Wayfair Taps OpenAI for Catalog and Support Overhaul

Wayfair is betting big on generative AI, embedding OpenAI models into its core catalog management and supplier support systems. This move, scaling from small-scale tests in 2024, aims to automate complex retail workflows across its vast product inventory.

The integration prioritizes two critical areas: improving the accuracy of tens of thousands of product attributes across a catalog of roughly 30 million items and optimizing supplier support workflows. This strategic approach moves beyond experimental AI applications to integrate the technology directly into daily operations.

Solving Catalog Quality at Scale

Ensuring consistent and accurate product tagging, like color, material, and size, is crucial for Wayfair's search, recommendations, and merchandising. Jessica D'Arcy, Associate Director of Catalog Merchandising, emphasized that data quality builds customer trust and reduces costly returns from misrepresented products.

Previously, catalog accuracy relied heavily on manual supplier and customer feedback, a process that couldn't keep pace with the sheer volume of products. While custom AI models showed promise, their expense and maintenance proved unsustainable for Wayfair's extensive tagging needs.

Building a Reusable AI Architecture

To overcome the limitations of bespoke models, Wayfair developed a tag-agnostic system powered by a single OpenAI model. A "definition agent" interprets contextual meaning for each tag, addressing the bottleneck of human time spent on defining tag semantics.

This contextual data, combined with aggregated product information, feeds into a framework capable of classifying attributes across diverse product categories. The team is now expanding model coverage for new attributes at a significantly accelerated rate, with the system already processing over a million products.

Early A/B tests indicate that improved attribute completeness directly impacts customer discovery, showing increases in impressions, clicks, and page rank. Wayfair, however, maintains human oversight, using structured audits and supplier validation to ensure data accuracy before automated updates, especially for high-risk tags.

Rethinking Supplier Support Workflows with Wilma

Managing support requests from tens of thousands of suppliers is a complex task. Historically, Wayfair associates manually reviewed, identified, and routed every incoming ticket, a process prone to errors and delays.

Wayfair augmented its supplier support product, Wilma, with agentic AI features, starting with ticket triage powered by an OpenAI model. This system reads incoming requests, adds missing context, and routes tickets to the appropriate team, moving from prototype to production in about a month.

Wilma provides associates with significant leverage, identifying ticket intent, enriching data from internal databases, and directing issues efficiently. Beyond triage, the company has deployed a dozen other AI-driven workflows, including co-pilots that assist human agents by analyzing case history and suggesting responses.

These AI assistants are trained on historical data to learn successful resolution patterns. The broader visibility provided by these models helps synthesize context across the entire supplier interaction lifecycle, leading to higher customer and supplier satisfaction.

Wayfair employs a staged rollout, shifting from assistive "co-pilot" to semi-autonomous "autopilot" modes once AI recommendation alignment with human decisions reaches a predetermined threshold. This approach ensures quality control and builds trust during implementation.

Measurable Impact Across Teams

The integration of OpenAI models has yielded tangible results. Wayfair has corrected 2.5 million product tags across its most visible products, with plans to quadruple this impact soon.

In supplier support, AI systems have automated 41,000 tickets per month, boosting throughput by up to 70% in some workflows and significantly reducing resolution times. This operational efficiency also leads to higher supplier satisfaction and fewer ticket re-opens.

Beyond these core integrations, Wayfair has deployed over 1,200 ChatGPT Enterprise seats across its workforce, facilitating ad hoc tasks and internal problem-solving.

A Long-Term Partnership with OpenAI

Wayfair's investment in machine learning and AI partnerships is long-standing. Advances in frontier models, particularly multimodal systems, are now expanding the scope of what Wayfair can achieve in the visually driven home retail sector.

Fiona Tan, Wayfair's CTO, highlighted the value of OpenAI's thought partnership, extending beyond model access to collaborative use-case development and rapid deployment. This strategic relationship is crucial as customer expectations evolve and they increasingly seek AI-driven shopping experiences.

The company is building for a future where AI is seamlessly integrated into the customer journey, from browsing and comparison to support and conversational interfaces, augmenting human expertise while scaling internal capabilities.