Wayfair boosts catalog accuracy and support speed with OpenAI

Wayfair has never had a small catalog problem. Millions of SKUs, each with dozens of attributes like color, material, dimensions, and style, all fed in by different vendors with different habits of describing the same beige sectional sofa. Keeping that data clean at scale has always been more of a labor problem than a technology one, which is exactly the kind of grunt work large language models turn out to be good at.

According to OpenAI, Wayfair is now running its models against that catalog to catch and fix attribute errors automatically, rather than relying purely on manual review or brittle rule-based scripts. The company is also using the models to triage incoming customer support tickets, sorting and routing issues so human agents spend less time reading and more time solving.

Neither of these is a flashy consumer-facing AI feature. There is no chatbot avatar, no flashy demo. It is quieter than that: infrastructure work aimed at making an enormous retail operation slightly less messy behind the scenes. But that is arguably where a lot of the real near-term value of these models actually lives, in unglamorous cleanup jobs that used to eat thousands of human-hours.

For a company the size of Wayfair, even modest gains in ticket resolution speed or catalog accuracy compound fast. Fewer mislabeled products mean fewer returns and fewer confused customers. Faster ticket triage means support teams aren't buried under a queue that grows every time a couch shows up dented. It's not the kind of story that makes headlines, but it's the kind that shows up in quarterly numbers.