resolve | Look up order status and tracking |
| Policy FAQ | Low | Auto-resolve | None (knowledge base only) |
| Address change, unfulfilled | Low | Auto-resolve | Change address |
| Address change, shipped | Medium | Draft and approve | None until carrier confirms |
| Order cancellation, unfulfilled | Medium | Draft and approve | Cancel order |
| Standard refund | Medium | Draft and approve | Refund |
| Out-of-policy return | Medium-high | Draft and approve | Refund or discount |
| Damaged item | Medium-high | Draft and approve | Refund, replacement, gift card |
| Discount or goodwill request | Medium | Draft and approve | Apply discount |
| Chargeback | High | Human only | None (evidence handled manually) |
| Legal threat | High | Human only | None |
| VIP complaint | High | Human only | Any, manually |
How do you set approval thresholds by order value?
Set thresholds by order value so that the money at risk decides the level of oversight. A common starting split is: under €30 auto-approve simple cases, €30-150 draft and approve, above €150 human review. These numbers are examples, so tune them to your average order value.
A practical method:
- Find your average order value (AOV).
- Set the auto tier near 0.5x AOV.
- Set human-only review near 2x AOV.
- Everything between goes to draft and approve.
A store with a €60 AOV might auto-handle under €30 and require human review above €120. Review the thresholds after 30 days of data. If you approve 98% of drafts untouched in a tier, loosen it. If you edit 30%, tighten it or fix the playbook.
Where does Heylio fit in this setup?
Heylio is the layer that applies these rules inside your inbox. It connects to Gmail, Outlook, Instagram DMs, WhatsApp, and an embedded chat widget. It drafts replies from your knowledge base and playbooks for FAQ, WISMO, refunds, returns, and cancellations. It can also act in Shopify.
Citation capsule: Heylio prices support by outcome. Under Heylio Care, a ticket resolved by the AI costs €0.49, or €0.39 with a monthly pack. A draft a human approves costs €0.10. Handing a ticket to a human is never billed. Prices are in euros excluding VAT, with pay as you go or monthly packs that do not roll over, and there is no free trial. In practice, that means risky ticket types cost little to route through approval, and tickets the AI should not touch cost nothing to escalate. Shopify actions available to the AI include order lookup, cancellations, refunds, discounts, address changes, and gift cards. You choose which actions each playbook may take.
Honest caveat: Heylio does not remove the need for judgment. You still define the playbooks and thresholds. It makes those rules enforceable.
What does a 4-week rollout plan look like?
Roll out in four weeks, adding one risk tier per week, so you catch errors while volume is small. Starting with the lowest-risk tickets typically gives you a working automation layer by week two and a full approval workflow by week four.
- Week 1: Audit and write. Export 100 recent tickets, label them by type, and update your knowledge base. Fill gaps in return and shipping policies.
- Week 2: Automate the safe tier. Turn on WISMO, FAQ, and early address changes. Read every reply for the first three days.
- Week 3: Add draft and approve. Enable refunds, returns, and damaged items in approval mode. Set value thresholds.
- Week 4: Lock handoff and tune. Define human-only triggers (chargeback, legal, VIP). Review corrections and update playbooks.
After week four, audit weekly for a month, then monthly.
Start with the boring 60%
The safest first move is also the most profitable: automate WISMO, FAQ, and pre-fulfillment address changes, then put refunds behind approval and keep chargebacks, legal threats, and VIPs human. Label your last 100 tickets against the decision table this week, and you will see exactly where your 60-70% sits.