Back to blog
Product

In-chat order tracking: solving "where is my order?" without a support agent

WISMO tickets are 30-40% of e-commerce support volume. How to answer them in chat, safely — with identity verification and zero AI access to personal data.

by Zedinga TeamJuly 18, 20262 min read

Ask anyone running e-commerce support what they answer most, and it's not a product question. It's "where is my order?" — WISMO, in industry jargon. Surveys consistently put it at 30-40% of all support tickets. It's also the most automatable question in your queue: the answer already exists, in structured form, in your order system.

So why don't most chatbots answer it? Because doing it safely is hard, and doing it lazily is a privacy incident waiting to happen.

The wrong way: let the AI see order data

The naive design pipes your order database into the language model's context. Now the model can, in principle, tell anyone anything — mix up two customers, hallucinate a delivery date, or get talked into revealing someone else's address by a creative prompt. Order data is personal data. An LLM is the wrong component to guard it.

How we built it instead

When a customer asks about their order, our bot doesn't try to answer. The language model's entire job is to recognize the intent and hand over to a form — and it never sees what happens next.

  1. Verify — the widget asks for the order number and the email used at checkout. Both must match the same order. Wrong email? The response is an identical "we couldn't find an order matching those details" — whether the order exists or not, so the endpoint can't be used to fish for information.
  2. Answer deterministically — on a match, the customer sees a status card: shipped/preparing, order date, item count, total, and the tracking number with a carrier link. This comes from a direct Shopify lookup, not from the model, so there is nothing to hallucinate.
  3. Minimize — the card never includes addresses or payment details, and the AI conversation history never contains the customer's email. Three failed attempts closes the form and points to your human contact channels.

What it changes

Every WISMO answered in chat is a ticket that never reached a human — at the exact moments (evenings, weekends, post-purchase anxiety) when customers most want an instant answer and least want to wait for one. And because the lookup verifies identity, you're not trading safety for speed.

See it live

This isn't a roadmap feature. It runs today on our own stores — try it on our French parts store's bot: ask "where is my order?" and watch the verification form appear (it will politely refuse to show you anything without a matching order number and email — which is exactly the point).

Setup for your store: the Shopify package (live catalog + order lookup) is an Enterprise feature — every connection is configured hands-on by our team, so get in touch and we'll set yours up. Start free to test the bot on your site content first.

Share
The Zedinga Team

The Zedinga Team

Founders & Engineers, Zedinga LLC

Zedinga is built by a founding team with over 20 years of professional software experience and more than 200 web and mobile projects delivered for clients worldwide. After two decades of building software for others, we now build our own products — including Zedinga AI, which runs the customer-facing AI on our own e-commerce stores and service businesses. Features get battle-tested on real businesses before they're called done.

More about Zedinga →

Keep reading