Paul Toms, Wayfair’s CMO, told Modern Retail he approaches AI by “not being kind of dogmatic about things, but being curious and being honest with ourselves about how people and how users and potential customers are receiving things.” It is a short sentence doing significant work.

The retail industry has spent two years sorting itself into camps. Believers mandate transformation; skeptics call it overhyped infrastructure spend. Toms refused both. Where AI demonstrably works at Wayfair — the Discover tab, creative iteration — he deploys it. Where it fails, he stops. His customers told him when to stop: early versions of the Discover tab served AI-generated room imagery filled with unshoppable products, and users called it “AI slop.” He rebuilt it.

This is a more demanding position than it sounds. “Not dogmatic” without measurement infrastructure is indifference.

The broader market makes Toms look prescient. BCG’s “The Widening AI Value Gap” (September 2025), across 1,250 companies, as reported by Master of Code, found only 5% achieve substantial AI value at scale; IBM’s 2025 survey of 2,000 CEOs, cited in the same analysis, found only 25% of AI initiatives delivered expected ROI. These numbers don’t indict AI. They indict the brands that have made AI transformation a core identity claim — whose earnings calls are structured around AI capability narratives — before the evidence arrived to support them.

Toms also said retail is “very hesitant” and he’d like more ambition. That isn’t a contradiction. Non-dogmatic is not the same as non-committal. The difference is measurement.

When the ROI questions arrive in earnest — BCG’s data suggests most firms are still waiting — belief won’t close them. Toms built the infrastructure to know whether it’s working. Most haven’t.