About 80% of customer service, handled by AI.

A print-on-demand fashion brand where every item is made to order in a factory. Most customer emails ask the same few things: where is my order, can I change the size, when will it ship. I built an app that reads every email, looks up the real order, and leaves a ready reply inside Gmail.

Book a free AI audit call →AI automation services

Built for a brand I run. A person still reads and sends every reply.

Tal Mor, AI Customer Service Automation Case Study

What the app does in Gmail

Drafts repliesTagsFlagsArchives

~80%of customer emails come with a reply drafted from real order data

1 to 2full-time customer service reps no longer needed

Built and running. Happy to show it on a call.

~80%

Of customer service handled by the app

1 to 2

Full-time customer service reps replaced

Drafts only

A person still sends every reply

The business

A fashion brand selling all-over-print hoodies, t-shirts and dresses designed by independent artists. Every item is printed and sewn to order by a factory, so each order passes through several hands before it ships.

That model creates a lot of email. Customers want to know where their order is, and the honest answer lives in three places: the store, the factory's production spreadsheet and the shipping tracking.

Industry	Print-on-demand fashion
Store platform	WooCommerce
Inbox	Gmail
Data the app reads	Store orders, the factory production sheet, shipping tracking
Built with	Claude, Supabase, Gmail and WooCommerce APIs
Status	Live, used every day

The problem

The same questions, all day. Where is my order, when will it ship, can I change the size. Each one took a few minutes of digging, and the digging was the expensive part.
The answer lived in three places. Order details in WooCommerce, production status in the factory sheet, tracking with the shipping hub. Nobody holds all three in their head.
Urgent emails got buried. A complaint or a refund request looked the same as a thank-you note in a busy inbox.
Nothing got archived. The inbox was the to-do list, so it never got shorter.

What I built

Order lookup on every email. The app matches the email to the customer's order and pulls the order details, the production status from the factory sheet and the tracking.
A draft reply inside the Gmail thread. Written in the brand's voice with the real facts filled in, so answering an email means reading it and pressing send.
Tags by topic. Shipping status, order changes, returns, complaints and everything else get their own label, so the inbox sorts itself.
Flags for anything that needs me. Money, an angry customer or a deadline goes to the top, flagged and never auto-handled.
Archive for the noise. Notifications and emails that need no answer get labeled and archived.
Its own order database. Orders are imported into a database the app owns, so lookups are instant and the daily order run reads the same facts.

How it runs

Step	What happens
01. Email arrives	The app reads it and works out what the customer is asking.
02. Order match	It finds the order by email address, name or order number.
03. Status check	It reads production status from the factory sheet and the latest tracking.
04. Draft	It writes the reply inside the same Gmail thread and tags it.
05. Triage	Urgent threads are flagged, noise is archived.
06. Human send	A person reads the draft, edits if needed and sends.

What changed

It does the work of at least one full-time customer service rep, closer to two in busy months.
About 80% of customer emails now arrive with a reply ready to send, built from the real order data.
Answering an email takes seconds, because the lookup is already done.
The inbox shows what needs a human, not everything at once.
Customer service and fulfillment read the same order data, so the answers match what the factory is actually doing.

How an AI project with me runs

Free call

Sixty minutes on your business. I tell you what I would automate first and what I would leave alone.

Map the work

I learn how your business actually runs, then rank what to automate by hours saved and risk.

Build one thing

We start with the system that saves the most time, with a person approving anything that touches money.

Hand over or run it

You get documentation and training, or I keep running and improving it monthly.

Customer service was the job I wanted gone most. Now the app does the digging and I just read and send.

I run my own ecommerce brands, so every system I build has to hold up on a busy Monday, not just in a demo. I build the same kind of system for other businesses: I learn how yours works, decide what to automate, and build it.

20 years operatingAI architect for ecommerceUpwork Top Rated · 100% JSS

Questions about AI customer service for online stores

Straight answers, from a system that runs every day.

Can AI answer customer service emails for an online store?

Yes, and the safe way is drafts. The AI looks up the order and writes the reply inside your inbox, and a person sends it. It takes the repetitive majority and leaves the judgment calls to you.

Does it work with Shopify, Gmail or a helpdesk?

This one runs on WooCommerce and Gmail. Shopify, Gorgias, Zendesk and Help Scout all have APIs, so the same design works there.

How is this different from a chatbot?

A chatbot guesses from your FAQ page. This reads your actual order, production and tracking data before it writes a word, and it never sends on its own.

How long does it take to build?

It depends on where your order data lives and how many systems it has to read. I scope it on the free call, and the first version covers your most common questions.