AI for Fashion

I recently joined a Fashion AI startup and have done a lot of deep dives into the conversation on the topic.

The vibes keep centering on generation. Feed a model some mood board images, get content or an avatar back. It’s interesting, but having worked in GenAI, I don’t know if it’s the future of what people care about. The output is directional, but is the utility functional for people who have to put real clothes on real bodies? Proportions that don't account for how fabric drapes. Colorways that look plausible on screen but would never work in production. It's the uncanny valley of taste. We might eventually get there. But right now I think the value lies elsewhere.

To me, the real opportunity is upstream. Curation, not creation.

This is what great buyers have always done for brands. They synthesize a massive amount of signal: what sold last season, what's moving on resale, what's showing up in street style, what their customer is about to want before she knows she wants it. That entire job is pattern recognition across messy, unstructured data. That's exactly what language models are good at.

The interesting question isn't "can AI imagine a piece of clothing on a body?” It's "can AI look at 50,000 SKUs across 200 brands and surface the 40 that belong in front of a specific customer right now." Not a recommendation engine. Those are backward-looking. This is more like giving every shopper a buyer who knows their closet, their calendar, and their taste graph.

Fashion has always been a taste business. AI doesn't replace taste. But it can make a small team with great taste operate at a scale that used to require a department store's worth of buyers.

The companies that win this won't be the ones generating images. They'll be the ones that figured out how to encode taste into a system and let it compound.