support triage
Refund request review
Route refund requests through policy checks, order context, and human approval.
Outcome
The reply is policy-aware, order-aware, and approved before sending.
Recipe snapshot
Ecommerce support teams45-90 minutesmedium difficultyhigh riskGorgiasZendesk AIShopify Magic
Trigger: A customer asks for a refund or cancellation.
Inputs
- Trigger source
- Customer or record context
- Owner
- Deadline
- System of record
Setup steps
- Capture the trigger and required context.
- Create or update the system-of-record entry.
- Draft the next action or message for review.
- Assign an owner and deadline.
- Log the result and check failures in the next daily review.
Prompts to reuse
- Summarise the context for refund request review in five bullet points.
- List missing information, risk flags, and the safest next action.
- Draft a short message that a human can review before sending.
Review checks
- Confirm the source data is real.
- Check customer-facing wording before sending.
- Confirm the owner and deadline are visible.
Failure modes
- Trigger does not fire.
- Record is duplicated or routed to the wrong owner.
- AI output sounds confident but lacks source evidence.
Privacy notes
- Only send fields needed for the workflow.
- Avoid uploading sensitive data unless the tool has been approved.
- Keep a human review step for high-risk outputs.
Related guides
GorgiasCustomer support platform for ecommerce stores with automation and AI support workflows.Zendesk AIAI features and agents inside Zendesk for customer service and support operations.Shopify MagicAI-powered Shopify features for product descriptions, store content, images, and ecommerce workflows.Support triage and reply draftA customer support message arrives.Run the related free toolTurn this recipe into a practical output.Workflow LibraryBrowse more workflow recipes.
Get the daily AI workflow brief
One practical workflow idea, tool check, prompt, or calculator to use each day.