AI for order support: automating „where is my order" (WISMO) questions
„Where is my order" (WISMO) is the most common and most repeatable category of contacts in eCommerce. How eCommerce customer service automation takes it over in a draft-first model: AI writes a draft with order context and status, and a human approves or escalates.
„Where is my order" is the most common question customer service gets in eCommerce, and also the most repeatable, because the answer almost always sits in the same two places: the order status and the shipment tracking. That is exactly why this type of ticket, known in the industry as WISMO (from Where Is My Order), is the best first step when you think about eCommerce customer service automation. The honest version of that automation is not a bot answering customers on its own. It is a draft-first model: AI assembles a ready reply with the current status and a tracking link, and a human approves it or escalates. Below: what exactly AI takes off your plate, which integrations it needs, and what deliberately stays with the human.
Key takeaways
- WISMO („where is my order") is usually the single largest category of support tickets in eCommerce and the most repeatable, because the answer almost always sits in the order status and the shipment tracking.
- The safe model is draft-first / human-in-the-loop: AI assembles a ready reply with the current status and tracking, and a human approves, edits, or escalates. Not full autonomy from day one.
- Without integrations, AI guesses. You need three sources: order statuses, shipment tracking, and a knowledge base with your policies and brand tone.
- Hard cases, decisions, and empathy deliberately stay with the human: complaints, lost parcels, upset customers. That is not a gap in automation, it is its proper scope.
What WISMO is and why it eats up support
WISMO stands for Where Is My Order. It covers a whole category of tickets about the status and location of an order: „has it shipped yet", „when will it arrive", „I have a tracking number but the status hasn’t changed in two days", „I changed my mind about the address". In many stores this is the single largest category of contacts, and during peak seasons, sales, and carrier delays it can dominate the entire support inbox.
These tickets eat up support for a simple reason: they are highly repeatable and mildly labor-intensive at the same time. To answer, someone has to find the order by number or email, check its status in the admin panel, open the carrier tracking, translate a technical status into a human message, and write a polite reply. That is two, three, maybe four minutes of clicking between systems, and the reply is almost identical every time. Multiply that by dozens or hundreds of tickets a day and you have a team that spends most of its time retyping the same information instead of handling the cases that genuinely need a person.
Why WISMO is a natural fit for automation
WISMO has two traits that make it an almost textbook candidate for automation. First, it is extremely repeatable, so every minute saved multiplies across a large number of tickets. Second, and more important, the answer barely requires judgment, only access to data: you take the order status, the shipment status, and a simple communication rule, and the reply follows. That is exactly the profile of a task worth automating, unlike tickets where the situation and the emotions are what matter. We unpack the rule „automate only what saves time, improves quality, or supports sales" in more depth in our piece on the anatomy of automation savings, but WISMO meets it almost by the book.
Vendors of support-automation tools claim that a large share of first-level (L1) queries, WISMO included, can be handled automatically, often in the range of roughly 80-90 percent. Treat those figures as vendor claims, not a measured result for your store. The real share depends on the quality of your status data and on how many tickets are genuine WISMO rather than hidden complaints that merely start with a question about a parcel.
The draft-first model: AI writes the draft, a human sends it
The safe way to automate WISMO is not an autonomous bot but draft-first with a human in the loop. In practice it works like this: a ticket arrives by email or chat, AI recognizes that it is a question about order status, and pulls the context, meaning the order number, its current status, the shipment tracking data, and the applicable delivery policy. From that it assembles a ready reply in your brand tone, with concrete information and a tracking link instead of vague reassurance.
What happens next is the crux. The draft does not go to the customer automatically. It appears next to the ticket in your helpdesk, and a human does one of three things: approves it with a single click, edits it in a few seconds, or escalates it if they see the case is trickier than it looked. When a ticket is ambiguous, emotional, or about a lost or damaged parcel, a well-designed workflow flags it to a human right away instead of pretending it can cope. This flips the balance of work: the human stops writing from scratch and starts approving and deciding. That is exactly how our AI in customer service works: AI writes drafts from order data and your policies, and a human sends them.
Which integrations it needs
Automating WISMO is only as good as the data it can reach. Without integrations, AI guesses, and guessing in support is worse than no answer, because it sounds confident and misleads the customer. You need three sources and one workspace:
- Order statuses (the store platform, ERP, or order system): the source of truth for what is happening with an order, whether it is packed, shipped, or on hold.
- Shipment tracking (a carrier API or a tracking aggregator): where the parcel physically is and the expected delivery date, without manually opening the carrier’s site.
- Knowledge base and policies (delivery and returns rules, brand tone): so the draft follows your rules, not the model’s general knowledge. We collect ready templates and skills for tasks like this in our free AI library for eCommerce.
- Helpdesk or inbox: the place where AI drops the draft next to the ticket and where a human approves, edits, or escalates it.
What AI takes over, and what stays with the human
Not every question that starts with „where is my order" is genuine WISMO. Some are hidden complaints, decisions, or situations where empathy is what counts, and those have to reach a human. The table below draws the line: what AI can prepare on its own as a draft, and what we deliberately keep with people.
| Query type | What AI does | What stays with the human |
|---|---|---|
| Where is my order / order status | Draft with current status, tracking link, and expected date | One-click approval |
| When will I get my order | Draft with the date from tracking and delivery policy | A check when the date is unusual |
| Change of address or delivery method | Draft with options and a „needs a decision" flag | Deciding whether we can catch it before dispatch |
| Where is my return / when is my refund | Draft with the return status and the date from policy | Reviewing disputed cases |
| Parcel delayed beyond the norm | Gathers context and proposes a draft, but flags for review | Deciding on compensation or a carrier intervention |
| Parcel lost or damaged (complaint) | Gathers data but escalates without sending a draft | Full handling, decision, empathy |
| Upset customer, threat, unusual case | Detects the tone and escalates immediately | A human takes over from the first message |
The line is deliberate: AI handles repeatable, data-based queries as drafts, while anything that needs judgment, a decision, or empathy goes to a human. That is not a gap in automation, it is its proper scope.
Look at the last two rows. Automating WISMO is not meant to handle complaints or defuse emotions, only to relieve the team of repeatable questions so it has the time and the headspace for the hard cases. A workflow that tries to automatically „sort out" an upset customer with a lost parcel does more harm than good.
How to roll it out safely: not full autonomy from day one
The most common mistake is turning on full autonomy on day one and letting AI answer customers by itself. One bad automated message sent to an upset customer costs more than ten correct ones saved, so the order of steps matters. A safe rollout goes in stages.
Start narrow: one ticket type, genuine WISMO, on one channel. Set draft-first mode, where a human sees every draft before anything goes out, and measure how often the draft ships without edits. That share grows over time, because every reviewer correction improves the rules and templates until quality stabilizes. Only then do you widen the scope: more query types and, where the risk is genuinely low, less manual control. Throughout, leave two things untouched: a hard escalation path to a human and a clear signal that the other side may be an assistant, not a person. Support automation should speed the team up, not pretend to be someone it is not.
Where this fits into the platform
For us this automation is not a separate bot bolted on the side, but a layer inside the platform, wired to real order statuses and shipment tracking. It is part of the AI Automation stage in the BEAM framework: we build the workflow where the numbers add up, in a draft-first model, and we leave the human the decisions that should stay with the human. We do not sell a bot that pretends to be a person. We build a layer that takes hundreds of repeatable „where is my order" tickets off the team and gives it back the time for the cases no automaton will ever handle for it.
FAQ
What is WISMO?
WISMO stands for Where Is My Order. It covers the whole category of customer service tickets about the status and location of an order. In many eCommerce stores it is the largest and most repeatable category of contacts, because the answer almost always follows from the order status and the shipment tracking.
What share of order-status questions can be automated?
Support-automation vendors often quote a range of roughly 80-90 percent for first-level queries, WISMO included, but those are vendor claims, not a measured result for a specific store. The real share depends on the quality of your status data and on how many tickets are genuine WISMO versus hidden complaints. It is safer to start with a draft-first model and measure your own result than to trust someone else’s number.
Does the AI reply to the customer on its own?
In the safe model, no. We work draft-first with a human in the loop: AI assembles a reply with the current status, tracking, and your policies, and a human approves, edits, or escalates it. Full autonomy is enabled later at most, and only for narrow, well-tested cases, never from day one.
Which integrations are needed to automate WISMO?
Three data sources and one workspace: order statuses (the store platform or ERP), shipment tracking (a carrier API or a tracking aggregator), and a knowledge base with your policies and brand tone, plus a helpdesk where AI drops the draft next to the ticket. Without these integrations AI guesses, and guessing in support is worse than no answer.
Journal
Co-founder of Seedlight · eCommerce platforms, AI, SEO and GEO
Newsletter
The Journal, straight to your inbox
New articles and lessons from real builds, every now and then. No spam, unsubscribe with one click.
