How generative AI boosts performance throughout our Group
An in-depth look at our strategy for implementing generative artificial intelligence: how we develop in-house tools, what we use them for, ethical considerations, data sovereignty, environmental impact, and more.
The second digital revolution
The first digital revolution at SNCF marked a shift from analog technologies to digital solutions. A prime example of this transformation is the SNCF Connect app, now used by 20 million people across France. Today we’re entering a new digital era with the rise of generative AI—cutting-edge tools with the potential to revolutionize key support functions like human resources, legal services, customer relations and rail operations. To harness this potential, we’re working with industry leaders and academic experts, and we’ve formed a dedicated committee to oversee the rollout of generative AI. This committee includes representatives from our subsidiaries and the Group ethics department.
We sat down with Julien Nicolas, Director of Digital Technology, Group AI and e.SNCF Solutions, to discuss how these tools are developed, how we’re helping employees to adopt them, real-world applications in our operations, their benefits and impact, and the key environmental, sovereignty and security issues they raise.
SNCF Group: how long has SNCF been using artificial intelligence solutions?
Julien Nicolas: We’ve been using AI for quite some time, particularly in industrial areas like maintenance, as well as in sales and marketing. However with the arrival of generative AI and the launch of ChatGPT in late 2022, we entered a new phase—exploring, testing and experimenting with these technologies.
We developed SNCF Group GPT under strict ethical guidelines. (...) All data is stored on our private cloud infrastructure, ensuring it is never used to train generative AI models.
SNCF Group: What’s our process for adopting generative AI at Group companies?
J.N.: We’re taking two approaches. First, there’s “generative AI for all”, where we’re introducing employees to the technology through meetings, conferences, demonstrations, workshops and e-learning sessions. To date 20,000 Group employees have already received training on these tools. The second approach is creating our own generative AI tool, which began in January 2024. We call it SNCF Group GPT, and it gives employees a secure space where they can upload and manage their documents and data.
SNCF Group: What was the internal development process like?
J.N.: Our e.SNCF Solutions teams developed an innovative in-house generative AI tool designed to assist employees with everyday tasks. It provides quick, accurate responses similar to ChatGPT, but with one key advantage—keeping our data confidential. We’re gradually rolling out this solution to more employees whose jobs require them to use AI, with a customized deployment strategy for each subsidiary. The AI models powering SNCF Group GPT—known as LLMs, or large language models—come from industry leaders, including OpenAI, Anthropic, and the French model Mistral. We select and combine these models based on the specific task, our business needs, and their energy consumption.
SNCF Group: If you’ve developed this tool internally, data sovereignty must be a critical issue for the Group?
J.N.: Absolutely. SNCF Group GPT was built with strict ethical safeguards to ensure that employees’ personal data and the Group’s data are protected. All data is securely stored in our private cloud infrastructure, which guarantees that it’s not used to train generative AI models. Instead, it’s segregated by user and by company.”
For us, generative AI is not just a fad. We’re committed to this technology because we believe in leveraging the best tools available to enhance our efficiency and rail operations.
SNCF Group: Beyond this generative AI “for all” tool, what other applications is the Group working on?
J.N.: One of the most obvious use cases for generative AI is in IT. Our developers already have access to various coding tools, but we’re taking it further with our “augmented developers” project—a program designed to test and implement AI-driven support for both consumer-facing applications like SNCF Connect and internal rail operation systems. Another key area where generative AI is making a difference is human resources. We’ve developed an HR chatbot that leverages our entire HR documentation database, enabling employees to quickly find answers to questions and improving overall HR efficiency.
SNCF Group: How do you decide which areas are suitable for generative AI solutions?
J.N.: Our generative AI steering committee makes those decisions. We’ve identified 10 key use cases where generative AI can make a real difference, particularly in customer relations, document searches, maintenance and passenger information. Two-person project teams consisting of a business specialist and an IT specialist are currently fleshing out these cases.
Real-life applications of generative AI at SNCF:
- Enhancing document search and automating content generation
- Augmenting program/app development and improving coding efficiency
- Modernizing our supplier relations
- Driving interactive passenger information kiosks
- Predicting delays
SNCF Group: When you talk about “real transformations”, what’s the first example that comes to mind?
J.N.: Predictive maintenance is a perfect example. For years, we’ve been fitting our rolling stock with IoT sensors, allowing us to collect vast amounts of data to predict incidents and malfunctions before they occur. With generative AI, this process has taken a major technological leap. By leveraging advanced AI models, we now have greater computational power. This provides our technicentre teams with richer, more detailed and more contextualized information. Which brings us closer to our goal of zero rolling stock malfunctions.
SNCF Group: How does it improve passenger information?
J.N.: We’re now using generative AI to manage passenger flows more effectively by providing real-time updates on RER and Transilien networks, displayed directly on station screens.
Another development is our SNCF Gares & Connexions interactive kiosks, which use Meta’s AI model. These were deployed at Paris-Nord station for the Paris 2024 Games. The kiosks feature a life-size human avatar that interacts with passengers, answering questions about train schedules, itineraries, next departures, and in-station services. Passengers can even continue the conversation on WhatsApp through a chatbot.
SNCF Group: How does our dynamic approach to generative AI enhance SNCF’s appeal?
J.N.: For us, generative AI is not just a fad. We’re committed to this technology because we believe in leveraging the best tools available to enhance our efficiency and rail operations.
This level of service builds trust with our clients, the Transport Organizing Authorities. Generative AI also boosts employee productivity, enabling us to deliver cost-effective, high-quality services.
The AI and Sustainable Mobility Chair we’ve established with École Polytechnique, France’s prestigious science and engineering institution, reflects our commitment to innovation and progress.
SNCF Group: But generative AI is also energy-intensive. What is SNCF Group’s stance on AI’s environmental impact?
J.N.: We take a pragmatic approach. While generative AI requires significant energy, it also helps us increase train traffic, improve service quality and reduce costs—all of which help accelerate a modal shift to rail travel, thus reducing overall transport sector emissions.
Of course, we are fully aware of AI’s energy demands. To minimize its impact, we carefully select Large Language Models (LLMs) that align with our needs, while avoiding unnecessarily energy-intensive AI systems. Lastly, when our employees use Groupe SNCF GPT, it tells them how much CO2 each prompt generates. That helps raise awareness in-house about the importance of responsible AI usage.
Share the article