The challenge
Orbital sold 60,000 products across four storefronts, most with thin supplier descriptions. Shoppers couldn’t find what they wanted, and the support team spent its days answering “where is my order?”.
What we built
A content pipeline that writes and enriches product copy in each brand’s voice, semantic search that understands intent (“a warm jacket for hiking in the rain”), and a support agent connected to order, returns and inventory systems that resolves common requests and hands everything else to a human with the context attached.
- Brand-voice product copy with human approval at scale
- Hybrid keyword and vector search
- Support agent with order-system tools and safe hand-off
- Kubernetes deployment with per-storefront scaling
The results
The support agent resolves 68% of tickets without a human, and semantic search lifted search-to-purchase conversion by 12%.