Most website builders — Shopify, Wix, Webflow — still follow the same basic interaction model: pick a template, drag some blocks around, manually fill in every section. It works, but it assumes the user already knows what a "good" store looks like.
I wanted to try a different model entirely: skip the blank page completely. Enter a business name and pick an industry — two fields, no free-text prompt, no lengthy form — and let AI generate a complete, working store — copy, products, theme, checkout — before the user makes a single design decision.
That's what I'm building with Storeaxe. Here's a look at how it actually works under the hood.
1- Merchant enters: business name + industry (two fields)
2- AI infers tone/style from the industry (no free-text to parse)
3- AI generates: homepage copy + sample product catalog
4- Theme matched from industry/tone (rule-based, not AI) 5- Merchant reviews / edits everything
6- Store published — live, real checkout, real orders
Keeping the input to two structured fields instead of a free-text prompt was a deliberate choice, not a limitation. A short open-ended prompt sounds more "AI-native," but it also introduces parsing risk — vague answers, missing info, inconsistent tone extraction. Two fixed fields are trivial to validate, faster to fill in on mobile, and give the downstream generation step reliably clean input to work with. The AI still does the interesting work — copy, catalog, tone inference — it just isn't asked to do information extraction on top of generation.
The key design decision beyond that: not every step needs AI. Generating copy and products benefits from a language model. Picking a theme doesn't — that's a straightforward tag-matching problem, and doing it with rules instead of a model call is faster, cheaper, and more predictable.
Card payment gateways assume a level of infrastructure and business registration that a lot of small, first-time sellers don't have yet — especially outside a handful of countries Stripe directly supports. Rather than forcing merchants through a payment gateway approval process before they can sell anything, Cash on Delivery is the default, zero-setup payment method. Card payments via Stripe/PayPal are opt-in, connected later if and when a merchant wants them.
This turned out to matter more for real-world adoption than any AI feature did.
The live product is at ** storeaxe.com** if you want to see the actual generated output. Happy to answer questions about the architecture, the AI pipeline, or the payment-method tradeoffs in the comments.