06 Services
AI for e-commerce that knows the order, not a script.
Storefront chat, WhatsApp, and email that can answer size, stock, and 'where is my order' from your catalog and policies. Scripted demos are not a live customer system. The real build is wired to the shop, with a person when it should be.
Problem
Support volume scales. Headcount should not have to.
A growing catalogue produces the same three tickets: Do you have it in my size? Where is order 4821? I want a return. Those are not strategy questions. They are lookups. Humans should not be the database.
Most store chat widgets fail in the same place. They cannot see live stock or the fulfilment status, so they apologise and offer a form. Or they invent a delivery date. Either way, the buyer writes again, or they charge back.
E-commerce automation is worth it when the agent reads the same sources your team uses: Shopify (or equivalent), shipping, and the return policy you actually honour. And when 'I need to speak to someone' is a real path, not a buried menu.
- Size, colour, and stock questions repeating all day
- Where-is-my-order tickets the warehouse already knows
- Return and exchange policy retyped by whoever is on chat
- WhatsApp and Instagram DMs that never reach the helpdesk
Who
Who this is for.
Stores whose bottleneck is size, stock, and where-is-my-order — not a merchandising bot.
-
01
The founder is still on chat
You are answering “where is my order” between packing. The warehouse already knows. You should not be the database.
-
02
A catalogue that outgrew the inbox
Size, colour, and stock questions repeat all day. Variants live in Shopify. Replies still live in a person.
-
03
DMs that never reach support
Buyers write on WhatsApp and Instagram. Those threads never hit the helpdesk. Broadcast is not a ticket queue.
-
04
A widget that cannot see the order
The chat apologises and offers a form. Or it invents a delivery date. Either way the buyer writes again.
Examples
What this looks like in practice.
Lookups from your catalog and orders. Not a widget that apologises and offers a form.
-
Where is order 4821
Status from Shopify and the tracking you already use. Not a guessed date. If fulfilment is messy, it hands off.
-
Size and stock from the catalog
Navy jacket, size M: looked up, not remembered. If the variant is gone, it says so. No invented stock.
-
Returns that follow the policy
The window and reasons you published. The agent collects the order and the reason. Refunds still hit a person.
-
WhatsApp as a support channel
Same catalog and order data as site chat. Instagram only when you already treat those DMs as support, not as a broadcast.
Process
How the work runs.
We start with the ticket type that burns the most hours, not a bot that claims to do merchandising.
-
01
Pull the ticket mix
Order status, stock, returns, or something else. We pick one as the first live path.
-
02
Ground it in the store
Catalog, variants, policy pages, and order data. The agent does not invent a navy jacket you do not have.
-
03
Connect the channels
Site chat, WhatsApp, or email, wired to Shopify, shipping, and the inbox a human already watches.
-
04
Launch with escalation
Live on one channel. Watch the misses. A person takes over for refunds, edge cases, and anyone who asks.
Integrations
Wired to what you already run.
We plug in. The workflow does not wait on a new stack.
- Shopify
- Site chat
- Order / tracking APIs
- Returns portals
- Helpdesk
- GA4 / ads pixels (sites)
Outcomes
What changes when it ships.
-
01
Fewer 'where's my order' tickets on the human queue
Status comes from the same system the warehouse uses.
-
02
Stock answers that match the storefront
Variants and availability are looked up, not remembered.
-
03
Policy that does not depend on who is online
Returns and exchanges follow the document you published.
-
04
A human for the cases that move money
Refunds, damage, and angry buyers are not left to the model.
Enquire
Tell us where the friction is.
Write where the work is stuck. I read it and reply. No build commitment.