Over the last few years, AI tools and plug-ins for every kind of fashion retail operation have promised to make processes stress-free and ultimately profitable. The frenzy of new software and available tech stack integrations made sense; in 2024, McKinsey forecast that AI would add between $150 and $275 billion to the fashion industry in the next five years.
But with the BoF-McKinsey State of Fashion 2026 recently reporting that up to 90% of transformative AI projects in fashion are stuck at the pilot stage, the evidence is becoming clear: the AI hype is deflating, and the fatigue is setting in. The results aren’t coming, the projects are stalling, and the trust is waning. So, what happened?
Many AI projects remain constrained by structural and cultural barriers within the workplace because the benefits of the tech aren’t understood or communicated to the right people. And, in most cases, the onboarding doesn’t gel with the existing workflows; it’s slow, clunky, and the support tickets just pile up.
By now, all those AI pilots put in place last year are currently waiting on their fate from the world’s busiest fashion companies - and the outlook isn’t promising. This is because the process of integrating AI into existing operations is underestimated, and companies are mistaking tools for plug-and-play solutions, resulting in no measurable return on investment.
The reality is fashion has changed, and traditional tools and legacy tech structures no longer match the pace of modern retail. Fashion needs new technology solutions for the industry’s biggest problems, and the 90% failure statistic shouldn’t be a reason for brands to distance themselves from it.
Here, we explore why AI in fashion is stalling, and how teams can integrate fashion-first tech that seamlessly solves real-world issues and relieves teams of manual processes.
1. AI projects rely on clean data
Before brands can even begin to get value from their AI ambitions, they need to be able to see the business clearly.
“A lot of brands are looking to AI to save them, fix their planning, or give them the competitive edge,” said Teneille Oakley, a fashion merchandising consultant whose career spans Australian retail giants like David Jones, Aje Collective, and Myer.
“However, in reality, these brands are not even tapping into the basic opportunities their data is presenting them,” she explained.
Teneille believes this is because the data is not easily accessible or accurate, and that most brands and retailers don’t have the tools or knowledge to interpret their own data correctly.
“Having clean data is an essential foundation before exploring what benefits AI can provide. Without it, I can totally understand why the majority of AI projects are not successful.”
“Having clean data is an essential foundation before exploring what benefits AI can provide."
Once the data is accurate, AI processes need to be practical and auditable for businesses to effectively embed the technology into everyday processes. Platforms should offer solutions on this basis, automating error-prone steps in the product cycle: like precision forecasting for product depth and size demand, generating instant post-season and in-trade analysis across channels, and digitizing one-source reporting across departments.
“I see Style Arcade being the necessary tool to provide these insights, interpret the data, and help guide the brand on the appropriate trading actions,” Teneille said.
2. AI needs to be treated like an assistant, not the authority
The teams stalling their tech stack projects are the ones chasing the big AI transformation.
However, as previously explained, Style Arcade’s Chief Technology Officer, Tristan Hoy, suggests the best way to position AI within your business is not as an unquestionable authority but as a transparent, explainable, and adaptable assistant.
And now, with the mainstream adoption of large language models (LLMs) like ChatGPT and Claude, he cautions against losing the aspects of what makes fashion roles human.
“Prompting isn't fun. It's addictive, but it's not fun,” said Tristan. “I mean, imagine if you opened Instagram, and instead of personalized, visual content, all you were greeted with was a blank page and a chat interface.”
“Our team uses Claude so much, and while there's definitely value in a natural language interface, we're going to lose something really important if it replaces everything,” he warned.
Tristan sees the future approach to AI in fashion retail as a collaborative environment. Teams need to recognise that their own knowledge and experience are crucial to the business, which in turn allows them to acknowledge the technology’s data analysis and automation as inherently valuable.
“There's also something about separating ourselves from the work,” Style Arcade CEO Michaela Wessels added. “Now, of course, it's dangerous to delegate to an LLM - but it's also deeply unsatisfying. And in fashion, so much of the magic happens in the inspired details.
AI is here to stay, but simply supervising LLMs is not the future of work. We need to make sure we keep things rewarding - and human,” she explained.
“...it's dangerous to delegate to an LLM - but it's also deeply unsatisfying. And in fashion, so much of the magic happens in the inspired details."
3. AI projects require effortless adoption
Many projects stay stuck in pilot mode because businesses are unsure how to measure their impact and test for their ongoing value.
And, it’s the early stages of introducing the software or tool within the business that are critical for gauging its true impact within the business. The project’s success requires the quick and seamless onboarding of multiple departments or entire teams to ensure minimal product data disruptions and to glean the return on the investment (ROI) of the software.
Before acquiring a new platform or software tool, retailers should ask themselves:
- What is the onboarding experience like? Early engagement should be enhanced by customized onboarding procedures that walk users through the system, highlight essential features, and provide tailored recommendations.
- Is it a fit? Do the platform’s features suit your business requirements and integrate seamlessly with your workflows?
- Is it user-friendly? Quick adoption will rely on a visually-led, intuitive interface with features relevant to your existing procedures, and that are easy to understand.
- What’s the customer service like? Timely and effective assistance helps users overcome obstacles, builds confidence, and encourages user adoption.
From there, to measure the value of generative AI platforms and tools, businesses can draw their own conclusions by discerning how much added revenue it contributes, how much time and cost it saves, and whether the software can be interrogated and the logic understood, so the user can continually trust it.
Fashion moves fast, but technology moves faster.
