Generative AI in design studios: ideas come easy, products don't

Clothing design is no longer just drawing. In studios, image generators prepare colour variants, suggest cuts and build visualisations that a few years ago required a photo shoot. The question of how much of the designer is left in that process came up again at the 48th WorldSkills in Shanghai. Chinese woodworking experts there said plainly that artificial intelligence can throw up good ideas and will replace some designers. The problem starts at implementation. Structures with many bends, bevels and sloped planes still need a human hand and eye. A garment's cut works the same way. A model will generate a silhouette, but fitting it to fabric, seams and a body remains craftwork.

A tool, not a decision

Red Dot runs one of the best known design competitions. In an interview the organiser published, the design heads of the Haier group describe AI as another tool in the chain of decisions, not an author. They are William Wu, vice president for design, and Federico Ferretti, who leads the Haier Europe design centre in Milan. Red Dot competitions run in three separate categories: product design, brands and visual communication, and design concept. Each judges the result of human work, not skill with a prompt.

The EU adds duties for designers

EU legislation creates a new context for studios. Directive (EU) 2024/825 on empowering consumers for the green transition entered into force at the start of 2024. Member states had until 27 March 2026 to transpose it, and full enforcement begins on 27 September 2026. In practice this bans hidden design tricks that shorten a product's life: glued housings and batteries that cannot be replaced. The scale of the problem has been counted. Premature disposal of goods accounts for 261 million tonnes of CO2 emissions a year in the Union, and the waste it generates reaches 35 million tonnes. For a clothing designer this is a new brief: instead of designing for one season, design for the next decade.

Who answers for a mistake

Generative tools bring a question of liability. If a generator reproduces an element of someone else's cut, the company is liable exactly as if a person had done it. The company puts the product on the market, not the model. Design therefore defends itself not with image rights but with a documented process: notes, variants, rejected samples and decisions that can be pointed to. Those determine whether a design is original, not how many seconds an image ran on a server.