Skip to main content

Expert opinion

AWS Transform Event: A Technical Deep Dive into AI-Assisted Legacy System Modernization

Published on 08/27/2025


Our TeamWorker Yahya Margad recently participated in the AWS Transform workshop, which focused on Amazon’s approach to modernizing applications using agentic AI. This post covers the technical architecture, capabilities, and practical implementation details across three main areas of transformation.

Architecture Overview

AWS Transform operates on a foundation of specialized AI agents that integrate with existing enterprise systems through secure connectors. The platform connects to code repositories and S3 storage, creating a processing pipeline that can analyze, assess, and transform workloads across VMware, .NET, and mainframe environments.

The service represents an evolution of capabilities previously demonstrated as Amazon Q Developer transformation features at re:Invent 2024, now available as a generally accessible service with core assessment and transformation features provided at no additional cost.

VMware Infrastructure Transformation

The VMware transformation component uses multiple data collection methods to build a comprehensive inventory of virtual infrastructure. The system deploys agents that perform automated application discovery and dependency mapping, creating a detailed view of the existing environment including databases and application relationships.

The analysis phase leverages machine learning techniques, specifically Graph Neural Networks, to understand complex interdependencies between workloads. This approach enables autonomous grouping of applications into optimized migration waves, reducing the manual effort typically required for migration planning.

A significant technical capability is the automated translation of VMware network configurations to AWS equivalents. The system converts existing virtual networking setups to VPC architectures, including subnet planning, security group definitions, and routing configurations. According to AWS documentation, this process delivers 80x faster conversion compared to manual network design approaches.

The platform provides EC2 instance selection and right-sizing recommendations based on analysis of current VM resource utilization patterns. The system aims to maintain application performance while optimizing costs through appropriate instance family selection and sizing.

.NET Application Modernization

The .NET transformation process begins with secure repository integration and encrypted code ingestion. The system performs iterative analysis, starting with an initial code scan to identify the application structure and dependencies.

Dependency resolution is handled through multiple analysis passes. When the system identifies missing dependencies or NuGet packages, it requests their upload and performs additional iterations until a complete dependency graph is established. This approach ensures comprehensive understanding before transformation begins.

The platform focuses on enabling cross-platform deployment of .NET Framework applications. According to verified performance metrics, the system delivers 4x faster application porting from Windows to Linux environments. The transformation process addresses architecture compatibility across x86 and ARM platforms.

The modernization effort targets cost reduction through elimination of Windows licensing requirements and improved scalability through cloud-native architecture patterns.

Mainframe Application Transformation

Current mainframe transformation capabilities support COBOL business logic and JCL (Job Control Language) batch processing scripts. The system ingests legacy code from S3 storage and performs comprehensive static analysis.

The analysis phase includes file categorization, line count metrics by file type, and program complexity assessment. This data provides the foundation for transformation planning and effort estimation.

The platform automatically generates technical documentation for COBOL programs, including program functionality descriptions, business logic explanations, and integration point mappings. This documentation capability addresses one of the primary challenges in mainframe modernization: understanding decades-old business logic without original documentation.

COBOL business logic is converted to Java applications, while JCL scripts are transformed to Groovy for modern job scheduling environments. The system performs intelligent decomposition of monolithic applications, organizing code into business domain-specific modules and feature-based components.

A final optimization pass refactors the generated Java code for improved maintainability and performance characteristics.

Toyota Motor North America implemented the mainframe transformation service for 40-year-old supply chain applications. The implementation generated comprehensive documentation for hundreds of COBOL modules in under 24 hours, providing context-aware information and business process insights that enabled modernization planning for previously intractable legacy systems.

Performance Metrics and Validation

AWS provides specific performance benchmarks for transformation activities:

These metrics are based on customer implementations and represent measurable improvements over traditional modernization approaches.

The platform operates within AWS’s existing security and compliance framework, with encrypted data processing and secure integration points. The agentic AI approach means the system can adapt to different code patterns and business logic implementations without requiring extensive configuration or training.

For organizations evaluating modernization approaches, AWS Transform provides assessment capabilities before committing to transformation projects. This assessment phase identifies potential issues, estimates effort requirements, and validates transformation feasibility.

The workshop session provided hands-on experience with sample COBOL code transformation. The demonstration showed the complete pipeline from code ingestion through documentation generation and Java code output. The process maintained functional equivalence while restructuring the code for modern deployment patterns.

Key technical observations from the workshop included the system’s ability to preserve complex business logic relationships during transformation and its approach to handling legacy programming patterns that don’t have direct modern equivalents.

Conclusion

AWS Transform represents a technical approach to enterprise modernization that addresses traditional challenges through AI automation. The platform’s ability to handle assessment, planning, and execution phases with minimal manual intervention could significantly change the economics and risk profile of large-scale modernization projects.

For technical teams managing legacy infrastructure and applications, the service provides concrete capabilities for addressing modernization backlogs that have been technically or economically impractical using traditional approaches.