The Stitch Fix story
Stitch Fix needed to recommend clothing to millions of customers without losing the personal touch that made the service work. Pure algorithmic recommendations felt cold; pure human styling did not scale and could not stay inside the price point.
Models could narrow a giant catalog quickly but missed nuance — a recent move, a wedding next month, a customer who said 'no patterns' three months ago. Humans caught the nuance but could not browse a million SKUs.
They built a hybrid. Algorithms narrowed the catalog to a personalized shortlist for each customer, and a human stylist made the final pick with context the model could not see. Speed from machines, judgment from people.
- Models surface candidates from a large catalog based on customer signals
- Stylists apply judgment the model cannot see in the data
- Customer feedback feeds back into both the recommendations and the stylist notes
- Outcomes tracked end-to-end, not just at the click
- Stylist time reserved for the decisions that actually need a human
Stitch Fix scaled personalized styling far beyond what either pure-AI or pure-human models could have reached, and the pattern became a reference for AI-plus-expert services.
AI plus human judgment beats either alone, when the work is split so each does what it is actually good at.