Alstom integrates Generative AI to its business processes with Azure AI technologies

“Industrials are facing existential and structuring challenges: how can they remain competitive, continue to lead the market share, continue to innovate while executing the operational backlog, be agile and adapt to the new and growing market needs, and meet customer requirements while maximizing profitability?” explains Mohamed Habib Mazouni, Alstom AI Innovation and Development Director.

The new capabilities of Generative AI to augment the workforce

Business opportunities, contract, specification, design, implementation, manufacturing, testing, supply chain, installation, commissioning, maintenance… Alstom is integrating AI at all stages of its value chain. Since 2020, Alstom has been driving a disruptive innovation with AI and Generative AI to anticipate the new era of Engineering 4.0, Augmented Workforce, and intelligent, autonomous products and services. “At Alstom, we believe that AI is critical, and we embrace the idea that employees will be leveraging generative AI-based co-pilots to generate engineering assets such as specification, software code, design mock-ups, faster, in a cost-efficient and sustainable way” adds Mohamed Habib Mazouni. “Many other co-pilots integrated at all stages of the value chain – into testing, to manufacturing, supply, installation, maintenance processes – can significantly accelerate software development by reducing the time required for various tasks such as documenting, optimizing, refactoring an existing code, or developing a new code” he adds.

The in-house tool developed by Alstom – turned this vision into tangible actions and use case development across businesses tailored to their very specific needs. The tool acts as an enabler supporting programs and engineering projects as well as transversal services like HR, learning, finance, project management.

One concrete example of its capabilities is regarding how AI is supporting engineering teams writing specifications for the development of a railway system to then be able to implement them. “A poorly written requirement costs an average of one day's salary" explains Mohamed Habib Mazouni. Specification quality is checked using fundamental AI service and then rewritten and even automatically generated using dedicated Generative AI services implementing Azure OpenAI Service. This allows us to gain 25% of the overall quality and consequently reducing the costs of non-quality just mentioned.

With over a thousand users and more than 15,000 operations conducted each month, Alstom’s AI services proves to be successful, grows rapidly and keeps evolving.

An AI approach designed to scale up over time

The railway industry is both specific and fluctuant, creating the needs to develop in-house applications. For deployment at scale, Alstom relies on Microsoft technologies by leveraging the full power of Azure, from intelligent and cloud-native application hosting solutions such as Azure Virtual Machines, Azure Kubernetes, Azure Cosmos DB, GitHub Copilot. This rich portfolio makes it easy to scale and manage scalable and resilient applications.

Alstom is now using a variety of services enhanced by generative AI and Microsoft Azure OpenAI Service. At the top of the list: content generation, translation, and document intelligence, which makes it possible to massively process data from forms and documents. “Microsoft Azure AI is what a turbo is to a race car” says Mohamed Habib Mazouni. “This technology allows you to go faster, do a lot more, and be in pole position.”

“We have a very strong and proactive partnership with Microsoft. Microsoft technologies are outstanding in terms of scalability and integrability. that’s for why, it was a natural choice to continue with this partner.” says Mohamed Habib Mazouni.

The success and impact of most digital transformation projects depends on how well they are adopted by end-users. To ensure that initiatives launched on application innovation, data & AI positively impact business, Alstom leveraged some models such as Sociocracy 3.0 for the governance and ADKAR for Cultural Change & Adoption at scale. These two models aimed at reducing the PoC factory syndrome. “The final advice to diffuse the innovation is so simple: be resilient and surround yourself with early adopters to create a snowball effect. At a second stage, you onboard the hesitant or neutral population, and finally, you welcome the laggards without showing them that they were mistaken in not believing in AI or in your ideas” concludes Mohamed Habib Mazouni.

The company continues to explore new avenues and push new services. There is a roadmap of continuous development and will continue over the next three years. “The best is yet to come” he concludes.