From Runways to Code: How Technology Is Reshaping Fashion

Fashion-tech coverage is obsessed with the part of the industry that photographs well: AI-generated runway concepts, virtual try-on, generative design tools that can spit out a hundred silhouettes in an afternoon. I think the actual disruption in fashion is happening somewhere far less photogenic: the backend logistics that decide whether a garment makes it from a factory to a customer's door profitably at all.

The industry's oldest, most expensive problem isn't creative. It's sizing and returns. A meaningful share of online fashion revenue gets eaten by return logistics, driven largely by inconsistent sizing across brands and a customer's inability to know, before ordering, whether a garment will actually fit. That's a data problem, not a design problem, and it's exactly the kind of unglamorous, structural inefficiency that AI is well suited to actually solve (using real fit and return data to predict sizing accuracy per brand, per body type, before the order ships) rather than the flashier problem of generating a hundred new silhouettes nobody asked for.

A brand that cuts its return rate by using better sizing data is making a bigger structural change to its business than a brand that used AI to generate its lookbook.

The generative design tools get more coverage because they're easy to demo and easy to be impressed by in thirty seconds. Reducing a return rate by using better fit-prediction data is a genuinely bigger commercial unlock: it improves margin on every single transaction, not just the ones involving a viral AI-generated piece, but it doesn't produce an image anyone wants to share, so it gets a fraction of the attention in coverage of “AI in fashion.”

The same pattern shows up in supply chain forecasting. Fashion has historically overproduced because demand forecasting per SKU, per region, per size has been genuinely hard, and the cost of guessing wrong (unsold inventory, markdown cycles, waste) is enormous and baked into how the whole industry prices things. Better forecasting models attacking that specific problem are less exciting to write about than a generative design tool, and they're worth more to the business running them, because they touch the industry's actual largest cost centre instead of its most visible one.

My honest view is that the runway-facing AI tools are a distraction from where the real reshaping is happening. The brands that end up structurally advantaged by technology in this industry over the next decade won't be the ones with the most eye-catching generative design pipeline. They'll be the ones that quietly fixed sizing prediction and demand forecasting well enough to run a fundamentally more efficient business than competitors still guessing at both, and that work, correctly, will never make it onto a runway.