One sentence to look something up; one sentence to reprice, publish or fulfil; and it watches your stores on a schedule, speaking up only when something is wrong.
Chat and action log
Interface mock — store names and figures are placeholders
Not a chat window that only talks: it reads the real data in your stores, and it changes what is actually selling on the marketplace.
Whether store setup is complete, where a product is stuck, how far a job has run, how many orders a store took today, how many are still unshipped, how many credits are left — ask once and it goes and checks, instead of you clicking through menus.
Data comes from the live business data in your account
Reprice (by store, site or variant), edit inventory, publish or unpublish, approve reviews, fulfil orders, or start a listing run — and it lists exactly what will change before it changes it.
Anything costing credits or touching a marketplace gets a confirmation card
Schedule checks over the stores you pick: stock-outs, failed jobs, overdue shipments, expired authorisation. You only hear from it when something is actually wrong.
Alerts on exceptions only; never reprices unattended
The AI Assistant has hands, but there is a gate on them. For these four it prepares, and waits:
Starting a listing run or re-running a batch spends credits — you see the estimate first.
Repricing, publish/unpublish and fulfilment change what buyers see and can buy.
Dozens or hundreds of products at once: the impact is spelled out — which stores, which variants, which values.
Deleting, archiving or voiding — the confirmation card states the consequence.
Same sentence, different result: a general assistant gives advice, the AI Assistant gives you an outcome — because it can read and act on what is inside your account.
Every figure comes from real data in your account; if it cannot find something it says so.
It runs as the signed-in account: a member can only do what a member can do, and someone else's login sees someone else's data.
Which turn, which tool, which products changed — all reviewable after the fact.
The entry point is always in the bottom-right corner — no extra site, no extra client.
“Which orders haven't shipped today?” “Where is P88 in review?” — just ask.
Before acting, it lists the scope and the result; adjust the conditions, or decline.
On approval it calls the right interfaces and reports back. Scheduled checks run on the time and stores you set once.
The assistant built into the product: chat with it to look up stores, products, reviews, jobs, orders and credits, and to edit prices, inventory, publish state, approvals and fulfilment on your instruction; it can also run scheduled store checks.
No. Repricing, publish/unpublish, fulfilment and starting a job all go through a confirmation card that states which stores, which variants and which values — nothing runs until you approve, and a declined action is not retried.
Yes. It checks the stores and times you set and only alerts you on stock-outs, failed jobs, overdue shipments or expired authorisation; it never reprices unattended.
Chats are billed as tokens on your credit balance and appear as their own line (“AI assistant chat”) on the same invoice as everything else. If credits run low it tells you before starting rather than stopping halfway.
It looks things up before answering: every figure comes from real data in your account, and when something cannot be found it says so. It currently replies in Chinese.
Once you sign up it is in the bottom-right corner: ask a question whose answer you already know, then let it reprice one product and see what the confirmation card looks like.
