Wayfair boosts catalog accuracy and support speed with OpenAI

Wayfair has integrated OpenAI models into critical internal systems to improve supplier support and product catalog quality at scale, moving beyond small-scale experiments to full production. The company focused on complex areas like routing supplier support requests and improving tens of thousands of product attributes across its 30 million-item catalog, which is essential for customer trust and reducing returns. Previously, improving product tags relied on manual effort or expensive, bespoke AI models that didn't scale across 47,000 tags. Wayfair implemented a tag-agnostic system using a single OpenAI model with a "definition agent" to ingest context, enabling them to expand model coverage for new attributes 70 times faster than before. This has resulted in correcting 2.5 million product tags across the most visible products. In supplier support, an AI-augmented product named Wilma handles ticket triage, routing, and context filling, automating 41,000 tickets monthly and increasing throughput by up to 70% in some workflows. Wayfair uses a staged approach, shifting from assistive "co-pilot" to semi-autonomous "autopilot" modes based on human agent alignment rates, ensuring quality control. The retailer also deployed over 1,200 ChatGPT Enterprise seats for general employee use, viewing the partnership with OpenAI as strategic for tackling ambiguity in home retail where products are visual and subjective.

(Source：OpenAI)