Wayfair Scales Retail Operations Using OpenAI Agentic Workflows
- •Wayfair automates 41,000 supplier support tickets monthly using OpenAI-powered agentic workflows.
- •Retailer scales product attribute tagging 70x faster, correcting 2.5 million tags across its catalog.
- •Wayfair implements Wilma, an agentic system that transitions support tasks from human co-pilot to autonomous autopilot.
Wayfair is moving beyond isolated AI experiments to embed OpenAI’s frontier models directly into the core of its global retail infrastructure. The furniture giant is tackling two of its most significant scaling challenges: maintaining data integrity across a catalog of 30 million items and managing a massive influx of complex supplier support requests.
To solve the catalog bottleneck, Wayfair developed a "tag-agnostic" architecture featuring a specialized definition agent. This system synthesizes information from the web and internal databases to encode the contextual meaning of product attributes like material and style. This allows the company to deploy new attributes at 70 times the speed of previous methods, correcting 2.5 million tags to date.
On the operational side, an agentic AI named Wilma has transformed supplier interactions by moving from simple ticket triage to semi-autonomous resolution. The system tracks an "alignment rate"—the frequency with which AI suggestions match human decisions—allowing high-performing workflows to transition from a "co-pilot" mode to an "autopilot" state.
With over 1,200 employees using ChatGPT Enterprise, Wayfair is positioning itself for a future where multimodal systems—AI that understands both text and visual data—help shoppers find products that are difficult to describe with simple keywords. This deployment illustrates how large enterprises are moving from generative novelty to tangible operational efficiency.