I sat down last weekend and went through 50 support tickets from the first wave of OrderWise stores. The plan was to label each one and see what the actual distribution looks like — not the marketing version, the real one.

Here's what 50 tickets looks like:

39 (78%) were "Where is my order?" / "When will it ship?" / "Tracking link?" — tickets that contain zero new information beyond the order ID.
7 (14%) were refund or cancel requests — humans still need to decide, but the conversation can be pre-structured.
4 (8%) were the actually-hard ones: sizing problems, custom requests, complaints with photo attachments, edge cases.

The 78% bucket is where every small store owner I've talked to is bleeding hours. They've typed the same five-line "your order ships from our warehouse in 2-3 days" response so many times that it has stopped feeling like work and started feeling like punishment.

Why "WISMO" eats your inbox

WISMO — "Where Is My Order?" — is the dominant ticket category in e-commerce support. Different sources put it at 50-80% of inbound volume depending on niche, but for digital-first DTC stores with sub-€100 AOV, the upper end is the norm.

The structure of these tickets is almost always the same:

Customer placed an order 2-5 days ago.
They want a status update.
They want a tracking link or an ETA.

You don't actually need a human to write the answer. The data is sitting in your Shopify admin. The customer asking just wants confirmation that the order exists, that it's moving, and roughly when it will arrive.

So the question becomes: why is anyone still answering these by hand?

The honest answer: most chatbots can't do it

I've sat with merchants who paid €29/month for a chatbot that, when a customer asked "where is my order #1047?", replied with:

Hi! To help you with your order, please share your order number.

This is the failure case I've seen in every chatbot vendor that doesn't actually connect to Shopify properly. They have a "WISMO intent" trained, they recognize the question, but they cannot resolve it because they don't have order context. So they fall back to either:

Asking for the order number again — even when it's literally in the customer's message.
Pasting a generic FAQ block — "shipping takes 3-5 days" — which is exactly what the customer is not asking.
Routing to email — which means the merchant ends up answering it manually anyway, but with extra delay.

When I prototyped the first version of OrderWise, the smallest possible MVP that I wanted to ship was: a chatbot that, given any plain-text message from a customer, looks up the relevant order without asking the customer for the order ID first. Just that one capability already kills 78% of inbound support work.

What "auto-resolvable" actually requires

I'll be specific because this matters for anyone building or evaluating these tools. To actually auto-resolve a WISMO ticket, the system needs four things:

Customer identity binding. The widget must know which logged-in customer is talking — or extract it from the email/order number in the message body. Anonymous chats are a dead end here.
Order matching that handles fuzzy input. Customers paste "#1047", "1047", "ORDER-1047", or sometimes the wrong format entirely. The regex on the input side has to be tolerant. About 30% of the messages in my sample had the order number written in some form the customer hadn't been instructed to use.
Real-time Shopify data, not a cache. Tracking status changes hour to hour. A daily-sync cache will tell the customer the package is in Frankfurt when it has actually arrived. The lookup has to hit the Admin GraphQL API at query time. Yes, that's a rate-limit risk; yes, that's solvable with per-shop request budgets.
A useful answer format. "Your order is in transit" is not enough. The customer wants the carrier name, the tracking number, the last-known location if available, and a clickable tracking link. The minimum acceptable response has roughly five data points in it.

If you have all four, the WISMO category collapses. If you're missing any one of them, you end up with the chatbot that asks for the order number it already has.

The 14% that needs structure, not autonomy

For refund and cancel requests, the right answer is not full automation. I tried it briefly in an internal build and the result was bad: the AI would correctly identify the intent ("customer wants refund"), correctly look up the order, and then make a refund decision the merchant disagreed with about 40% of the time.

Merchants want control over money decisions. That's not a UX preference, that's a basic business requirement. What they don't want is to manually parse a half-paragraph email to extract the order ID, the reason code, and the requested action.

So in OrderWise, refund and cancel flows do something different: the AI takes the conversation, asks the missing structured questions (reason, refund-or-replacement, photo if relevant), and produces a single inbox card with:

Order context inline (number, status, value, items)
Conversation summary
Suggested action with explanation
A one-click "Approve" or "Reject" button

The merchant now spends 10-15 seconds per request instead of 5-10 minutes. The decision quality is the same — the merchant is still the one deciding — but the manual data-gathering step is gone.

That's the 14% bucket. It can't be fully automated. It can be pre-structured so that the human only does the part that requires human judgment.

The 8% that genuinely needs a person

This is the bucket I'm careful not to over-claim about. Sizing complaints with attached photos, custom engraving requests, gift-recipient address corrections after the package shipped, complaints that escalate quickly when handled poorly — these tickets benefit from a human reading them carefully and writing a thoughtful response.

The AI's role here should be routing, not answering. Identify that the ticket falls outside the auto-resolvable categories, classify it (complaint, custom, post-shipment edit), surface it in the inbox with appropriate priority, and stay out of the way.

I've seen tools that try to handle this bucket with AI-drafted responses that the merchant reviews. In my experience, the time saved on writing is roughly equal to the time spent reading the AI draft and rewriting half of it. The honest answer is: for 8% of your volume, just write the response yourself. It's worth it.

Numbers that hold up in production

The 78/14/8 split has been stable across the 50-ticket sample and the next ~200 conversations we've watched go through. The variance you should expect:

Niche-dependent. Apparel and shoes skew higher on the 8% bucket because sizing issues are common. Digital products or shippable consumables skew higher on the 78% bucket.
Pre-purchase vs post-purchase. I excluded pre-purchase questions ("do you ship to Austria?") from this analysis because they have a different structure. If you include them, the 78% drops to roughly 65% — pre-purchase questions tend to need product knowledge that the AI either has via your FAQ or doesn't.
Volume changes the math. A store doing 20 tickets a week feels the 78% as "saved a few hours." A store doing 200 tickets a week feels it as "hired half a support agent."

What I'd build differently if I started today

Two things, in hindsight:

Bind customer identity at the widget level earlier. Anonymous chat sessions are a UX nightmare. The first version of OrderWise let any visitor open the widget without being logged in, and the order-matching success rate was lower than it should have been. Tightening that flow to "you must enter an email if you're not logged in" recovered about 12 percentage points of auto-resolution.
Don't try to be smart about edge cases until the 78% works perfectly. The temptation is real. Every merchant has a "but what about this weird case?" question. Resist. Get the boring 78% bucket to feel instant and reliable, then go after the next layer.

If you're building or evaluating customer support tools for Shopify and want to compare notes on what actually works, my email is in the footer. I'm always interested in seeing real ticket distributions from other niches.

Try OrderWise free → Free plan: 25 conversations/month, full Shopify order context, no credit card. Install from the Shopify App Store.

78% of 'Where is my order' tickets are fully auto-resolvable