AI still needs taste
Trade coverage argues fashion firms rushed to deploy large language models but the competitive edge isn’t the tech itself—it’s knowing where AI works and where human judgement must stay in charge. (businessoffashion.com) Workers who use AI as a learning tool may increase their value over time, underlining that supervision and creative direction are becoming the scarce skills. (stltoday.com)
Fashion companies spent the last two years treating large language models like a universal intern, asking them to write product copy, generate campaign ideas, and even sketch design directions. The new argument from The Business of Fashion is that the edge is no longer getting access to the tools in 2026; it is knowing which jobs can be sped up and which ones still break without a person steering them. (businessoffashion.com) That distinction is unusually sharp in fashion because the industry sells taste, not just output. A model can produce 50 blouse descriptions in seconds, but it cannot reliably decide which one sounds like The Row, which one sounds like Zara, and which one quietly cheapens a luxury brand’s voice. (businessoffashion.com) The easy wins are the jobs with clear patterns and lots of repetition. The Business of Fashion says companies are getting useful results from large language models in tasks like writing code, drafting text, and generating first-pass images, where speed matters and a human can quickly reject bad work. (businessoffashion.com) The weak spots show up when the brief is fuzzy or the brand risk is high. A tool that predicts the next likely word or image can mimic last season’s language, but fashion still needs someone to decide whether a hemline, casting choice, or campaign mood feels early, tired, or off-brand before it reaches customers. (businessoffashion.com) That shifts value upward from pure execution to supervision. If software handles the first draft, the scarce skill is increasingly the editor who can spot the one wrong adjective, the merchandiser who knows a trend is peaking, or the creative director who can tell when “good enough” will damage a label that spent 20 years building its image. (businessoffashion.com) The labor evidence outside fashion points in the same direction. A Washington University in St. Louis summary of new research says workers who treat artificial intelligence as a complement to their own effort tend to increase learning and effort over time, while the payoff depends on whether people use the tool to build human capital rather than coast on it. (source.washu.edu) That pattern matches one of the clearest findings in recent workplace research. A Quarterly Journal of Economics study of customer-support agents found artificial intelligence assistance raised productivity by 15 percent on average, with the biggest gains going to less experienced workers, which suggests the tool can act like on-the-job coaching as much as automation. (academic.oup.com) For fashion, that means the junior employee who uses artificial intelligence to learn line planning, copy tone, or technical vocabulary may become more valuable, not less. The person who simply pastes prompts into a chatbot without developing judgment is easier to replace than the one who uses the same software to sharpen judgment faster. (source.washu.edu)
