AWS re:Invent 2025 - Modernizing Mercedes-Benz’s Global Ordering System with Gen AI (IND218)

Modernizing Mercedes-Benz's Global Ordering System with Gen AI

Overview

• Mercedes-Benz is undertaking a major project, called "Go to Cloud", to migrate its critical global ordering application from a mainframe to the cloud
• The global ordering application is the "heartbeat" of Mercedes-Benz's sales operations, supporting over 8,000 users in 150 countries
• The application is massive in scale, with over 20,000 interfaces, 5 million lines of Java code, and processing over 5.1 billion messages per year on the mainframe

Replatforming Approach

• Mercedes-Benz chose a replatforming (or rehosting) approach to migrate the application to the cloud, rather than a full refactoring
• This approach aims to keep the application as untouched as possible while moving it to a new platform
• The project is being led by Capgemini as the general contractor, working with AWS as the cloud provider and Rocket Software for mainframe emulation
• The migration is being done in stages, starting with stateless services to validate the new platform before migrating the full application

Leveraging Generative AI

• Mercedes-Benz and Capgemini are using Generative AI, specifically a tool called "Gen Revive", to assist with the migration and modernization process
• Gen Revive uses a multi-agent AI system to automatically transform legacy Cobol code into Java, reducing manual effort
• In one example, the AI was able to migrate 1.3 million lines of Cobol code for a pricing service to Java in a matter of months
• The AI-generated code was able to match the performance of the original mainframe application, with some manual tuning required for database access optimization

Integration and Testing

• A key component is the "Global Ordering Facade", a gateway built using Mulesoft that allows routing traffic to either the mainframe or cloud-based application
• This facade enables parallel testing and gradual cutover, minimizing disruption to downstream systems
• Automated testing is performed by comparing results between the mainframe and cloud-based systems, ensuring parity in functionality and performance

Business Impact

• The migration to the cloud is expected to deliver significant cost savings compared to the mainframe
• It also lays the foundation for further modernization and optimization of the global ordering system
• The successful migration of the pricing service demonstrates the potential for using Generative AI to accelerate legacy modernization projects

Lessons Learned

• Thorough assessment and understanding of the existing application is crucial before starting the migration
• A phased, incremental approach helps manage risk and validate the new platform
• Leveraging Generative AI can dramatically speed up the transformation of legacy code, but requires careful integration with human expertise
• A robust integration layer and automated testing are key to ensuring a smooth transition and maintaining business continuity