Key Takeaways:
- The WTO projects AI could expand global trade in goods and services by up to 37 percent and lift global GDP by 12–13 percent by 2040. These gains are contingent on developing economies closing infrastructure gaps and adopting AI effectively, a caveat that highlights the risk of unbalanced adoption in a trade-dependent industry like fashion.
- The “Workslop” phenomenon – a term coined by the Harvard Business Review to describe superficial AI outputs in other industries – offers a caution for fashion. If brand headquarters focus mainly on speed and cost, the outputs may look complete but still lack the technical depth required, leaving specialist suppliers to pick up the slack.
- Fashion’s core supply chain expertise – patternmaking, sizing, and construction – sits with specialist suppliers, and AI only creates real productivity when it compounds that knowledge rather than bypassing it. Tools like defect detection, grading accelerators, and yield models may seem unglamorous, but by amplifying the judgment already present in these supply chains they offer the most credible path to efficiency.
In its latest report, the World Trade Organisation sketches a future where, by 2040, AI could expand global trade in goods and services by as much as 37 percent, and lift GDP by 12-13 percent. The report also suggests that developing economies could see especially large gains – if they can close infrastructure gaps and adopt AI effectively.
But those adoption gaps are not the only caveats attached to the WTO prediction. Implicit in their analysis is the commitment to the idea that trade requires two parties with mutual benefit and understanding. And in fashion’s case that reminder is especially important, since, in a very real way, clothing and footwear does not get made without creative input from one party and specialist engineering and operational capabilities and capacity from another.
For additional context, the same mood of optimism was easy to spot at the start of the year, when our joint survey with Fashion by Informa found that more than four in five professionals saying AI would add value to their business in 2025. How that value was defined may have been extremely broad, but it’s also straightforward to condense it all down to improvements in the way goods are made and sold.
But what actually sits under that umbrella, in fashion? On one hand, it’s the creation of great brands, well-defined and marketed, with ongoing improvements to product outcomes. On the other it’s the ability to execute on those visions, and to translate them – through cutting, sewing, patternmaking, platform engineering, shopfloor efficiency improvement and so on – into tangible things. And the concern raised by the WTO report is that adding AI to just one side of this equation will increase output and throughput, but it will fall to those upstream specialists to then separate quantity from quality.
And this same concept is also becoming visible across other industries. In a recent article the Harvard Business Review coined a label for a kind of AI-generated output that meets usage targets but falls short of technical expectations and requires extensive rework from other people: “Workslop”. It shares some DNA with the “AI slop” shorthand used to describe the flood of low-grade written and visual content appearing online, but this term sits closer to, as the name would suggest, the work environment. It denotes, essentially, employees using AI to generate material (which, for fashion’s purposes, could be generative product photography, generative sketches, product descriptions, technical specifications, or any number of other deliverables where workers are employing AI as an assistant. Often, in other industries, this is the kind of work that looks superficially acceptable, but, when it lands in the hands of people equipped with the specialist skills to scrutinise it, its flaw become evident.
We touched on this, albeit approached from a different angle, two weeks ago through the lens of vibe coding and the newly formed job title of Vibe Code Fixer, but this new report shows that the scale of the problem extends far beyond the realms of programming. No matter the industry, the promise of time saved turns into more work for whoever is then asked to translate the idea into reality.
It’s already evident that fashion is working to find the right way to bring AI into the product creation process, because the time and efficiency improvements can feel too compelling to ignore. Silhouettes created in seconds, mood boards pulled together from prompts, and presentation decks built without breaking a sweat.
The challenge is that there will sometimes (or often, depending on the setup) be a disconnect between something that looks believable enough on-screen, in a meeting, and something that’s ready to carry forward into development and production. And if the use of generative AI is confined to experimentation and ideation then this doesn’t pose an issue; but if the same systems are being adopted with the view that they can replace existing solutions and workflows that extend deeper into the product journey, then this is where the issues will arise.
This is why the World Trade Organisation’s emphasis on developing economies rings so true. Much of the industry’s hands-on expertise and specialist patternmaking, sizing, and construction knowledge sits in supplier bases in Europe, Africa, and Asia. Contrasted against the objectives set for the actors that make up that supplier landscape, the use of AI inside brand headquarters has tended to circle a much narrower set of goals: speed and cost. Quicker sketches that don’t tie up design time, campaigns that can be turned around leaner, copy that doesn’t need another freelancer commissioned. Each case makes sense on its own. But the risk isn’t only that more is being made, faster. It’s that what’s made looks finished enough to move along, further up the chain, until someone who works for a partner spots the fundamental issues with a brief or tech pack.
Yet, the upstream AI picture isn’t all cautionary. There are examples where AI doesn’t compete with skill so much as compound it. Systems that catch defects before they spread down a line. Grading tools that turn days of work into hours. Yield models that shave waste while trimming costs. None of these feel especially futuristic, they’re ordinary, almost boring, which is what makes them work. They don’t ask anyone to rethink the process, they just sit inside it, amplifying the judgement that’s already there.
We’ve said this before in different ways, and it’s worth repeating: prompts are cheap, but the taste and judgment required to make physical sense of them is not, or at least shouldn’t be when we consider just how much value it adds to the typical product journey. Big AI labs like to talk about productivity and efficiency, but it’s vital to remember that those don’t come from the flood of ideas, drafts or images on their own. Productivity benefits are realised when those outputs meet the skills that can carry them through to reality.
For The World Trade Organisation’s projections to hold, the issue isn’t whether AI can generate more ideas, more options, or more open-ended possibilities. The real test is whether fashion will avoid the pitfall of leaping into AI’s accelerating effect on output, without investing in a commensurate way in the people who can determine how useful that output really is.
