YundaDeskDamaged-item support goes wrong when the first reply tries to apologize, assign fault, and promise compensation all at once.
The customer is usually stressed. The item may be a gift, a sale-period order, or something they needed this week. But for cross-border e-commerce teams, a damaged item is not just an emotional reply. You need photos, packaging evidence, order context, carrier clues, and a clear approval path for replacement, refund, or discount compensation.
The better workflow is simple: AI support catches the first message, calms the customer, and guides photo upload. Once evidence is collected, AI summarizes the case and hands off any money-related action to a human agent. These damaged item reply scripts are written for both agents and the knowledge base.
Calm the customer first: do not assign fault too early
The first message should not blame the carrier, the warehouse, or the customer. It should acknowledge the problem and move the case into a review flow.
I am sorry the item arrived damaged. To help us review this quickly, please send your order number and a few photos. I will organize the information first. Any replacement, refund, or compensation will be confirmed by a human agent.
Got it. I understand this may affect your plans. Please keep the item and the outer packaging for now, because the team needs photos of the damage, packaging condition, and shipping label.
I will help record the case clearly. To avoid a wrong judgment, please provide photos of the item, outer package, inner protection, and shipping label. Once the evidence is complete, I will hand it to our team for review.
The goal is to catch the case, not judge it. AI support is useful for de-escalation and information collection, but it should not promise “we will replace it right away” or “you will definitely get a refund.”
Photo checklist: help customers send the right evidence once
The slowest part of damaged-item support is often not the final decision. It is asking for missing photos again and again. The first reply should make the evidence checklist clear.
| Photo needed | How to take it | Why it matters |
|---|---|---|
| Full item photo | Show the complete item in good light | Confirms the damage scope |
| Damage close-up | Focus on cracks, dents, breaks, leaks, or stains | Shows whether use is affected |
| Six sides of the outer package | Include dents, holes, wet areas, or compression marks | Checks shipping or delivery damage |
| Inner protection | Show fillers, bubble wrap, dividers, or inserts | Reviews packaging quality |
| Shipping label | Keep order number or tracking number visible; private details can be covered | Matches the case to the order |
To help the team review this in one pass, please upload five types of photos: full item, damage close-up, all sides of the outer package, inner protection, and shipping label. You may cover private information, but please keep the order number or tracking number visible.
Fault language: say “under review”, not “we know why”
A common support mistake is deciding too quickly: “the carrier damaged it”, “the warehouse packed it poorly”, or “this happened after use.” Once that message is sent, it is hard to walk back.
A better damaged item reply keeps the review open:
We will review the item photos, package condition, shipping record, and after-sales policy together. I do not want to make a rushed judgment before all information is available. Once the evidence is complete, the team will confirm whether replacement, refund, discount compensation, or further review applies.
From the photo you shared, the item does show visible damage. We still need the outer package and shipping label photos before confirming the next step.
If the outer package has compression marks, holes, moisture, or other visible damage, please include those photos as well. This helps the team understand where the damage may have happened.
AI can classify the case as “possible shipping damage”, “possible arrival damage”, or “missing evidence.” It should not lock in fault automatically, and it should never execute refund or compensation by itself.
Damaged-item speed depends on whether the first round collects complete evidence, not on how fast support guesses the cause. Design the flow as evidence collection, missing-evidence follow-up, and human approval.
Damaged-item evidence flow (illustrative)
Fix options: replacement, refund, or discount
Damaged-item fixes usually fall into three buckets: replacement, refund, or keep-with-discount. Different product categories need different handling, so scripts should leave the final choice inside the review process.
| Option | Typical use case | Recommended script |
|---|---|---|
| Replacement | Item cannot be used, and the customer still wants the product | Once the evidence is complete, we will first check whether a replacement can be arranged. Replacement depends on stock, destination, and after-sales policy, and needs human confirmation. |
| Refund | Item cannot be used and replacement is not suitable | Refunds require human review based on order status, damage evidence, and after-sales policy. I will send the full context to the team. |
| Keep with discount | Minor damage that does not affect main use | If you are willing to keep the item, we can record a discount compensation request. The specific amount needs human approval. |
| Further review | Missing photos, unclear fault, or unusual amount | We do not have enough information to confirm a fix yet. Please send the missing photos or details, and the team will continue the review. |
Ready-to-use damaged-item scripts
These templates can be stored in your knowledge base. For broader support templates, maintain them alongside your cross-border support scripts.
First damaged-item report
I am sorry the item arrived damaged. I will help record the case and send it to the team for review. Please send your order number, a full item photo, a close-up of the damage, outer package photos, and the shipping label. Any replacement, refund, or compensation will be confirmed by a human agent.
Customer only sent one photo
Thanks, I received this photo and can see part of the damage. To avoid a wrong review, we also need photos of the outer package, inner protection, and shipping label. You can cover private information, but please keep the order number or tracking number visible.
Customer asks for an immediate refund
I understand you want this resolved quickly. Refunds need human review based on the order, damage photos, and after-sales policy, so I cannot promise or execute a refund directly. I will send your request and the available evidence to the team.
Customer wants to keep the item with compensation
If you are willing to keep the item, we can record a discount compensation request. The compensation method and amount need human approval. Please first send a damage close-up and a full item photo so the team can review the impact.
Use AI for the first half: collect, summarize, hand off
Damaged-item cases are not a reason to avoid AI. They are a reason to set the boundary clearly. The right split is:
- AI answers first: de-escalates, explains what photos are needed, and reminds the customer to keep the packaging
- AI summarizes: order number, channel, product, damage type, photo completeness, and customer request
- AI hands off: refund, replacement, compensation, and discount amount requests go to a human with context
- Human approves: responsibility, fix option, and any money-related action
- Team improves: agent corrections become learning suggestions for review
This is where “gets smarter over time” matters. Missed answers and agent corrections become learning suggestions only. They take effect after owner approval, and they should remain traceable, testable, and revertible. For the full loop, see how to teach AI support that gets smarter.
Review: turn damaged-item cases into a better process
Review damaged-item cases weekly. The point is not to blame agents. It is to find weak spots in packaging, policy, scripts, and AI boundaries.
- Which SKUs have the most damage reports? Should packaging be changed or the product checked?
- Which channels send the least complete evidence? Is the first reply unclear?
- Which compensation requests are often rejected? Are approval rules too vague?
- Where does AI risk overpromising? Should the handoff rule be stricter?
- Which agent replies worked well? Should they become knowledge base entries?
Appendix: damaged-item knowledge base entry
Title
How to reply when an item arrives damaged
Customer phrasing
damaged item reply / item arrived broken / package was crushed / product is leaking / item damaged in transit
AI can handle directly
De-escalate, ask for order number and five photo types, remind the customer to keep the item and packaging, and summarize the case.
Must hand off
Customer asks for refund, replacement, compensation, discount amount, platform complaint handling, threatens a bad review, or reports severe damage to a high-value item.
Learning rule
New SKU packaging rules, photo requirements, or approval language from agents become learning suggestions only. They enter the knowledge base after owner confirmation.
Handled well, a damaged-item case tells the customer that your team has a process. Handled poorly, it feels like you are asking them to repeat the same evidence forever. Let AI handle the first half, keep compensation decisions under human approval, and fold each case back into the knowledge base. That is how damaged-item support becomes an operating capability instead of a daily fire drill.