"Where's my order?": Automating the No. 1 Question in E-commerce Customer Service

James
Customer Service
- 8 min reading
Published on
October 16, 2026
Packages routed from a single point to multiple order-tracking scenarios, most of which result in an automated response

To automate the question “Where is my order?”, connect your AI to the order and carrier tracking systems, then define for each situation (in transit, delayed, delivered but not received, not yet shipped, etc.) what it should check, what it should respond, and when it should hand the matter over to an agent. Most of these inquiries are then resolved without human intervention, and your agents only handle cases that require a decision: lost packages, disputes, or goodwill gestures.

This guide is a practical supplement to our article on order tracking automation (WISMO), which outlines the key strategies and tools. Here, we’ll dive into the details at the ticket level: eight real-world scenarios, along with the logic you should program into your AI for each one.

What does AI need to know to answer "Where is my order?"?

A good response depends on four pieces of information. Without them, the AI can only redirect the customer to the tracking page—which the customer has often already visited.

  1. Who is the customer: The AI must identify the correct order based on the order number, email address, or phone number.
  2. Order status: paid, being processed, shipped, partially shipped, canceled.
  3. Shipping tracking: latest update, estimated delivery date, pickup location, proof of delivery.
  4. Your Rules: preparation deadlines, the timeframe beyond which a delay is considered abnormal, and your return or refund policy.

The first three come from your tools (online store, ERP, shipping carrier). The fourth one is up to you to write. It’s what makes the difference between an AI that gives accurate answers and one that makes empty promises.

Which data source should be connected?

Connect to the source that has the most up-to-date information. This point seems obvious, but it holds up a lot of projects. An e-commerce manager put it this way during a demo:

"When we make a change to an order in our ERP system, it doesn't get updated in the store. How can the bot take that into account?"

If your ERP updates the online store (Shopify, Magento, PrestaShop) with a one-day delay, the AI connected to the store will respond with outdated information. Before running the automation, check the following:

  • where the shipping status is updated first;
  • if the tracking number is indeed included in the order;
  • If orders consisting of multiple packages do have a tracking number for each package.

If the information exists only in the ERP system, it must be made available to the AI, for example via an API. Otherwise, it is best to exclude these cases from automation.

How should AI handle each situation?

Here are the eight most common scenarios, along with an example of a customer message and the expected behavior.

1. The package is in transit and on schedule

"Hello, I placed an order three days ago. When will I receive my package?"

The AI retrieves the order, reads the latest carrier update, and responds with the status, expected delivery date, and tracking link. No agent is needed. This is the most common scenario, and the one that clogs up the queues the most.

2. The order has not yet been shipped

"My order is still being processed—is that normal?"

The AI compares the order date to your fulfillment time. If the order is on schedule: it reassures the customer and provides the expected shipping date. If the order is past the deadline or on hold (pending payment, out of stock): it notifies the customer that an agent is looking into it and forwards the ticket along with the context.

3. The package is late

"The delivery was scheduled for yesterday, but I haven't received anything."

The AI checks the discrepancy with the announced date. If there’s a slight delay but the tracking information is updating, it apologizes, provides the new estimate, and shares the link. If the delay exceeds your threshold, or if the tracking has been stuck for several days, it hands the matter over. Set this threshold in writing, with the carrier if necessary.

4. The package is marked as delivered, but the customer hasn't received it

"The tracking information says it was delivered, but I haven't received anything."

AI can handle the initial steps: checking the mailbox, the doorman, the neighbors, and the designated pickup location. It must then hand the matter over to a representative, since the next steps—investigating the carrier, reshipping, or issuing a refund—involve the brand. The customer needs to know that a human will take over and within what timeframe.

5. The package is waiting at a pickup location

"I didn't receive a code to pick up my package."

The AI provides the pickup location address, the hours of operation (if provided by the carrier), the pickup deadline, and the documents required for pickup. If the pickup deadline is missed and the package is returned, it forwards the information to an agent.

6. The order was shipped in several packages

"I received my package, but one item is missing."

Before concluding that an item was omitted, the AI checks whether the order was shipped in multiple shipments. If so, it specifies which item is included in which package, along with each tracking number. Otherwise, it’s a complaint: agent.

7. The customer wants to change the shipping address

"I entered the wrong address. Can you change it?"

If the order has not yet been shipped and your tools allow it, the address can be corrected—either by the AI or by an agent—depending on your policies. If the package has already been shipped, the AI explains the carrier’s options (redirection, pickup location) and notifies you if manual action is required.

8. The package is lost or damaged

"My package arrived open, and the product is broken."

The AI immediately requests the necessary information: photos of the package, the packaging, and the product, as well as the order number. Then it hands things off. A customer service manager described this process to us in the context of furniture delivered in damaged condition: “The initial response could be automated, because people always say pretty much the same thing: ‘Please provide us with photos of the boxes and packaging.’ After that, a human needs to take over.” The agent receives a complete case file and can make a decision right away.

When should AI hand things over to a human agent?

The table below summarizes the logic. Adjust the thresholds to fit your deadlines and carriers.

LocationWhat the AI ChecksWhat she saysAssignment to an agent
In transit, on scheduleLatest event, scheduled dateStatus, Date, Tracking LinkNo
Not shipped yetOrder Date, Processing TimeEstimated Shipping DateIf the deadline has passed or the order is on hold
LateDifference from the announced dateApologies, revised estimateBeyond your threshold or fixed monitoring
Delivered but not receivedProof of delivery, addressFirst StepsYes, always
At a pickup locationAddress, deadline for pickupPickup InformationIf the package is shipped again
Shipping in multiple packagesNumber of packages, contents of eachDetails by Package and Tracking InformationIf an item is really missing
Change of AddressShipping StatusCorrection or carrier optionsIf manual action is required
Lost or DamagedOrder, StatusRequest for Photos and InformationYes, with the complete file

Add two rules that apply in all situations: a dissatisfied customer or one who contacts us for the third time is routed to an agent, and the AI never promises a goodwill gesture that it is not authorized to grant. Our guide on escalating issues to support details how to write these rules.

How can you verify a customer's identity before providing information?

An email address entered in a chat does not prove a person's identity. If the AI displays the contents of an order and the shipping address based solely on an email address that was entered, anyone can view another customer's orders.

Best practices :

  • by email, reply only to the email address associated with the order;
  • In the chat, ask for the order number in addition to the email address, or use the logged-in customer's session;
  • Never include the full address or payment details in your reply.

Ask your service provider about this and request a demonstration. Our Security and GDPR page explains our approach in detail.

Can we notify customers before they ask?

Yes, and it’s often more effective than even the best response. When a delay is known (supplier shortage, strike, carrier capacity issues), reach out to the affected customers before they contact customer service: provide a new date and possible options (wait, exchange, cancel). This helps you avoid a flood of identical support tickets.

This type of proactive message requires a reliable source (ERP, warehouse, or carrier) and a clear policy regarding what you’re offering. It’s particularly useful during peak periods: see our guide on how to handle sales and Black Friday spikes.

How can automation be implemented without risk?

  1. Start with Case 1 (in transit, on time). This is the most common and least risky scenario.
  2. Start in suggestion mode: The AI suggests a response, and the agent approves it. You catch data errors before a customer sees them.
  3. Set approved cases to automatic processing, one channel at a time (email, then chat).
  4. Add the other scenarios one by one, along with their thresholds.
  5. Each week, review a sample of responses and the tickets forwarded to agents.

If your store runs on Shopify, our article on AI for customer service in a Shopify store details some useful integrations.

At Klark, AI works directly within your help desk (Zendesk, Gorgias, Salesforce, Freshdesk, Intercom) and connects to your online store to retrieve order information and tracking details. On average, our customers automate 43% of their tickets, and order tracking requests are often the first to be automated.

Frequently Asked Questions

Can AI answer 100% of "Where's my order?" inquiries?

No. Lost packages, packages that were delivered but not received, or damaged packages require human intervention. AI, however, can handle standard tracking requests on its own and prepare the case file for other situations.

Is a package tracking tool necessary in addition to AI?

Not necessarily. If the tracking number and carrier updates are synced back to your store, the AI can use that information. A post-purchase tracking tool is still useful for sending notifications and reducing the number of inquiries beforehand.

What should I do if the customer doesn't provide their order number?

The AI can find the order using the sender's email address, if it matches the one on the order. In the chat, it asks for the order number to verify the identity before responding.

How do I handle orders with multiple packages?

Verify that each package in the order has its own tracking number. The AI can then display the status of each shipment and prevent false claims of missing items.

Does automation also work via email?

Yes. Follow-up requests often come in via email or the contact form. The same process applies: the AI reads the request, locates the order, and responds in your helpdesk.

Want to see how Klark handles order tracking requests on your own tickets? Request a demo.

You might like

Klark blog thumbnail
- 5 MIN READING 

Sales and Black Friday: How Can You Handle Peaks in Customer Service?

Ticket volume will double—or even more: Here’s how to prepare your customer service team for Black Friday 2026 and the 2027 sales. A comparison of options, a preparation plan, and KPIs.
Klark's author
Co-founder and CEO
Klark blog thumbnail
- 5 MIN READING 

How can you automate order tracking requests (WISMO)?

Are "Where's my order?" inquiries overwhelming your customer service team? Here are the strategies for automating WISMO, the tools to compare, the cases to escalate, and the KPIs to track.
Klark's author
Co-founder and CEO
Klark blog thumbnail
- 5 MIN READING 

How can customer service automate returns and refunds?

Are return requests overwhelming your customer service department? What can be automated, what requires a human touch, up-to-date legal requirements (including the cancellation button), and how to get started.
Klark's author
Co-founder and CEO
Klark blog thumbnail
- 5 MIN READING 

Customers don't go on vacation with your team

Preparing Customer Service for the Holidays with a Reduced Staff: What to Automate, What to Handle First, What to Escalate, and When to Ask Customers to Wait Before, During, and After the Holidays.
Klark's author
Co-founder and CPO
Klark blog thumbnail
- 5 MIN READING 

Why Your Knowledge Base Is No Longer Enough—and How RAG Is a Game-Changer

Support information already exists in the FAQs, as well as in thousands of resolved tickets. The challenge is no longer just to write it down; it’s about finding the right source at the right time.
Klark's author
Co-founder and CPO
Klark blog thumbnail
- 5 MIN READING 

Alternatives to Intercom and Fin: How to Choose Without Replacing Your Support Tools

Salesforce's acquisition of Fin provides an opportunity to take stock. Public facts, a simple framework for deciding whether to migrate, and criteria for evaluating alternatives for your tickets.
Klark's author
Co-founder and CPO