If you run CX for an ecommerce brand, you already know the shape of your ticket queue: it's a bell curve with "where's my order" at the peak and "does the medium run small" a close second. These aren't hard questions. They're just volume. And they arrive on the slowest, most expensive channels you own — email and live chat, both of which force a human (or a chatbot with a spinning cursor) to answer something the tracking page already knows.
Full disclosure: I work for Ready, an SMS platform. So I have a horse in this race. But the argument here isn't "buy SMS." It's narrower and more defensible: a huge slice of your inbound is repetitive, status-driven, and answerable from data you already have — and text is the right surface for it because your customer is already getting order texts from you.
Why WISMO is a text problem, not a chat problem
WISMO — "where is my order" — is the single most common ecommerce support query. Industry estimates put it somewhere between 30% and 50% of all inbound contacts for brands that ship physical goods. Call it roughly a third, conservatively.
Here's the thing that makes it a text problem specifically: you already text these people. If you send a shipping confirmation over SMS, the customer now has a thread with your brand sitting in their messages app. When their package is late, their most natural move is to reply to that thread — not to hunt down your support email, not to open a chat widget and wait in queue.
So the reply comes in as "where's my order??" on a thread that's already tied to an order number. All the context you need to answer is one lookup away. That's the whole opportunity: the customer self-selected the channel where you have the most context and the lowest per-interaction cost.
The math: cost per resolved query
Let's price a resolved WISMO query on each channel. These are approximations — your loaded agent cost will differ — but the ratios hold.
| Channel | What one resolution costs | Notes |
|---|---|---|
| Live chat / phone | ~$4–7 | Loaded agent time, ~5–8 min per contact incl. queue |
| ~$1.50–3 | Async, but still human time + back-and-forth | |
| SMS with AI reply agent | ~$0.05–0.10 | 1–3 outbound segments, no human touch |
On Ready, an outbound SMS segment is $0.02 plus a $0.0045 carrier pass-through — $0.0245 all-in per segment on the Standard tier (0–50,000 segments/month). A typical WISMO answer — "Hey Sam, your order shipped Tuesday and is out for delivery today. Track it here: [link]" — is about 110 characters, so one segment. Even a chattier two-turn exchange (they reply, you answer again) is maybe 3–4 segments total: under 10 cents to fully resolve.
Run that against your queue. Say you field 10,000 support contacts a month and a third are WISMO — 3,300 tickets. If those currently hit chat at ~$5 each, that's $16,500/month in agent time on a question a database lookup answers. Deflect them to text at ~$0.08 resolved and you're at roughly $264/month for the same volume.
That's not a rounding error. That's a headcount decision.
What the AI reply agent actually does (and doesn't)
The reason this works — and doesn't read like a 2015 chatbot — is that the automation isn't trying to converse. It's trying to answer one of about six known questions and hand off the rest.
Ready's two-way messaging lands every inbound reply in a unified inbox, and the optional AI reply agent runs in one of three modes:
- Off — replies just sit in the inbox for a human.
- Suggest — the agent drafts a reply and an agent approves/edits before it sends. Good for your first few weeks while you build trust.
- Auto — the agent answers directly for high-confidence, known intents (order status, return window, sizing) and escalates anything ambiguous to a human.
The trick to not sounding like a bot is scope discipline. Point the agent at your order data and a short knowledge base (return policy, sizing chart, shipping timelines). Let it handle exactly those. Everything else — "this arrived broken," "I want a refund," "you charged me twice" — should route to a person immediately. A bot that confidently mishandles an angry customer costs you far more than the ticket you deflected.
Start in Suggest mode. Read the drafts for two weeks. When you can predict what the agent will say and you agree with it ~95% of the time, flip the intents you trust to Auto and leave the rest on Suggest.
Sizing and fit: the second-biggest deflection win
Apparel and footwear brands, this one's yours. "Does this run small?" and "what size am I if I'm normally a medium?" are pure knowledge-base questions. If the agent has your sizing chart and fit notes per product, it answers instantly — and a fast, correct sizing answer prevents a return, which is a bigger dollar win than the deflected ticket.
A returned $60 apparel order can cost you $15–25 all-in once you count return shipping, restocking, and the item that ships back unsellable. Answering "our runners fit true to size, but if you're between, size up — the toe box is snug" before purchase quietly kills a chunk of that. The deflection savings are the appetizer; the return prevention is the entrée.
Where this loses — and you should send anyway or not at all
I'd rather you set this up correctly than oversell it, so here's the honest boundary:
- Anything emotional or money-losing for the customer — damaged goods, wrong item, double charge, "I've been waiting three weeks." Auto-answering these is how you turn a recoverable customer into a chargeback and a one-star review. Escalate on the first sign of frustration.
- Anything requiring a policy exception. The agent shouldn't be inventing goodwill credits.
- Consent boundaries. WISMO replies to a transactional thread are fine — the customer initiated. But don't let a support thread quietly become a marketing channel. If you want to promote off the back of a support win, you need the right opt-in. We wrote about the clean way to do that in the one-tap post-purchase consent upgrade.
If you're not sure whether a moment belongs on text at all, 5 ecommerce moments where SMS loses to email is a useful gut-check.
Setup: the keyword routing underneath it
Before you turn on the AI agent, get the deterministic routing right — because some replies shouldn't touch a model at all. "STOP" must opt the contact out and propagate that across every campaign. "HELP" returns your compliance help text. Those are carrier-mandated, non-negotiable, and Ready handles STOP/opt-out automatically. Order-status intents route to the agent. Everything else routes to a human.
If you want the mechanics of building that tree — what gets a canned reply, what gets the agent, what gets a person — building a keyword routing tree that handles STOP, HELP, and YES differently walks through it. And for GHL users, inbound replies sync two-way into your location's conversations, so your agents see the same thread whether they're in Ready or GHL — the GHL SMS setup guide covers wiring that up.
A rough rollout order:
- Turn on two-way replies for your shipping-confirmation number.
- Confirm STOP/HELP handling and quiet-hours enforcement are active.
- Load your order-status data + a short KB (returns, sizing, shipping windows).
- Run the AI agent in Suggest for 2 weeks; audit the drafts.
- Flip only the high-confidence intents to Auto; keep escalation aggressive.
The practical takeaway
You're already paying to text these customers their order updates. When they reply — and a third of them reply asking where their order is — you can either route that into a $5 chat queue or answer it for under a dime from data you already have. The AI reply agent makes that scalable; scope discipline and fast escalation keep it from sounding like, or behaving like, a bot.
If you want to sketch your own numbers, our ecommerce SMS ROI calculator post is a decent starting frame. And if you'd rather just try it, Ready ships 2,500 free credits with no card required — enough to run WISMO deflection on a real week of your queue and see the deflection rate for yourself. Start here.