WISMO automation uses an AI employee to match each “where is my order?” case to the real order and draft a reply with tracking and ETA for a human to approve — typically covering the ~50% of e-commerce tickets that are pure lookup work. Returns get triaged against your policy the same way. Nothing is sent or refunded automatically, and angry customers always escalate to a person.
If you run an online store, you already know the shape of the inbox. It’s not complaints. It’s not complex problems. It’s the same question, hundreds of times, in slightly different words: where is my order? The industry shorthand is WISMO — where-is-my-order — and depending on whose numbers you trust, it’s anywhere from a third to half of all e-commerce support volume. Add returns and exchanges and you’ve described most of the queue.
Here’s what makes WISMO maddening: it isn’t hard. Every one of those tickets is answered by looking up an order, reading a tracking status, and writing three friendly sentences. No judgment, no policy interpretation, no relationship management. Pure lookup work. And yet a human sits there doing it, one tab-switch at a time, because the customer wrote an email and someone has to write one back.
A Tuesday in November
Take an example scenario: a store doing 3,000 orders a month with one support person. In February, she handles maybe 20 tickets a day and has time to fix product page errors, chase a carrier about a lost shipment, and write the returns FAQ.
Then Q4 arrives. Order volume doubles, and support volume more than doubles — shipping delays turn one order into three anxious emails. Now it’s 60, 70, 80 tickets a day. She’s answering WISMO from 8am to 6pm, first-response time slips from four hours to two days, and the two-day delay generates more WISMO tickets from people who assume they’ve been ignored. The queue feeds itself.
The tickets that actually needed her — the damaged item, the customer threatening a chargeback, the exchange that would have saved a sale — sit in the same undifferentiated pile as “just checking on my order!!” She gets to them late, or tired, or both.
Why does seasonal hiring keep failing?
The traditional answer is seasonal staffing, and everyone who has tried it knows the trap. You start recruiting in September for a November peak. You spend October training people on your products, your policies, your tone — knowing they’ll be gone by mid-January. The math is rough: recruit, onboard, and train a temp for six-ish productive weeks, then do it all again next year with different people who make the same first-month mistakes at your busiest moment.
And the alternative — staffing permanently for peak — is worse. A team sized for November is idle in February. You’re paying twelve months of payroll for two months of load.
The problem isn’t the people. It’s that ticket volume is spiky and headcount is flat, and every way of bending headcount toward the spike costs money, quality, or both. What you actually want is capacity that scales with volume — which is exactly the shape of the problem AI is good at, because the marginal ticket costs nearly nothing. (We walked through that cost curve in the ROI of an AI employee.)
What does an AI employee actually do with a WISMO ticket?
In Turtle’s E-commerce CX and Returns pack, the AI employee is Remy, an AI CX concierge. Here’s the honest, concrete version of what Remy does — no more, no less.
Order status, handled on arrival. When an order-status case lands, an automation kicks in: Remy matches the case to the real order in your orders table — not a guess, the actual record — and drafts a friendly status reply with the tracking number and ETA. The draft waits in your queue. A human reads it, maybe tweaks a sentence, and sends. What took eight minutes of tab-switching becomes a thirty-second approval. This is the same triage-and-draft pattern we covered for general support in AI customer support ticket automation, pointed at the e-commerce-specific version of the problem.
Returns triaged against your policy. When a return or exchange request arrives, Remy checks it against the policies in your knowledge base — your return window, your condition requirements, your exclusions — and triages it accordingly. Every eligibility decision traces back to a policy you wrote, not a model’s mood.
Exchange-first steering. This is where margin protection is built into the triage. When the issue is size or fit, Remy steers the conversation toward an exchange or store credit first — the resolution that keeps the revenue — before a refund enters the picture. A tired human at ticket #63 defaults to “refund processed” because it closes the ticket fastest. A system follows the playbook every time.
Refunds prepared, never fired. When a refund is the right answer, Remy prepares it for approval. No money moves until a human says so. Full stop.
Happy endings become reviews. When a case resolves well, Remy turns it into a review request. The best moment to ask for a review is right after you’ve just fixed someone’s problem — and it’s exactly the moment a busy human forgets to ask. Over a season, that’s a steady flywheel: resolved cases feeding the social proof that drives the next season’s orders.
Behind all of this sit four tables — support cases, orders, returns and exchanges, and daily metrics — that both you and Remy read and write, so there’s always one shared picture of the queue.
What stays human — by design
The guardrails aren’t a disclaimer; they’re the operating model.
Everything is draft-and-approve. No reply is sent and no refund is issued automatically. Remy drafts; your team reviews and sends. During peak, your person’s job shifts from writing eighty replies to approving eighty drafts — same control, a fraction of the time.
Four things always escalate to a human, no exceptions: denials (telling a customer no is a human conversation), refunds over your threshold, damaged items, and angry customers. Anger is a signal that the relationship needs a person, and the system treats it that way instead of pattern-matching a cheerful template onto a furious email.
That escalation list is doing something subtle: it means the tickets your best person handles are, by construction, the ones that actually need her. WISMO stops burying the chargeback threat.
Getting started without a re-platform
You don’t need a systems integration project to try this. The pack works out of the box with orders you import — load a CSV of recent orders, replace the placeholder shipping and returns policies with your own, and Remy is answering against real data the same day. When you’re ready, connect Shopify to sync orders automatically, Stripe for refund context, and Gmail to send approved replies from the platform. Setup for the pack itself is about 20 minutes; budget a couple of hours to make the policies genuinely yours, because every reply and eligibility decision follows them.
Frequently asked questions
Does it connect to Shopify?
Yes — connect Shopify and orders sync automatically, so every case is matched against live order data. But you don’t have to start there: the pack works out of the box with imported orders, so you can run a pilot on a CSV export before touching your store’s integrations.
Can it issue refunds on its own?
No. Refunds are prepared for approval, never issued automatically — a human confirms every one before money moves. Refunds over your threshold are always escalated, and refund denials are always handled by a person rather than the AI.
What happens with angry customers?
They’re always escalated to a human. An angry customer is one of the four hard escalation rules built into the pack, alongside denials, over-threshold refunds, and damaged items. The AI never tries to placate rage with a template; it routes the case to a person who can.
Does it work if I’m not on Shopify?
Yes. The pack runs on its own tables and works with any orders you import, regardless of platform. Shopify, Stripe, and Gmail are optional connections you add later — the triage, drafting, and policy logic don’t depend on them.
If your inbox is half WISMO and peak season is already circled on the calendar, the practical next step is small: install the E-commerce CX and Returns pack, import a week of orders, replace the placeholder policies with yours, and watch how the queue feels when the lookup work drafts itself. The judgment calls stay yours — there are just finally hours in the day to make them.