Subscribe to Tech Mode with Marc Bain, a deep dive into the most intriguing developments in artificial intelligence and its impact on the fashion industry.
Welcome to the relaunch of The Business of Fashion’s Technology Newsletter. We’re calling it Tech Mode, and here’s what you can expect: We’re introducing a new format designed to fill you in on the important news at the intersection of technology and fashion once a month. There’s an endless stream of updates to follow, and it’s hard to keep up with it all, so you won’t have to. You’ll get a quick download of the bits and bytes you need to know.
First, in case you missed it, Lululemon just appointed its first chief AI and technology officer, Ranju Das. In this edition, I dig into the latest controversy over AI-generated models, the tepid response to GPT-5 and a fresh sighting of fashion’s white whale: automated sewing machines.
Let’s get into it.
Who Isn’t Using AI-Generated Imagery?
Sign language: The newsletter Blackbird Spyplane published a story this week accusing J.Crew of using AI to “counterfeit their own vibes.” It pointed to posts on the retailer’s Instagram done in the style of J.Crew’s old catalogues. J.Crew’s posts included no mention of AI, but Blackbird Spyplane did an accounting of the slightly off details in the images they identify as telltale signs of the technology. They even tracked down the apparent creator — Sam Finn, an “AI photographer” — and his agency, which attributed the imagery to Finn on Instagram.
J.Crew confirmed to The Business of Fashion that Finn’s studio did create the imagery. The partnership, it said, is “one of many examples of how we engage artists of all genres to interpret our brand and experiment with different art mediums.” It has since updated its Instagram posts to credit Finn.
Fool me once: The problem with the AI images, according to Blackbird Spyplane, is that they run counter to what the authors presume J.Crew was aiming for — “when you realize that the seemingly rarefied locales and energies depicted in these new pictures are in fact replicas, assembled by a computer that seems to have scraped all those old catalogs and counterfeited their vibes, then that kills the aspirational fantasy,” they wrote.
To disclose or not to disclose: Part of the complaint is that J.Crew didn’t disclose its use of AI. That’s become a tricky subject for brands. They’re a bit damned if they do, damned if they don’t. H&M was slammed when BoF reported the company would start using AI digital twins of real models. People were also outraged last month when Guess ran a spread in Vogue that revealed in tiny type that the images were made with AI.
The argument right now is that AI imagery can’t transport or move viewers the way imagery made by and featuring real humans can. But it’s also getting harder to spot AI images. If AI gets so good that people can’t tell the difference, and brands don’t disclose it, does that solve the problem? Presumably not, but we’re likely to reach a point where more of the imagery we see is AI. Most of the time we’ll have no idea.
What GPT-5 Says About the State of AI
What happened: The biggest news of the month was the Aug. 7 release of OpenAI’s long-awaited new AI model, GPT-5. The expectation was that it would be a major advance in the abilities of large language models, ushering in new capabilities for businesses in all sorts of fields — including fashion. It wasn’t. While it improved on some benchmarks, overall it was an incremental step forward from OpenAI’s previous releases.
So what: The fact that GPT-5 was not a giant leap for AI seemed to confirm one of the biggest fears in the space: that progress has stalled. The assumption for years was that building systems with more compute and more data would translate to increasingly astonishing capabilities, ultimately getting us to what’s called AGI, or artificial general intelligence — AI that can perform any task as well as or better than a human. Instead, returns have plateaued.
What’s it all mean: The significance of all this depends on who you ask. AI boosters believe there are still dramatic gains to be had from techniques such as post-training, which entails refining a model after the fact. Sceptics say LLMs have hit a wall, and the quest for AGI will only be achieved by other means.
For fashion businesses, these arguments only matter insofar as they result in tools that change how companies work. AI isn’t going to replace everyone just yet, but it does already have some powerful applications for tasks such as design, image creation and copywriting. It’s already affecting jobs, and businesses will likely be finding uses for existing and emerging AI tools for years still.
Is Generative AI’s Value to Businesses Overblown?
Tell me about it: A study out of MIT has been getting attention after it concluded that, despite as much as $40 billion in enterprise investments in generative AI, 95 percent of businesses are getting no discernible return.
What’s the problem: The researchers identified a few things standing in the way of companies making the most use of AI. Many businesses struggled to integrate tools into their workflows, and they were reluctant to learn custom enterprise tools, defaulting to general products like ChatGPT. The study’s conclusion wasn’t that AI is a waste of money; it was that many companies take the wrong approach to using it.
What’s the solution: The 5 percent of companies wringing lots of value out of AI shared some traits. They started by focusing on narrow tasks where they could use easy-to-setup tools that provided some immediate improvement, they created systems that learned from feedback and they scaled from there. Just using a generic off-the-shelf tool or building some elaborate custom programme and expecting it to suddenly change the whole organisation was pretty much guaranteed to get the project stuck in the pilot phase.
Return of the Sewbots
Show me the money: Softwear Automation announced $20 million in fresh funding earlier this month, with Danish fashion company Bestseller leading the round as a strategic investor. Softwear Automation makes sewing robots — aka Sewbots — that it believes could fully automate garment manufacturing.
Sew what: The company has been around for years, but after announcing some projects in the late 2010s, it went quiet. In 2022, CEO Palaniswamy “Raj” Rajan told BoF that it “ran into technology challenges working with knitted fabrics,” which are difficult for robots to manipulate because they’re soft and stretchy. It seemed like the company might be stuck. Now it’s back with money and a new partner.
High stakes: Companies aren’t trying to automate sewing because garment workers in countries like Bangladesh are too costly. The goal is often to make it cost effective to produce garments in markets where labour is expensive, like Europe and the US. That would enable more on-demand production, since you wouldn’t be dealing with the lead times that come with shipping from Asia. This seems to have been at least part of the logic behind the deal. In a statement, Softwear Automation’s Rajan said the partnership would be a “powerful catalyst for the future of on-demand and localised apparel manufacturing.” Has Softwear Automation solved the technological challenges that stalled it before? We’ll just have to wait and see.