Out of everything you did as a designer at Hermès, what has AI replaced? Nothing completely, in her view. The creative process still starts the same way: a point of view, creative direction, drawing on white paper. What changed is everything around it. Photo editing is automated, and research that used to mean days in a library now takes minutes. "It's just the same, but enhanced."
This year's luxury "AI slop" campaigns. What's actually going on? Some of it may be intentional. Gucci's DNA has always been provocation, so an uncanny film that creates a reaction is not off-brand, and "I think they got what they wanted." For the rest, it's a language problem. Content is the language between a brand and its customer, customers are human, and they respond to imperfection. She points to wabi-sabi, the Japanese idea of beauty in imperfection: push too far into the hyper-digital and the brand loses the language entirely.
Which brands are getting AI right? The ones using it inside the iteration process rather than end to end. They start with a unique point of view and their brand DNA, use AI to visualize and iterate, and a human refines the result until it aligns with who they are. AI is one piece of the process, never A to Z.
For a brand on a marketplace, what does scaling imagery actually involve? It starts with defining the brand DNA and translating it into a design language a model can ingest. With the right inputs and preset workflows, bulk generation at scale works. But the model has to be trained on who the brand is, tagged correctly, reinforced, and refreshed. Train it on generic data or on who the brand was five years ago and you get what she calls creative entropy: a closed loop that drifts further from the brand with every generation.
What's the most common mistake brands make with AI imagery? Letting the tool drive. "We all have a tendency to let the tool define what we create, instead of telling the tool what we want to create." Output gets shipped because it looks impressive, not because it matches the intention or the product. On a marketplace that's expensive: if the image doesn't match what arrives in the box, the customer's trust breaks and the returns cycle starts.
AI is collapsing the production gap between small and big brands. Who wins? Launching a brand ten years ago meant hiring a technical designer, a marketer, a social media manager. Now one person does it at scale with half the budget, tests demand before producing, and sends the factory an exact rendering of the finished product instead of a description. And she's betting on a counter-trend that favors small brands: the return of craftsmanship, tactility and the human touch, with AI running quietly underneath to speed it up.
Should AI imagery be labelled? Smart regulation over blockage regulation. For news, yes: misinformation is the real risk. For fashion and creative work, she believes in internal labels for traceability, but not public ones on product imagery, provided the output is accurate to the product and models are contracted and paid properly for the use of their image. Long term, she thinks AI ends up like the pen or Photoshop: part of the everyday process, not a special category.
