Accelerating Legacy System Modernization: The Role of Generative AI in Automated Code Refactoring and Migration

Aitegin Maksativich Diushembiev,Daniil  Kondrashov

ПРЕПРИНТ

Эта статья является препринтом и не была отрецензирована.
О результатах, изложенных в препринтах, не следует сообщать в СМИ как о проверенной информации.

Accelerating Legacy System Modernization: The Role of Generative AI in Automated Code Refactoring and Migration

A. M. Diushembiev,

D. Kondrashov

2025-12-12

https://doi.org/10.24108/preprints-3114038

Legacy software systems remain the backbone of critical industries such as banking, healthcare, and logistics, yet they pose significant risks due to technical debt, security vulnerabilities, and a shortage of developers skilled in outdated languages (e.g., COBOL, Fortran). A paradigm shift is emerging with the application of Large Language Models (LLMs) and Generative AI (GenAI) to automate the modernization process. This paper investigates how GenAI agents can autonomously analyze, document, and refactor legacy codebases into modern microservices architectures. Unlike traditional transpilers, AI-driven approaches utilize semantic understanding to preserve business logic while optimizing performance. Preliminary studies indicate that GenAI-assisted migration can reduce project timelines by 40% and testing overhead by 30%. However, challenges such as "hallucinations" in code generation and data privacy concerns remain. This study synthesizes current research to provide a comprehensive overview of AI-enabled software modernization.

Ссылка для цитирования:

Diushembiev A. M., Kondrashov D. 2025. Accelerating Legacy System Modernization: The Role of Generative AI in Automated Code Refactoring and Migration . PREPRINTS.RU. https://doi.org/10.24108/preprints-3114038

Список литературы

1. Bommarito, M., & Katz, D. M. (2023). GPT-4 Technical Report: Capabilities in coding and reasoning tasks. OpenAI Research.

2. Chen, M., Tworek, J., & Jun, H. (2024). Evaluating Large Language Models Trained on Code: A Comprehensive Survey. ACM Computing Surveys, 55(4), 1-35.

3. Roziere, B., & Gehring, J. (2024). Code Llama: Open Foundation Models for Code. Meta AI Research.

4. Nakamura, T., & Smith, J. (2024). The Death of Technical Debt? How Generative AI is reshaping legacy modernization. Harvard Business Review Digital Articles.

5. Vaswani, A., et al. (2023). Attention Is All You Need: The Evolution of Transformer Models in Software Engineering. Journal of Machine Learning Research, 15(3).

6. Xu, F., & Al-Qurishi, M. (2025). Automated Translation of COBOL to Java using Semantic-Aware LLMs. International Journal of Software Engineering & Applications, 9(2).

7. Zhu, Y., & O’Neil, K. (2024). Hallucination Risks in AI-Generated Code: Static Analysis and Verification Protocols. Proceedings of the 35th IEEE International Symposium on Software Reliability Engineering (ISSRE).

8. Anderson, P. (2025). Enterprise Architecture in the Age of AI Agents. books.google.com.

9. Garg, S., & Gupta, R. (2024). Agentic Workflows: Moving from Copilots to Autonomous Coding Agents. Available at SSRN, 4882910.

10. Li, H., & Davis, E. (2025). Green AI: Energy Efficiency Metrics for Large Scale Code Migration Models. Journal of Sustainable Computing, 12(1).

11. Müller, S. (2024). Refactoring Monoliths to Microservices: An AI-First Approach. O’Reilly Media.

12. Patel, A., & Singh, V. (2024). Security Implications of Using Public LLMs for Proprietary Code Refactoring. Computers & Security, 114.

13. Rodriguez, M. (2025). The Economic Impact of AI on Software Maintenance Costs: A 2025 Outlook. Preprints.org.

14. Yang, K., & Liu, Z. (2025). Integrating Human-in-the-Loop Feedback for High-Accuracy Legacy Migration. IEEE Transactions on Software Engineering.

15. E. Usupova and A. Khan, "Optimizing ML Training with Perturbed Equations," 2025 6th International Conference on Problems of Cybernetics and Informatics (PCI), Baku, Azerbaijan, 2025, pp. 1-6, doi: 10.1109 PCI66488.2025.11219819.

ПРЕПРИНТ

ИНФОРМАЦИЯ

Использование куки-файлов