Inside LVMH’s AI Factory
By Maghan McDowell, Vogue Business
Where does AI belong in luxury fashion? In some ways, it’s already discreetly arrived.
For the past few years, LVMH has been building out an “AI Factory”, a library of algorithms for use by its brands, which include Louis Vuitton, Dior and Tiffany & Co. The factory also helps its brands make use of AI, and educates the company’s workforce.
LVMH’s AI Factory team builds and provides modifiable, modular algorithms that can be deployed and tuned according to the brand’s needs, like Legos, to make the team’s work more efficient and the tools more readily accessible. “We don’t want to reinvent the wheel,” says LVMH’s Axel de Goursac, director of the AI Factory. “You can adapt to the business needs and the data, because it’s very important that algorithms adapt to the data, and not the contrary.” De Goursac explained the group’s evolving AI strategy during the recent Vogue Business AI Luxury Summit with Google, held this month at Google’s Paris headquarters.
The team of data scientists was originally formed in 2020, but it has grown since last year as generative AI tools reached common consciousness. To help guide its strategy, LVMH is working with Stanford University’s Institute for Human-Centred Artificial Intelligence. At a time when governmental regulations and industry-wide best practices are still in development in multiple countries and regions, the goal of the partnership is to proactively establish LVMH’s principles for ethical and responsible uses for AI that will guide how it makes decisions.
The group’s core tenets are to use these technologies to improve the customer experience and to aid in offering exceptional, desirable products, de Goursac says. He echoes what has emerged as a guiding principle for many proponents of AI’s potential: “It’s not automating or replacing humans, but really augmenting our employees to better serve these pillars,” he says.
That means that while a number of brands have experimented with customer-facing search tools that use generative AI, or chatbots that make recommendations or answer questions, LVMH brands are not likely to offer customer-facing AI tools in the near future. Instead, its strategy is to keep the wires under the hood and the tradition at the forefront. “We are not so keen to put the AI models directly in front of customers because we think that our competitive advantage is our client advisor workforce,” de Goursac says.
Already, LVMH has deployed a number of uses. In addition to the recommendation algorithms used in e-commerce, it helps client advisors make better recommendations to personal clients. It can suggest the most high-potential customers to contact, and for customers who aren’t the “VICs” (very important customers), it can supplement associate intuition by suggesting product category affinity. While the recommendations aren’t “deterministic”, meaning they don’t strictly dictate a specific outcome, de Goursac says, it has proven to help inspire relevant conversations and is more successful in leading to sales than pure associate instinct alone.
Additionally, the company is using AI in forecasting, spanning which products to produce when, and where to distribute them. De Goursac says that in terms of forecasting that accounts for scarcity, it had to build luxury-specific predictions that are different from standard fashion retail. The algorithms can make recommendations with varying levels of granularity, including which zones are best for certain products, and which stores.
Going forward, LVMH plans to continue educating its workforce, upskilling all 10,000 in AI in the coming years, and has worked with Google on an internal AI tool, called MaIA, that can provide business insights and answer questions. Part of its training includes how to successfully write prompts for generative AI. Already, 1,500 staff have received training.
Expect more generative AI pilots soon; during this summer’s LVMH Innovation Award, two companies that use generative AI to create content — Fancytech, which creates marketing videos (that are especially useful in the China market), and Bling.AI, which creates photorealistic 3D renders — were awarded prizes. Future projects might also help address sustainability, with plans underway already for a way to detect “end of life” products and produce quantities accurate enough to reduce unsold products.
For now, data science — not specifically generative AI — is still the most useful, he says. “We don't forget the fundamentals: data, data quality, data governance and traditional AI — which still brings higher value than generative AI.”