To build the AI Storefront, we need to first fix ecommerce’s data problem

The past two decades of e-commerce have seen steady growth, driven by shifts in consumer behavior, innovations in logistics, payments, and advertising, and accelerated by platform shifts to mobile and the internet.

AI is the next generational platform shift, and potentially the biggest: It promises unlocking entirely new ways to discover, evaluate, and purchase products, fundamentally better shopping experiences, and better economics for brands and retailers.But none of this happens without the right foundation.

We’re building the infrastructure to make it possible.

The Data Problem

AI needs to be grounded in reality. But the reality of e-commerce data is messy.

The data that power e-commerce lack a standard structure or a shared ontology. Every retailer operates on a bespoke stack: there is no common language that vendors use to describe products or users. Instead, every commerce site is a snowflake (Yes, even on Shopify).

I’ve seen this firsthand. I spent a decade as cofounder and CTO at Bluecore, building and scaling a unicorn ecommerce marketing automation platform.

We had an entire engineering unit whose job was to go retailer by retailer, writing custom integration to ingest, normalize, and maintain an understanding of each client’s product data. Every catalog was structured differently. Every integration was custom. It worked, but it scaled linearly at best – each new client required more engineers and more maintenance..

AI Can Transform Shopping – But First, The Data Must Be Fixed

AI infrastructure for commerce starts with clean, structured, and richly annotated data. We built that first.

Our system ingests, structures, and enriches e-commerce data into a highly detailed common taxonomy. It is powered by agents that continuously learn and improve, requiring no lift from brands or retailers and zero marginal human labor is required.

We started with apparel because it is one of the most complex categories in commerce. It requires understanding visual variants, sizes and measurements, fabrics, silhouettes, and closures. You can’t buy clothes from a spec sheet.

Cosimo

Cosimo is the first thing we built on top of it: a retail discovery experience that’s live today.

What’s Next

AI traffic to retail sites grew 693% year over year. The shift is already underway.

We’re building the tools that make our infrastructure available to brands, retailers, and AI agents, preparing them for a world where agents and users are both accessing the primary interface to commerce.

For now: try Cosimo and let us know what you think.

We’re adding brands every week. If yours should be one of them, get in touch.

— Mahmoud

You can subscribe here to stay tuned for updates. Our team is mostly engineers who don’t love writing, so we promise not to spam.