E-commerce
The marketing side of your shop, prepared before you get to it
You already sell, and you already spend to grow. What eats the week is everything around it: working out which spend brought the sale, filling in the same product in every language and channel, keeping campaigns in step with the catalogue, and answering the same questions in every inbox. We automate that on the tools you already run, and a person approves anything that reaches a customer or moves money.
Measurement & attribution
Google Analytics says one thing, Meta another and the shop a third, so nobody can say which spend brought the sale. We set up PostHog to count orders from the shop's server, send purchases back to Meta and Google so they bid on real sales, and set ad spend against real orders. Every week an AI read of the numbers is drafted through the PostHog MCP, ready before Monday. We are building this on our own brand first, and the case shows how far it is.
Human in the loop
The weekly read is a draft, not a decision. Nothing in your campaigns changes until a person has checked what it says.
Multilingual product launches
A new product means the same work in every place it appears: the shop listing, the translations, the Google Shopping feed and the ad copy, each filled in by hand and each a chance for a price or a size to disagree. Now that the EU bans environmental claims a brand can't back up, a claim that slips through a translation is a legal problem too. We turn one product sheet into all of it, checked for language and for claims you can't prove, and nothing goes live until you approve it. We are building this on our own brand first.
Human in the loop
Nothing goes live until a person approves it. The system flags a claim it can't find on your list; it never decides what is true.
Feed-driven advertising
Amogh set this up for Dutch retailers with 1,000+ products before founding Flowstate Works: one feed running Shopping, Search and DV360, so the campaigns followed the shop instead of a spreadsheet. The platforms now write and optimise from the feed themselves, so a wrong attribute reaches every channel at once. What we build today is the check on the feed before it goes out.
Human in the loop
Budgets, margins and the rules that decide what gets advertised stay yours. Changes to the feed are checked and approved before any channel sees them.
Customer-service agent
Where is my order, how do I return this, does it run small, can I swap the size: the same questions, all day, through email, the chat on your site, WhatsApp and Instagram. The agent answers the simple ones itself, from the order and tracking data, in the customer's language. Returns, exchanges and refunds it prepares with the order, the policy for that market and the customer's rights already checked, and they wait for a person. It builds on the support-email agent we built for a client; the shop version runs on our own brand first.
Human in the loop
Simple answers from your own data go out on their own. Anything involving money, stock or an exception is prepared with the full context and waits for a person to approve it.
E-commerce support agent
One agent for email, chat, WhatsApp and Instagram answers order and FAQ questions on its own, and prepares returns, exchanges and refunds for a person to approve.
per week on the email agent it builds on
AI customer-support email agent
A retrieval-augmented agent reads support emails, pulls the customer record and the right FAQ answers, and drafts an on-brand reply for one-click review — response times drop from days to minutes.
of support time saved per week
Find out which of these your shop needs first
The audit is where that gets decided, and the services page sets out what it includes. If you'd rather talk it through first, the intro call is free.