The pitch aimed at online stores is almost always the same. Buy this AI and your ads get cheaper. Sometimes that's even true. More often the store is already paying to send people to a site that answers "where is my order?" with a contact form, and then a human, two days later, with the tracking link that was in the courier portal the whole time.
Four jobs pay back faster than a new ad tool. None of them are glamorous. All of them are sitting in the ecommerce work we actually scope.
1. Where is my order?
This is the highest-volume email in almost every small store I've looked at. The answer is usually in the system. The delay is a person copying it out. A decent setup reads the order number, fetches the courier status, and replies with the date and the link. If the status is "exception," "failed delivery," or anything that isn't a clean in-transit, it doesn't guess. It hands the thread to a person.
That last rule is the whole design. Automate the boring 80 percent of status mail. Keep the angry and the ambiguous.
2. The cart that went quiet
A single reminder, a few hours later, in plain language, with the actual product name, beats a five-email sequence that gets more desperate. If you wouldn't send email five to a friend, don't let a model send it to a stranger who looked at a kettle.
Check two things before you switch it on. Does the reminder respect people who already bought? Does it stop if they reply with a question? A sequence that keeps selling after someone asked about a cracked delivery is how you get the review you deserve.
3. The packing-table questions
Size, colour, "does this fit a standard filter," "can you leave it at the gate." Half of these are already answered on the product page, which nobody found. The other half need a person who has held the product. Split them.
- If the answer is on the page, the reply can quote that page. Don't paraphrase your own spec into something slightly wrong
- If the answer isn't written down anywhere, write it down once, then let the replies use that sentence
- If it depends on a photo of the customer's setup, that's a person
Stores that skip the "write it down once" step end up with a bot that has opinions. Opinions about thread size are how returns happen.
4. The Monday trading note
Orders, refunds, which product suddenly started getting the "where's my order" mail, ad spend if you want it in the same place. One note, Monday morning, pulled from the store rather than rebuilt in a sheet. This is the job founders swear takes "just twenty minutes" and then lose the morning to.
| Job | Automate when | Stop and fetch a person when |
|---|---|---|
| Order status | The courier status is a normal scan | The parcel is stuck, damaged, or the address looks wrong |
| Quiet cart | One reminder, right product, no purchase since | They replied, or they bought a different variant |
| Product questions | The answer exists in your own words | It needs a photo, a measurement, or an apology |
| Monday note | The store admin is the source of truth | A number looks impossible. Don't let a summary hide a broken feed |
A shopping assistant that doesn't know today's stock will sell tomorrow's disappointment. Connect it to the catalogue or don't put it on the site.
Ads, after the inbox can cope
I'm not anti-advertising. I'm anti pouring traffic into a store that still answers delivery questions by hand on a Wednesday because Monday and Tuesday were dispatch. Fix the four jobs, even just the first one, and then look at spend. The method for choosing that first job is the same as any other business, written up in how to integrate AI when you don't know where to start.
If you want those four checked against your actual admin, rather than a generic Shopify sermon, book the discovery call and have an order number ready. A real order teaches more than a tour of the theme.
Questions people ask
Which platforms does this apply to?
Shopify, WooCommerce, and the custom builds that still run on email and a courier portal. The job matters more than the logo on the admin screen.
Should the first project be an AI shopping assistant on the site?
Usually no. A chatbot that doesn't know your stock, your shipping cut-off, or your returns rule will invent all three. Start with order status and the inbox you already drown in.
Will customers hate automated replies?
They hate waiting three days for 'where's my parcel?' They don't hate a correct tracking link in twenty minutes. They hate a cheerful wrong answer.
What about the ads?
If the site converts and the inbox keeps up, then spend on traffic. If half the buyers email you because the delivery page is vague, more ads buy you more of that inbox.