AWS re:Invent 2025 - Modernize Legacy Applications with Agentic AI (DEV333)

Modernizing Legacy Applications with Agentic AI

The Challenge of Legacy Applications

• 65% of enterprise applications are now considered "legacy" - relying on outdated technology, hard to maintain and extend
• Legacy applications typically consume the majority of IT budgets, which is expensive
• Traditional modernization efforts are often long, manual projects

The Agentic AI Approach

• AI agents can perform reasoning, use tools, and execute multi-step workflows to dramatically increase the speed and quality of modernization tasks
• Key characteristics of AI agents that enable application modernization:
  • User interface for understanding goals
  • Ability to decompose goals into executable tasks
  • Access to tools, knowledge bases, and feedback loops

Modernization Patterns and Decomposition

• Rehosting, replatforming, refactoring, and rearchitecting are different modernization patterns
• Technical decomposition (e.g. monolith to microservices) and business decomposition improve agility and flexibility
• Strangler pattern enables incremental migration from legacy to modernized applications

AI-Powered Modernization Capabilities

Reverse Engineering and Documentation Generation

• AWS Transform can automatically analyze legacy code, extract business logic, and generate technical documentation
• Leverages large language models (LLMs) hosted on Bedrock to create specification documents from source code
• Provides a human-in-the-loop review process to validate the generated documentation

Runtime and Dependency Upgrades

• AWS Transform Custom can automatically upgrade Java, Spring Boot, and AWS SDK dependencies in legacy applications
• Generates a detailed transformation plan, executes the upgrades, and validates the results
• Provides granular visibility into the changes made and any build errors encountered

Cross-Platform Replatforming

• Kira, the new agentic IDE from AWS, can replatform .NET 3 applications to .NET 8 for cross-platform portability
• Leverages steering files, MCPs, and a knowledge base hosted on Bedrock to ensure the generated code adheres to corporate standards
• Provides a multi-agent architecture using the open-source Strands framework for orchestrating the modernization workflow

Refactoring and Reforging for Maintainability

• AWS Transform can refactor mainframe COBOL applications into Java, including transforming VSAM data stores, JCL scripts, and BMS screens
• The "Reforge" agent further improves the generated Java code for better readability and maintainability using LLM-based techniques

Operationalizing Modernized Applications

• The Kira CLI can be used to investigate and troubleshoot issues in modernized applications running in production
• Provides access to cloud resources and logs, allowing the CLI to diagnose problems and provide recommendations
• Demonstrates the power of agentic AI for ongoing operations and support of modernized applications

Key Takeaways

• Agentic AI can dramatically accelerate and improve the quality of legacy application modernization tasks
• AWS provides a range of tools and services, including AWS Transform, Kira, and Strands, to enable AI-powered modernization
• Modernization patterns like rehosting, replatforming, and refactoring can be automated and optimized using AI agents
• AI agents can handle reverse engineering, documentation generation, runtime upgrades, replatforming, and refactoring for improved maintainability
• The operationalization of modernized applications can also be enhanced through agentic AI capabilities

Next Steps

• Explore the AWS Agent AI Solutions to learn more about the capabilities demonstrated
• Review customer stories and modernization blueprints to understand real-world applications
• Consider scheduling an Experience-Based Acceleration (EBA) workshop to apply the techniques to your own legacy applications