Most large groups approach AI through optimisation: automate customer service, speed up the back office, assist teams. It is useful. But it is the smallest part of the opportunity. Above all, AI lets groups turn assets they already own, and do not exploit, into new companies that sell to the market.
We see three ways to do it.
1. Sell what your data knows
Many groups accumulate data they only use for themselves. With AI, that data becomes a product.
- Walmart Luminate. Walmart sells its suppliers insights drawn from its sales and customer data. The service launched internationally in 2024.
- Airbus Skywise. Born as a data platform for airlines, Skywise became a fully fledged Airbus subsidiary in April 2026, which also includes Navblue.
- Altametris. Created by SNCF Réseau to capture and analyse the state of its infrastructure by drone, the company sells this expertise to other network operators.
2. Use a business asset to enter a new sector
An asset built for the core business can open a completely different market.
- Exaion (EDF). Created within EDF Pulse, Exaion moved into high-performance computing and AI infrastructure, drawing on the group's energy expertise. In 2025, MARA signed an agreement to take 64% of it, for $168M.
- Unlimitail (Carrefour and Publicis). Carrefour turned its store and customer data into a retail media business, open to other retailers in Europe and Latin America. It is a joint venture between two groups, but the mechanism is the same: a retailer's asset becomes a media business.
3. Charge for access to your content and archives
AI models need quality content. Those who own it have a new customer.
- AFP and Mistral AI. In January 2025, AFP signed an agreement with Mistral for the use of its news wires.
- Le Monde. The group signed with OpenAI in 2024, then with Perplexity in May 2025.
- Reddit. Its agreement with Google, revealed in 2024, is reported to be worth about $60M a year.
A licence is not yet a business. It becomes a company when you build a product, a team and an offer on top of it.
Why these businesses are better built outside
An AI business needs rare profiles, a startup pace and a technical freedom that a group's rules make difficult. It also needs the group's asset: the data, the distribution, the brand. That is exactly what a company created with the group rather than by it allows: the group's rare asset, the execution of entrepreneurs who own part of the capital, and funding committed over several years.
Three questions to start
- What data, know-how or customer relationships do we own that others would pay to use?
- Who, outside our current customers, has the same problem we solved for ourselves?
- Are we ready to let a new company sell this asset, including to our competitors?
This is the purpose of the opportunity studies we run with large groups: finding, among their assets, those that can become a new company, and deciding which ones deserve to be launched.
Sources
- Walmart, global launch of Luminate (May 2024)
- Airbus, Skywise subsidiary (April 2026)
- Railway Gazette, SNCF Réseau forms Altametris
- EDF, MARA and EDF Pulse Ventures agreement on Exaion
- Publicis Groupe, launch of Unlimitail
- Journal du Net, AFP and Mistral AI agreement (January 2025)
- Reuters via The Star, Le Monde and Perplexity (May 2025)
- The Decoder, Reddit and Google agreement (February 2024)