E-Commerce Operations Agent

The Problem

A D2C brand with 30 years of manufacturing expertise was stuck in manual mode. They know their products better than anyone: material grades, finish types, weight tolerances, sizing conventions, which surface treatment works for which application. But that knowledge lived in the founder's head and a stack of spreadsheets. Product photography was done one-by-one. Marketplace listings were written manually for Amazon, Shopify, and two other platforms. Catalog updates took days. They couldn't scale because scaling meant hiring people who would take years to learn what the founder already knew.

What Made It Hard

The domain knowledge is the product. A material science expert knows that "304 stainless steel" and "18/8 stainless" are the same thing, that the finish affects corrosion resistance differently in coastal vs. inland environments, that certain dimensions require different manufacturing tolerances. An API call to Claude doesn't know any of this without the right context architecture.

Each marketplace has different listing requirements. Amazon SEO is keyword-density driven. Shopify is brand-story driven. The same product needs different descriptions, different image specs, different attribute formats. Manual approach: 45 minutes per listing per marketplace. With 200+ products across 4 platforms, that's 600+ hours of listing work.

The Architecture

Knowledge Encoding: The founder's 30 years of expertise structured into a queryable product knowledge base. Material properties, manufacturing constraints, application guidelines, competitive positioning. Not a flat database. A knowledge graph that understands relationships between materials, processes, and applications.

Image Pipeline: AI-generated product imagery from specifications and reference photos. Background removal, lifestyle context generation, marketplace-specific formatting. Each platform's image requirements (dimensions, background color, hero image positioning) handled automatically.

Listing Engine: Claude API with domain-specific context generates marketplace-optimized listings. Amazon gets keyword-dense bullet points. Shopify gets story-driven descriptions. Both get technically accurate specifications because the knowledge base enforces correctness.

Catalog Sync: Automated catalog management across 4 marketplaces. Price updates, inventory sync, new product rollouts. What took days now takes minutes.

The Outcome

Active deployment. 30 years of accumulated manufacturing expertise is now encoded in a system that generates technically accurate product content, optimized per marketplace, at a pace the manual process could never match. The founder reviews and approves rather than writes from scratch. The 8-to-80 multiplier in action: the 8 took 30 years of manufacturing experience. The 80 took 6 weeks of engineering.