YundaDeskWhen a customer says a package is lost, support teams often make one of two mistakes: they send a vague “please wait a little longer”, or they promise a reship, refund, or compensation before the facts are clear. A better lost package reply follows a sequence: verify the customer and order, collect evidence, check shipping status and address, then send reshipment or compensation decisions through human approval. AI can handle intake, but decisions that move goods or money need a person.
First classify it: delay, misdelivery, or likely loss
| Situation | Common signals | First action |
|---|---|---|
| Shipping delay | Tracking has not moved from transit, customs, or out-for-delivery status | Explain the current status and set the next follow-up time |
| Misdelivery or not received | Tracking shows delivered, but the customer says nothing arrived | Confirm address, delivery proof, nearby pickup points, and possible recipients |
| Likely loss | Tracking is abnormal, the carrier confirms loss, or your internal time limit has passed | Collect evidence and route to human approval for reship or compensation |
Start with language that is calm and specific:
I will verify the order and shipping record first. To avoid misclassifying the case, I will check the parcel status, delivery address, and delivery confirmation. If the record shows a real loss risk, we will submit it for specialist review for reshipment or compensation.
It shows action without promising an outcome before the facts are clear.
Collect evidence: ask once, not five times
Lost-package cases become frustrating when the customer has to repeat themselves. Use a fixed checklist:
- Order number or order email
- Recipient name, ZIP or postal code, and the last part of the shipping address
- Tracking number or a screenshot of the latest tracking status
- Whether the customer checked with household members, reception, neighbors, or pickup locations
- If tracking shows delivered, any notes about the doorway, mailbox, parcel locker, or delivery photo
This is a good job for AI support: ask for missing details, look up the order, and prepare a structured summary for humans.
Lost-package intake should narrow before human approval
That sequence turns waiting into visible progress: AI gathers the standardized facts first, then humans decide whether goods or money should move.
Verify the order: match the story against your records
After intake, check three things:
- The order exists: the email, phone number, or order ID matches your records.
- The address matches: the address the customer gives now matches the order address, including any address-change history.
- The fulfillment path is complete: the order was shipped, handed to the carrier, and has no return, refusal, customs, or exception event that explains the issue.
Give the customer a clear status:
I have found your order and tracking number. The latest record shows the parcel was updated as {status} on {date/time}. I still need to verify the delivery or scan details before treating this as a lost package. We will update you again within {timeframe}.
If the same customer contacts you through email, WhatsApp, Instagram, and your website widget, those messages should land in one omnichannel inbox, so agents do not restart order lookup.
Reply scripts: what to say at each stage
Use these templates as a base.
| Stage | Script |
|---|---|
| First complaint received | Hi, we have received your message that the package has not arrived. I will verify the order, tracking history, and delivery record first. Please share your order number or order email so we can investigate it quickly. |
| Tracking still moving | I checked the tracking record and the parcel is still in transit / customs / transfer status. At this point we cannot confirm it as lost yet. We will keep monitoring it and update you again by {date}. |
| Delivered but not received | The carrier record shows the package was delivered on {date/time}. To help us verify the case, please check the doorway, mailbox, parcel locker, reception, or neighbors if applicable. We will also review the carrier record on our side. |
| Moving to human review | This order now meets the conditions for manual review. I will submit the order details, tracking record, and the information you provided to our team for a reshipment or compensation decision. We will notify you as soon as the review is complete. |
Promise verification, review, and follow-up. Do not promise an immediate reship or guaranteed compensation before evidence and approval are in place.
Reshipment and compensation: AI prepares, humans approve
Once a likely loss moves toward reshipment or compensation, make the boundary explicit.
| Action | AI can do | Humans must do |
|---|---|---|
| Reshipment | Summarize order details, inventory, address, and logistics exceptions | Decide whether to reship, which carrier to use, and whether shipping is covered |
| Compensation | Gather evidence, match policy, and draft a recommendation | Approve the amount, coupon, refund, or other remedy |
| Refund | Explain the process, collect materials, and share status updates | Approve the refund and execute any money movement |
YundaDesk’s AI agent is built for intake and preparation: answer first, collect evidence, look up orders, and hand off when the answer is uncertain, the customer asks for a person, or the action is high risk. Refunds, compensation, and price changes always require human approval. That is AI answers first, humans back up: speed without losing control.
Follow up across channels: stop making customers repeat the story
A customer with a missing package rarely sends only one message. They may email first, send an Instagram DM later, and then open your website widget. If every channel is separate, agents repeat the investigation and the customer feels nobody owns the case. Standardize three things:
- Bring email, website widget, WhatsApp, Messenger, Instagram, TikTok, LINE, and other customer messages into one workspace.
- Keep country, language, time zone, social IDs, order history, and conversation history on the customer profile.
- Let AI follow the customer’s language automatically, while high-risk decisions still follow your internal approval rules.
With that setup, the next agent sees full context instead of restarting from order lookup.
Review and improve: turn this case into the next standard answer
Every lost-package case should leave behind three kinds of learning:
- Which tracking statuses customers often mistake for loss.
- Which replies calm the conversation, and which ones make it worse.
- Which edge cases belong in the knowledge base, such as “delivered but not received”, “carrier requires investigation time”, or “customer wants a reship before verification is complete”.
In YundaDesk, cases the AI missed, human answers, and agent corrections can generate learning suggestions you confirm. They only become skills, knowledge, or customer memory after review and adoption. Each item is traceable, testable, and revertible; learning never takes effect automatically. For setup guidance, see build a knowledge base that actually feeds AI.
The point of lost-package scripts is not to talk the customer down. It is to show real progress: collect the right information once, verify the order cleanly, route risky actions through approval, and follow up with ownership.