How does LVMH organise AI? The AI Factory, maison agents and the Innovation Award
LVMH runs AI as shared group infrastructure with maison-level control: a central data and AI team, internal assistants, clienteling agents and a startup award that feeds the pipeline.
KEY TAKEAWAYS Summary by the editors
- LVMH built the core of its group data platform in 2021 and 2022 on Google Cloud, which its CIO Franck Le Moal says left the group ready when generative AI arrived.
- In 2022 Journal du Net reported that LVMH's data factory, which includes an AI factory, had 60 to 70 staff and had industrialised about 200 reusable use cases in under three years.
- LVMH's internal generative AI assistant Maia, launched in 2023, is open to about 200,000 employees and handles 1.5 to 2 million prompts a month, according to Le Moal speaking in November 2025.
- Client advisor agents that suggest products and connections from client data are in use at Tiffany & Co. and Celine, while LVMH keeps a firewall between maisons' customer data.
- Le Moal said in late 2025 that more than 70% of LVMH's sales forecasts are supported by AI and that classic machine learning, not generative AI, still delivers most of the efficiency.
LVMH organises AI as shared group infrastructure that each maison uses on its own terms. A central data and AI team builds platforms, models and agents once, the 75 maisons decide what to deploy, and customer data stays separated by brand. Generative AI sits on top of years of classic machine learning in forecasting and clienteling, and the annual LVMH Innovation Award brings outside startups into the same pipeline.
What is the LVMH AI Factory?
The term describes the group's central capability for building and industrialising data and AI use cases. In November 2022, Journal du Net reported that LVMH's data factory employed 60 to 70 people, about half of them business experts working alongside data and machine learning specialists, and that it included an AI factory. According to that report, the team had industrialised about 200 use cases in under three years, each designed to be reusable by any brand in the group, and had trained 900 to 1,000 employees through an internal data academy over 16 months.
The model matters more than the label. Instead of every maison building its own forecasting or recommendation engine, the group builds a component once and offers it to all. Group CIO Franck Le Moal described the aim at the time as strengthening how each business function, from marketing and retail to supply chain and manufacturing, uses data.
Which technology platform does LVMH use for data and AI?
In a June 2025 interview published by Google Cloud, Le Moal said LVMH had built the core of its data platform in 2021 and 2022, with BigQuery as a central component, and had worked with Google Cloud for about four years. He said this put the group in a position where it "was very data aware" when generative AI arrived. LVMH uses Vertex AI to explore agents and applications, which it then adapts into its own custom agents. In 2022, Journal du Net also named Dataiku as the group's machine learning platform.
Governance is built into the architecture. Le Moal told Google Cloud that LVMH "always maintain[s] a firewall between the brands and the data they've collected", while shared benchmarks let maisons compare performance on measures such as digital conversion. Each maison remains responsible for its own wider IT ecosystem.
| Area | What has been reported | Source year |
|---|---|---|
| Clienteling | App with product recommendations per client used by advisors at about ten maisons; agents that adapt client messages at Tiffany & Co. and Celine | 2022, 2025 |
| Forecasting | Demand and stock forecasting; more than 70% of sales forecasts supported by AI | 2022, 2025 |
| Internal productivity | Maia assistant open to about 200,000 employees, 1.5 to 2 million prompts a month | 2025 |
| Brand voice | Voices agent developed at Bulgari, planned for reuse in other maisons | 2025 |
| Product and cost | Manufacturing cost estimates for bags in minutes instead of weeks | 2022 |
| Store allocation | Data used to decide which stores receive very high-end jewellery pieces | 2025 |
How do LVMH's AI agents work for client advisors?
Clienteling is the core use case. In the Google Cloud interview, Le Moal explained that an AI agent is layered on top of client data so that an advisor "can chat with this data to find what they need even faster", and that the agent can surface new opportunities for additional items or experiences. He named Tiffany & Co. as one of the leaders in working with client advisor agents. Speaking at Tech for Retail in Paris on 25 November 2025, as reported by LeMondeInformatique, he said agents in the clienteling app that adjust messages to clients were already installed at Tiffany and Celine.
The principle is explicit: "you still have a client advisor, and that's who we want to empower with data and AI," Le Moal said. The agent drafts and suggests; the advisor decides and owns the relationship. For a luxury group, that division is also a brand safeguard, because tone and judgement remain human.
How widely is generative AI used inside LVMH?
LVMH's first agent, the internal assistant Maia, launched in 2023. According to LeMondeInformatique, it started on ChatGPT and now also offers Gemini, Claude and Mistral models, is open to about 200,000 employees and receives 1.5 to 2 million prompts a month from more than 30,000 weekly users. Uses mentioned by Le Moal include document digestion, translation, product descriptions and customer service. A second agent, Voices, was developed at Bulgari to keep a consistent tone of voice and is planned for reuse in other maisons, adapted to each one's data.
Le Moal also stressed proportion. "Even though everyone talks a lot about GenAI today, classic AI is still what makes us more effective," he said, citing roughly 50 classic AI algorithms deployed across the houses.
What role does the LVMH Innovation Award play?
The award is the group's external scouting channel. The ninth edition was held on 12 June 2025 at VivaTech in Paris, where 15 projects involving 11 maisons were shown. Three partners were recognised: Kahoona, which works with Dior on real-time predictive audience segmentation for anonymous online visitors; Genesis, which works with Moët Hennessy on measuring soil health; and OMI, which works with Guerlain on a 3D and AI content studio that produces product visuals faster and at lower cost. The pattern is consistent with the internal model: startups are judged on a working collaboration with a maison, not on a pitch alone.
Where is LVMH cautious about AI?
The group's public statements are notably restrained. Le Moal said in November 2025 that LVMH is accelerating on generative AI and agents but is "not in a logic of submission" and is "not a technology company". Points of caution he raised include:
- Conversational commerce is still being observed because it "still lacks some robustness", and security and confidentiality questions remain open.
- Brand differentiation: answers will differ by maison, and Sephora in beauty will not respond the way high jewellery does.
- AI search: purchases driven through generative AI engines account for only about 3% of sales, and the group is monitoring its visibility in those answers.
- Cybersecurity: securing AI apps and agents, segmenting customer data and controlling partners for resilience.
- People first: creatives, artisans and sales advisors remain the priority, with AI in support.
What are the limits of what is known?
Most published detail comes from the group's own CIO, partner vendors and trade press interviews. LVMH has reported usage figures and forecasting coverage but has not published audited financial returns for specific AI programmes. Figures from 2022, such as the 10 point improvement in forecasting accuracy reported by Journal du Net, describe that period and should not be read as current results.
Frequently asked questions
Does LVMH have its own AI team?
Yes. LVMH runs a central data and AI capability that builds reusable use cases for its maisons. Journal du Net reported in 2022 that the data factory, including an AI factory, had 60 to 70 staff, about half of them business experts.
Which AI tools does LVMH use?
LVMH's CIO has described a data platform on Google Cloud with BigQuery and Vertex AI for agents. Its internal assistant Maia offers several large language models, including ChatGPT, Gemini, Claude and Mistral, according to LeMondeInformatique.
Which LVMH brands use AI agents for client advisors?
Tiffany & Co. and Celine have clienteling agents that help advisors adapt client messages and suggest products, according to LVMH's CIO in 2025. Bulgari developed a brand voice agent called Voices.
Who won the 2025 LVMH Innovation Award?
Kahoona (with Dior) won Best Business Prize, Genesis (with Moët Hennessy) won Best Impact Prize and OMI (with Guerlain) won Most Promising Prize at the ceremony on 12 June 2025 during VivaTech.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
