Connect with us

LIFESTYLE

Fashion AI Pays the Designers Who Built Taste First

AI moodboards speed veterans who already have an eye. Junior fashion creatives lose the slow work that used to build one.

Published

on

Fashion studios now build moodboards, pitch decks and shot lists with generative AI before a person has gathered their own references. The public fight is still about sloppy campaign images. The private split is about who still owns a first idea that did not come from a model.

Seniors who already trained an eye treat the tools as a visualization pass and pocket the time. Juniors, asked to move at the same calendar speed, start where the model starts.

Inside the Moodboard, Before the Idea

The fashion calendar has always punished slow research. Moodboards, agency decks, lighting references and rough visualizations now get a first pass from a chatbot because the next meeting is tomorrow and the sample factory is already booked. Creative agencies, photographers and some designers describe that sequence as standard, even as the same industry scolds consumer-facing AI imagery for looking lazy.

The risk is the order. When a suggestion arrives before a person has formed a private reference set, the suggestion becomes the reference. People who open the same class of model, with the same class of brief, begin in the same place.

Giorgia Lupi, a partner at Pentagram in New York, puts the damage on that first shortcut. Even building inspiration, she says, can now mean asking a chatbot for ideas for a brief before anyone sits down and thinks.

Who Cashes Fashion’s AI Speed Dividend

The people who cash that speed already have something to direct. McKinsey’s State of Fashion 2026 found that more than 35 percent of fashion executives were already using generative AI in online customer service, image creation, copywriting, consumer search or product discovery. That is an executive stack. It is not a sketchbook.

THE SPEED THE CALENDAR WILL PAY FOR

  • Executive use: More than 35 percent of fashion executives already deploy generative AI in listed commercial functions, per McKinsey’s State of Fashion 2026.
  • First-pass work: Moodboards, decks, shot lists and rough visualizations are where agencies now quietly plug the tools in.
  • Factory math: New York designer Jane Wade has said that visualizing a look before cutting it could save about $1,000 at the factory.

Google’s Envisioning Studio tested that math in the run-up to Spring 2027 New York Fashion Week, co-building a styling suite with Wade through its Flow studio and a runway visualization tool with Sergio Hudson. Hudson’s line on the old method was blunt: every mock sent has a cost. Wade’s studio had been hand-sketching styling decisions. The suite let her team try colorways, fabrics and accessories on images of real cast members from fittings, then still put every look on a person.

That is a veteran’s bargain. She already had a collection in cloth. The model helped her see variants of work she had already done.

What 54 Essay Writers Showed on EEG

The lab picture behind that bargain is not a fashion study, and the people who ran it keep saying so. In June 2025, Nataliya Kosmyna and colleagues at the MIT Media Lab posted a preprint, still not peer-reviewed at the time they published it, on 54 participants writing essays over four months. Eighteen of those 54 came back for a fourth session with the tools swapped. People were split into three groups: ChatGPT, a search engine, or their own heads. EEG tracked how regions of the brain coupled while they wrote.

Brain connectivity scaled down with outside help. The brain-only group showed the strongest, widest networks. Search sat in the middle. LLM help showed the weakest coupling. People in the LLM group reported low ownership of the essays and fell behind when asked to quote work they had written minutes earlier. Named entities, n-grams and topics also clustered inside each group. Kosmyna’s team called the pattern cognitive debt: fluency now, less effort later.

THE THREE GROUPS IN THE MIT ESSAY STUDY

Group Tool Brain pattern What they kept
Brain-only No tool Strongest, widest networks Highest ownership, better at quoting their own text
Search engine Web search Intermediate engagement Strong ownership, still less than brain-only
LLM ChatGPT Weakest overall coupling Low ownership, weaker recall of their own essays

In session 4, people who had used the model for three sessions and then wrote without it still showed weaker alpha and beta networks. People who had written unaided and then tried the model showed higher recall and re-engaged visual and prefrontal areas. Starting with your own draft changed what the tool did to you.

Kosmyna is blunt about effort. Mental effort, she says, is extremely important for learning, memory and creativity. Bypass the frustration of a new idea by turning to a model first, and you become a consumer of thinking that does not belong to you.

WHERE THE PAPER AND ITS RECAPS DIVERGE

  • The viral version: Recaps of the preprint push round numbers about ruined brains and treat a small essay task as a verdict on all thinking.
  • The lab’s own limits: The project page lists a small Boston-area student sample, a ChatGPT-only protocol, an educational essay task, and a request that journalists not use words such as dumb, harm or brain rot.
  • The read-the-paper view: John Gallagher, a professor at the University of Illinois Urbana-Champaign who assigned the full text, wrote that the results cannot be distilled into a summary or a tweet.

None of that makes the EEG traces go away. It does mean a fashion moodboard is not an SAT essay, and the honest use of the study is narrower: people who start with the model remember less of what they produced, and the output clusters.

Directors Keep a Gap the Model Cannot Close

Roger Beaty, who directs the Cognitive Neuroscience of Creativity Lab at Penn State, has been measuring the same split in the act of making. Memory, he says, is the raw material for ideas. You need to know what has been done in order to make something original, and you use episodic memory to test whether an idea will work and how other people will take it.

His team’s Creative Direction work scores, from chat transcripts, how much a person contributes their own ideas and how hard they push back on the model. Across lab sessions and direction reduced sameness across users, people who directed rather than handed the job over produced more original work with large language models, with a meta-analytic correlation of.28 (95 percent confidence interval.25 to.32), including 6,290 real chats. Direction also cut the so-called homogenization effect, the tendency for models to give different people the same kinds of ideas.

A sister pair of studies with 442 people found the uncomfortable career fact. People who wrote more original stories without a model also wrote better ones with GPT-4o. The tool did not close the gap. Beaty’s line in interviews is the practice version of that result: more you, less AI, do the creatively meaningful work yourself, then use the model for feedback you cannot get on demand.

Simone Luchini, a Penn State researcher who works with Beaty, tested the collaboration itself. In a pre-registered comparison of 68 human-human pairs and 72 human-AI pairs on GPT-4o, people working with the model generated fewer original ideas than human partners. The apparent AI edge was mostly extra words, not extra originality, and the humans on those teams produced weaker ideas of their own. Two people working together got better at original ideas as the session went on. Luchini’s summary is the one fashion should sit with: working with AI reduces the diversity of ideas and pulls them toward fewer common themes, while working with humans preserves that diversity.

The Missing Years of Looking

That is the junior problem, and it is not a vibe. The first years of a fashion job used to be slow looking. You pulled references by hand. You sat in fittings. You built a private index of why some ads stuck and others vanished. Lupi’s fear is that a generation starting now will not build their own way of seeing the world, or their own way of designing, because the shortcut is always on the desk.

I’m concerned about what younger designers who are starting out now will not be able to build, their own way of seeing the world and their own way of designing, because of the temptation to shortcut.

Giorgia Lupi, partner, Pentagram

Taste in this industry is not a moodboard export. Knowing that Tom Ford’s G-string for Gucci walked in Spring 1997, and then kept getting cited because it changed how luxury sold sex, is not a retrieval task. It is a stack of rooms, people, and later copies. A model can summon the images. It cannot tell you why they still matter in a room of editors.

WHAT THOSE EARLY YEARS USED TO TRAIN

  • Private references: Building a set of images, shows and ads you had actually sat with, not a board anyone with the same login could regenerate.
  • Context, not keywords: Knowing why a campaign stuck, who it was for, and how later houses borrowed it.
  • The slow make: Sketching, draping, fittings and the small failures that teach what cloth will do.

Working designers who publish their tool stacks still keep the hard steps analog. Flats stay by hand. The sample is the only score that counts. The model can fill a board. Someone still has to kill the looks that will not grade across sizes. That kill step is taste. If you never built the archive, you cannot do it except by agreeing with the model.

Lupi’s own counter-practice is almost anti-fashion. Two years ago she started making things for no reason, painting on weekends, sewing paper patterns, writing on tissue in handwriting she keeps hard to read, with no plan to show it. Making without a deadline, she says, is how you stay in a real relationship with your own instincts. Seniors can afford that. Juniors on a production calendar cannot.

Jane Wade Draws the Line at Generation

Wade is the clean case of a winner who knows where the line is. She calls herself a pretty techy person. She still starts in cloth. Sewing, draping, pattern-making and sketching come first. AI, for her, is refinement, not generation: colorways and fabrications she would otherwise render for hours, not the idea of the garment.

I like to describe my relationship with AI as using it as a visualization tool rather than a creative tool. It helps me communicate and explore ideas I’ve already developed, rather than generating those ideas for me.

Jane Wade, designer, on her SS27 process

She is explicit about the muscle. Efficiency can help parts of the process. She does not want it to replace the trial and error, the problem-solving, or the personal view that make the work hers. After the Google test she still fitted the collection on people. Photographers, stylists and models still sat on the campaign side of the line.

Erica Dhawan, a leadership expert whose coming book is Use Your Brain: Think For Yourself in the Age of AI, gives the same sequence in office language. Go first. Do not open the chatbot until you have a rough draft, bullets, or a voice memo. Make the model argue with you. Keep a human-only list of the parts of the craft it never touches. Those rules only work if you have a draft. They are a veteran protocol. They are not a junior’s first week.

A Board Anyone With the Same Login Could Build

Fashion spent 2026 talking about taste as if it were a shield against Silicon Valley. Raf Simons, after Prada’s Fall 2025 show, said designers have to work from the heart, the mind, the feeling, knowledge, respect for history, and interest in the future. Kosmyna’s version is colder. Creativity can be enjoyed by other humans, she says, and mathematics and artificial intelligence do not have a concept for that.

Beaty’s research says people who are already creative without AI are more creative with it, which is how you differentiate yourself when everyone has the same models. That sentence is a labor market. The veterans keep the heart-and-history speech and a faster deck. The juniors inherit a board that could have been produced by anyone with the same subscription, then get judged on originality they were never given time to practice.

Wade still puts every look on a person before it goes out. The factory still cuts cloth. Those steps do not care how finished the first file looked.

Harry is the editor and lead writer of STUDIO ONE NETWORKS, an independent title he owns and runs himself. Ten years in journalism, reporting first and editing later, taught him that entertainment and business are one beat seen from two sides: a box office figure is a company number, a streaming deal is a contract, a casting rumour is not a story until someone puts their name to it. He works from the record, whether that is a distributor's statement, a licensing agreement, an interview transcript or a set of published ratings, and checks every number against it before publication. The same rule holds for the rest of the site, which covers news, technology, science, sports, lifestyle, travel, auto and gaming for an audience spread across the world. When he gets something wrong, the article is corrected and the change is noted and dated, under a corrections policy anyone can read. Reader mail is answered by him at support@studioonenetworks.com.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending