Deutsch
English
We are our own first customer.
Kurd Art Collective is our founder’s art print shop — and the proving ground for Alpheris. The agent has been running there since July 2026: every feature you see in Alpheris first had to survive this real shop with its real edge cases. And because our product is “a chatbot that invents nothing”, this case study invents nothing either.
A real shop, with real edge cases.
Kurd Art Collective sells art prints on Shopify — and brings exactly the awkward cases where generic bots fail: pre-sale items whose shipping starts weeks later. Gift cards. Customers writing in German and English. A catalogue that keeps changing.
- ✓A Shopify shop with 659 products and 93 collections — read in fully automatically, nothing maintained by hand
- ✓76 shipping, returns and policy blocks taken straight from the shop pages
- ✓21 FAQs the crawler recognised in the content on its own
What Alpheris read from the shop.
On connect, Alpheris crawls the shop and builds the agent’s knowledge base from it — nobody types up product data. This is the current state for kurdartcollective.com:
| Content type | Examples | Blocks |
|---|---|---|
| Products | Prints, editions, pre-sale items, gift cards | 659 |
| Collections | Collections and category pages | 93 |
| Shipping & returns | Delivery times, return windows, payment methods | 76 |
| FAQs | recognised automatically in the shop content | 21 |
| Info pages & blog | About, contact, background stories | 18 |
Read directly from the Alpheris database · as of 2 September 2026
Three things only a real shop teaches.
Anyone can answer test questions. The cases that made Alpheris better came from day-to-day operation:
The pre-sale that isn’t in the shipping table
“When will my poster arrive?” has two answers for pre-sale items: the normal delivery time — and the shipping start date, which only lives on the product itself. A generic bot quotes the shipping table and gets it wrong. Alpheris answers from the specific product, because both live in the knowledge base and every answer is checked against its source.
Admitting instead of guessing — and learning from it
37 customer questions could not be safely backed by the shop’s data. Instead of improvising, the agent collected them: they land on the knowledge-gap board in the dashboard, the owner answers each one once — and from then on the agent can. The knowledge base grows exactly where real customers ask.
From product advice to a filled checkout
When a chat visitor decides on a print, they don’t get a “have a look around the shop” reply — they get a cart link with the right item in the right variant. One click, and the checkout is filled. Built and first put into production on kurdartcollective.com.
Why this case study looks different from others
There is no “25% revenue increase” here and no glowing customer quote. Kurd Art Collective belongs to our founder — it is our pilot and stress-test shop, and that is exactly what we call it.
Every number on this page comes straight from the Alpheris database (as of 2 September 2026). We build a product whose core promise is to invent nothing — that applies to our marketing too. If stories of other shops ever appear here, the same rules apply: only backed numbers, only real voices.
The same agent, on your shop.
The live demo crawls your shop and answers real questions from your real data — within minutes, no sign-up needed.