YundaDeskRefund declines fail fast. A blunt “not eligible” can trigger reviews, disputes, or chargebacks. A rushed refund promise can break policy and make every similar case harder later.
A better boundary is simple: AI explains policy and prepares context; refunds, compensation, and price changes always go through human approval.
Refund-Decline Scripts That Explain Policy Without Escalating: the experience cost high-risk handling cannot ignore
Set the Boundary: AI Explains, Humans Approve
Refund scripts should start with permission rules.
| Scenario | AI can do | Must go to a human |
|---|---|---|
| Customer asks about refund policy | Explain windows, conditions, and required proof | Customer asks for an exception |
| Request clearly falls outside policy | Explain why and offer next options | Customer is upset or threatens a complaint |
| Money-related action | Collect order number, photos, and item condition | Refund, compensation, price change approval |
Use a Script Structure: Acknowledge, Then Explain
A good refund decline script usually follows four steps:
- Acknowledge the concern so the customer does not feel ignored.
- State what was checked such as order date, item condition, or policy window.
- Explain why the order is not directly eligible without blaming the customer.
- Offer a next step such as exchange review, troubleshooting, store credit, or human review.
Use this base script:
I understand you want this order resolved quickly. We checked the order details and our return policy, and this order is not eligible for a direct refund because of [specific reason].
We cannot promise a refund without approval, but we can help with [option 1] or [option 2]. If you believe there is a special circumstance, I can submit the order for human review.
Use “not eligible for a direct refund” instead of “we cannot refund you.” It leaves room for review.
Late Refund Requests: Show the Timeline
For late requests, show the dates.
| Customer message | Suggested reply |
|---|---|
| I am only a few days late. | I understand why that feels frustrating. The order was delivered on [date]. Our refund request window is [number] days from delivery, and this request is now outside that window. Under the current policy, it is not eligible for a direct refund. |
| I did not have time earlier. | I understand you were not able to handle it sooner. The refund window is calculated from the delivery date so every order is reviewed consistently. This order is now outside that window, but we can check whether exchange, repair, or human review is available. |
| Other customers got refunds. | We review each order against the same policy. For this order, the delivery date and request date fall outside the direct refund conditions. If there is additional context we should consider, please share it and I can submit the case for human review. |
Non-Quality Issues: Separate Preference From Defect
Non-quality requests can turn into taste debates. Keep the reply focused on defect, preference, and policy.
I understand the item did not fully meet your expectations. Based on the information provided so far, we have not found a product defect, so this does not fall under a quality-related refund reason.
If the item is unused and the packaging and accessories are complete, we can still check whether it qualifies for return or exchange under the store policy. If it does not meet those conditions, we can offer [alternative option] for you to consider.
For cross-border e-commerce, explain return cost early:
Because this is a non-quality-related request, return shipping is usually paid by the buyer, and the warehouse needs to inspect the returned item before the next step is confirmed. You can send photos of the item condition first, and we will help you assess whether it is worth continuing the request.
Used Items: Describe Condition, Not Character
For used items, describe the verifiable condition instead of sounding accusatory.
| Item condition | Softer decline wording |
|---|---|
| Visible signs of use | From the photos provided, the item shows visible signs of use. Under the return policy, items that may affect resale condition are not eligible for a direct refund. |
| Missing packaging | The packaging and accessories appear incomplete, so the warehouse may not be able to accept it as a returnable item. Please send photos of the missing parts, and we can submit the case for human review if another option may apply. |
| Hygiene or safety-sensitive item | This item category has hygiene and safety requirements. Once opened or used, it is usually not eligible for a non-quality refund. We can still check whether there is any quality issue. |
A useful buffer sentence is:
We are not dismissing your experience. Refund approval has to be based on the current item condition and the policy requirements.
When Emotion Escalates: Stop Explaining and Hand Off
More explanation is not always better. If the customer threatens a bad review or chargeback, hand off.
Add these signals to your handoff rules:
- Refund, compensation, chargeback, complaint, lawyer, media, bad review
- Customer rejects the AI explanation twice
- Customer asks for a manager or human agent
- The case involves platform disputes, payment disputes, or customs holds
Use this handoff script:
I understand this is important to you. To avoid slowing this down with a template-only reply, I will submit the order details, your explanation, and the policy check result for human review. A support teammate will review the full context and decide the next step.
If you run an AI answers first, humans back up workflow, AI should collect evidence, summarize the case, and flag the risk here. It should not keep trying to persuade the customer.
Feed the Knowledge Base: Make Scripts More Accurate
Refund decline scripts should not be a one-time document. New edge cases belong in the knowledge base:
- Refund policy: time windows, item condition, category limits, return shipping rules
- Standard scripts: late request, non-quality issue, used item, missing packaging, restricted category
- Handoff rules: high-risk keywords, customer requests, exception approval conditions
- Learning suggestions: cases where AI could not answer, agents filled the gap, or agents corrected AI
One point matters: learning suggestions must be confirmed by the owner or support lead before they take effect. Refund scripts involve money and disputes. AI should not automatically learn a new refund stance from one conversation. Each learning item should be traceable, testable, and revertible.
Appendix: Refund decline scripts you can adapt
Late request
The order was delivered on [date], and the request is outside the [number]-day refund window. It is not eligible for a direct refund, but we can check exchange, repair, or human review.
Non-quality issue
We have not found a product defect, so this does not fall under a quality-related refund reason. If the item is unused and complete, we can check return or exchange options.
Used item
The photos show signs of use. Items that may affect resale condition are not eligible for a direct refund, but we can check for quality issues or submit human review.
A refund decline should not feel like pushing the customer away. It should make the policy, evidence, and approval boundary clear. Let AI handle repeated explanations, keep sensitive decisions with humans, and feed approved scripts back into the knowledge base.