Could generative AI solve fashion’s excess stock problems?

To receive the Vogue Business newsletter, sign up here. Amid fluctuating global demand and macro-economic pressures, optimising and allocating inventory for multiple locations, across multiple channels and without holding excess stock remains a challenge for many fashion businesses. Could generative artificial intelligence be the answer? Generative AI has the potential to transform core processes in the luxury sector, including supply chain and logistics, helping luxury brands improve warehouse operations and inventory management, a report by management consultancy BCG and Italian luxury association Altagamma published this month predicts. Luxury CEOs “need to experiment” to discover and decide on the best use cases, the report advises. There are several big barriers to overcome, however. Generative AI is still far from being considered fit for purpose for use within a supply chain setting, relies heavily on large cohorts of training data, and is complex to implement. Proponents say getting the foundations of machine learning and data science right now will help supply chain directors smooth the way for using generative AI in the future. “Generative AI enables retailers to have better conversations with their data, helping both customers and employees,” says Marco Limena, CEO of business intelligence firm Board International. “It has a number of key benefits for retailers, such as reinforcing customer centricity — whether through helping businesses with consumer research, scenario forecasting, or creating new products with synthetic customer data. This serves to significantly speed up new ranges, product design, and sourcing while improving time to market.” AI is already transforming supply chain management. It enables more accurate demand forecasting by tracking data such as store footfall and automating decision-making in areas such as pricing, promotions, replenishments and forecasting. Brands are also exploring how it could help with traceability: in 2019, Google announced a partnership with Stella McCartney to pilot a tool that uses machine learning — a subdiscipline of AI that uses data and algorithms to imitate the way humans learn, gradually improving its accuracy — to trace raw materials throughout the supply chain, using data from the brand. The difference with generative AI is that it can analyse vast amounts of data from various sources to offer answers. Feed in inventory data and ask: “Where am I at risk of over and understocking based on current sales numbers?” and the newest generative AI technology can use tools such as pricing platforms and Google search to provide an answer that is structured to the business’s needs — rapidly speeding up the inventory management process, explained Dr Hardy Kremer, vice president of data science and data engineering at digital strategy consultancy Ommax, in a post accompanying the publication last week of its new report, ‘Transforming luxury fashion inventory management with generative AI’. Big players such as Louis Vuitton and Dior parent company LVMH Group and US department store chain Macy’s are keen to explore generative AI’s potential. Franck Le Moal, chief information officer at LVMH, predicts that the solution to many supply chain problems will lie in a blend of technologies that use AI, including machine learning and generative AI. “We believe that all these technologies will work together to enable better management of our supply chain, from inventory and stock allocation through to production capabilities and having a strong impact on sustainability,” he says. The rewards of bringing generative AI into the mix could be huge. Supply chain challenges continue to impact global fashion business after hikes in the price of raw materials and the war in Ukraine compounded problems caused by the pandemic. Demand volatility across global markets is adding further pressure. Luxury spending has notably slowed in the US as inflation dampens consumer spending. VF Corp, Nike and Under Armour are among the companies that have recently flagged problems with excess inventory. “By analysing data related to suppliers, production capacities, lead times and transportation logistics, generative AI models can help brands make data-driven decisions about sourcing, production planning and inventory allocation,” says Ommax’s Kremer. “This enables luxury brands to streamline their supply chain processes, reduce costs, and improve overall operational efficiency.” Building the foundations LVMH’s Le Moal says machine learning has been pivotal in enabling the company to fine-tune its sales forecasts, to help with distribution management and relocating inventory. This is helping answer questions, such as, “Should it be at country level or within our global distribution centre?” He explains: “We’re able to really allocate and optimise location of our stock depending on the demand and depending on our production capabilities. It enables us to allocate very, very quickly or relocate very quickly,” he adds. LVMH is also investigating the use of machine learning on the production side, to ensure items are “being produced at the right time, in the right place, with the right quantity,” says Le Moal. Not only does this lead to more sustainable purchasing practices, but the ability to respond quicker to global demand fluctuations in terms of distribution. Le Moal says “smart” warehouses — which use AI to enable real-time decision-making to drive improved efficiencies when it comes to order fulfilment — will play a key role in inventory management across the LVMH portfolio in the future, with the technology currently being developed across brands including Louis Vuitton, Fendi and Tiffany & Co. Sandra Han, SVP merchandise planning & allocation for Macy’s, Inc., says it has been embedding machine learning into the company’s day-to-day decision-making since since February 2020, which has helped strengthen its forecasting abilities, to plan and adjust inventory to meet customer demand. “Our forecasting abilities, enhanced by machine learning, have increased our visibility and predictability, which allows us to plan and adjust our inventory to effectively meet customer demand,” Han says. “We predicted that the consumer was going to shop more in-store and return to purchasing occasion-based categories [last year]. Through the increased flexibility we’ve achieved by modernising our supply chain, we moved quickly to adjust the timing, amount and composition of receipts by channel, category and brand.” Generative AI has the ability to go one step further by producing realistic outputs based on the patterns and information present in training data, enabling retailers to have better conversations with their data. Le Moal sees the biggest opportunity for generative AI in relation to product engineering and optimisation — enabling different areas of the business to quickly access information. This would range from sales associates being able to access product information or real-time stock location questions in-store to factories being able to access data around stock levels or answering crucial questions around materials such as, “Can I still get this material 10 days from now?” The knock-on impact will be the ability to adjust the distribution of production forecasts “from monthly five years ago to daily now, and soon it will be in real-time”, Le Moal says. “Generative AI will enable us to fine-tune our supply. So, as soon as we see a signal at sales forecast or market level, this will be sent immediately, in real-time, up to the distribution warehouse and off to the production factory.” Generative AI is already being tested in Dior and Louis Vuitton stores to improve the customer experience, he says. A sales associate can speak into their iPhone and ask a question, such as ‘Can you tell me more about this red bag?’. “Generative AI algorithms offer the ability to access, very quickly, the product history and product information,” Le Moal says. Data limitations Underpinning all of this is data. It’s why Hugo Boss is building a €15 million data campus in Portugal that will employ at least 250 people, including data scientists, data engineers, data visualisers and business intelligence specialists. LVMH is also investing heavily in this space as it looks to transform into a “data-driven company”, says Le Moal. It has 70 data scientists currently, a proportion of whom are focused solely on supply chain operations. “Having the capabilities to really rely on accurate forecasting, amplified by data, is a great opportunity for us to minimise any material impact we may have been facing until recently,” Le Moal explains. However, the limitations can’t be ignored. Dr George Bargiannis, subject area leader for the department of computer science at the University of Huddersfield, says state-of-the-art generative AI technology — such as ChatGPT-4 (the latest tool offered by OpenAI, which uses a large language model to understand and generate text in a human-like fashion) or Bard (Google’s latest large language model-powered tool) — is not ready for commercial use within a supply chain setting. “As with any deep learning technology, availability of large amounts of good quality data is a prerequisite for generative AI to be useful, as their success rests on their ability to self-improve by learning to filter out low quality or unsuitable information and curate both their input and output. If data is scarce or of low quality, then so will the outputs of generative AI,” he says. Ommax’s report also notes that generative AI models may struggle to accurately predict demand during unprecedented events, such as the Covid pandemic, which can introduce behavioural changes that aren’t captured by historical data. “So, in future emergencies, while humans will likely rely on AI to help them make decisions, ultimately, it still needs to be humans that make the big decisions,” wrote Kremer. The report also points to the challenge of building scalable and reliable generative AI systems that interact with other complex inventory management systems. Le Moal says ensuring data is accurate and reliable is one of the most significant challenges for LVMH. “We’re really working on the foundation of our high data governance, making sure that we’re training and upskilling our teams, making sure that every single piece of data added is qualitative. Nothing can work without strong data and qualitative data, we’re making a significant effort to organise our company to become a data-driven company.” Aarthi Ramamurthy, chief product officer at e-commerce platform CommerceHub, which works with a range of companies, including Coach and Kate Spade owner Tapestry, agrees that generative AI currently relies too heavily on large cohorts of training data and is too complex to implement for everyday inventory management. Still, she argues that generative AI can play a “complementary role in inventory management by providing valuable support and enhancement”. She adds: “AI can handle routine, mundane tasks and generate insights and recommendations, which can be reviewed by experts and actioned upon.” This allows the experts to focus on setting strategy and providing customer value, while AI systems can focus on data and insights to feed to the experts. Lee-Anne Doltis, marketing director at London-based menswear business Savile Row Company, emphasises the need for human input in the supply chain, where relationships with suppliers are so important. “While generative AI can bring significant advantages, it's not a cure-all. We need to focus on close collaboration with our suppliers to ensure we have an effective product lifecycle management to keep waste to a minimum and a comprehensive sustainability strategy that addresses every aspect of our operations.” Holger Harreis, senior partner at global consultancy McKinsey, warns that inventory management strategies “need to be holistic” and “shaped by starting from the customer experience lens, looking at where the value really is, and then finding the appropriate AI approach”. He adds: “The underlying foundations need to be considered. Without the right data set and the right quality, no AI model will do a good job.” Comments, questions or feedback? Email us at feedback@voguebusiness.com. More on this topic: The fashion exec’s guide to generative artificial intelligence Can AI transform the skincare buying experience? Generative AI is coming to Google search

Make better business decisions

Make better business decisions