YundaDeskWhen a customer says “the item arrived broken”, “my package is lost”, or “one piece is missing”, support teams usually make one of two mistakes: promise compensation too early, or repeat vague requests like “please send photos” until the customer has to explain everything again.
Compensation claims are high-risk conversations. YundaDesk’s boundary is clear: AI can greet the customer, collect evidence, verify the order and summarize the case, but refunds, compensation and price changes must always go through human approval. This compensation claim script keeps evidence, responsibility and resolution clear for human review.
Separate the issue first: damaged, lost or missing
Different claims need different evidence. Do not use one generic script for every case.
| Scenario | Common customer wording | Collect first |
|---|---|---|
| Damaged item | Arrived broken, packaging crushed, accessory snapped | Product photos, outer packaging photos, shipping label, unboxing timing |
| Lost package | Tracking says delivered but nothing arrived, tracking is stuck | Order number, address confirmation, tracking screenshot, nearby pickup checks |
| Missing item | Bundle missing one piece, accessory not included, wrong quantity | Photo of received items, packing slip, SKU/color/size |
Do not open with a compensation amount. Use a steadier first message:
Sorry your order arrived this way. To check whether this is shipping damage, a warehouse miss or a carrier issue, I will first verify the order and collect the required evidence. Once confirmed, we will send the case for human approval and review the right option: replacement, refund or another resolution.
That message acknowledges the issue, explains process and avoids early promises.
Step one: collect evidence with clear instructions
Many customers are willing to cooperate. They just do not know what kind of photo is useful, so tell them exactly what to send.
Damaged item script
Please send us three photos: 1) a close-up of the damaged area; 2) a full photo of the outer packaging; 3) a photo of the package with the shipping label visible. We just need to clearly see the damage and label details so we can check the order and carrier record.
Missing item script
I will help you check this. Please take one photo showing all items you received, then tell us which style/color/size is missing. If there is a packing slip, please include a photo of that too.
Lost package script
I can see the tracking status needs further checking. Please first confirm that the delivery address is correct, then check your door, front desk, parcel locker or neighbor pickup area. If you still cannot find it, please send us a screenshot of the tracking page and we will continue the investigation.
From claim intake to approval review (illustrative)
Step two: verify the order, not just the complaint
Once the evidence arrives, agents need the full order context. Check four things:
- Whether the order exists, and whether the recipient, email and social IDs match the same customer record.
- Whether the SKU, quantity, batch and shipping warehouse match what the customer describes.
- Whether the tracking status shows abnormalities, such as no update for a long time, mismatched delivery location or return in progress.
- Whether the customer history shows repeated claims, unusual complaints or a high-value order.
Handling this inside an Omnichannel inbox is steadier than jumping between email, WhatsApp and TikTok DMs. Messages from every channel land in one workspace, and AI places the order, customer record and conversation summary side by side before human review.
Step three: assign responsibility with human judgment
Claim decisions are not as simple as “the customer sent a photo, so pay.” A practical review path:
| Review point | Likely handling |
|---|---|
| Item is visibly damaged and packaging is crushed | Treat as likely shipping damage; review replacement or compensation |
| Item is intact but an accessory is missing | Check warehouse fulfillment and SKU configuration first |
| Tracking says delivered but customer did not receive it | Verify address, delivery proof and pickup locations before replacement or carrier claim |
| Customer cannot provide key evidence | Ask for the missing evidence, then let a human decide any exception |
The boundary matters: AI may suggest a handling option, but the suggestion must enter a pending approval state. An authorized owner confirms before replacement, refund, coupon or other compensation is executed. This is the same principle behind AI answers first, humans back up: repetitive context goes to AI, high-risk decisions stay with people.
Step four: offer a resolution with precise wording
Resolution scripts should make the customer feel the case was reviewed and make the approval result clear.
Replacement script
We have reviewed the order and evidence, and this case qualifies for a replacement of the missing/damaged item. The replacement order will be created after approval, and we will share tracking through your original contact channel.
Partial refund script
We have completed the review. Based on the damage level and the order details, we can submit a partial refund request for approval. Once approved, the refund will be processed back to the original payment method. We will notify you as soon as it is confirmed.
Insufficient evidence script
I understand this has been frustrating. We reviewed the order, tracking record and evidence so far, but we cannot yet confirm product damage or a missing item. To continue, please send the following details: … Once received, we will submit the case for another human review.
Do not turn “not approved yet” into a cold rejection. Say what is missing and what happens next.
Use one evidence checklist to reduce back-and-forth
Put this checklist into your knowledge base so AI can call it when it detects a compensation claim intent.
- Order number or order email
- Clear photos of the damaged or missing item
- Outer packaging photo, ideally with the shipping label visible
- Tracking page screenshot or delivery-status screenshot
- Missing item’s SKU, color, size and quantity
- Customer’s preferred handling option: replacement, refund, coupon or human follow-up
Review and teach the system after each claim
After a claim is handled, do not leave the experience inside one agent’s head. Review:
- Which questions did AI fail to answer and should be added to the knowledge base?
- Which human replies can become standard compensation scripts?
- Which rules should change to “must hand off” or “requires owner approval”?
In YundaDesk, cases where AI did not answer, agents added the missing answer, or agents corrected AI can generate learning suggestions you confirm. They only take effect after an owner accepts them, and every suggestion is traceable, testable and revertible. The point is not to let AI act alone. It is to turn team experience into controlled support capability.
Appendix: claim handling checklist
- Is the issue classified as damaged, lost or missing?
- Has the order number and customer identity been collected?
- Have photos, tracking screenshots or delivery details been collected?
- Have SKU, quantity, warehouse and tracking status been checked?
- Has the customer been told that human approval is required?
- Has the final resolution been recorded back to the customer record and knowledge base?
Good compensation communication is not just polished wording. It helps the customer see progress, the agent know the next step and the approver see the basis for the decision. AI collects, organizes and reminds. Humans assign responsibility, approve compensation and own the outcome. Keep that line clear, and claims stay both humane and controlled.