AI-powered application rewrite: Revolutionizing legacy code transformation

An article authored by: Stefan Aulbach, Head of Application Modernization Innovation Lab, Deloitte Consulting LLP

Co-authors: Bjoern Langmack, Global App Modernization & Migration Leader, Deloitte Consulting LLP; Matthias Lachmann, Manager, Deloitte Consulting LLP; Martin Eberle, Senior Manager, Deloitte Consulting LLP; Yihong Zeng, Managing Director, Deloitte Consulting LLP

The IT landscape is evolving rapidly, with new technologies and paradigms quickly emerging. One of the most significant advancements in recent years is the advent of AI-powered application rewrite, fueled by large language models (LLMs). This technology is poised to revolutionize the way legacy code is handled, transforming it into modern, efficient, and maintainable applications. In this article, we will explore how AI-fueled application rewrite is transforming the software development landscape and ushering in the next level of abstraction in translating business requirements into code.

Legacy code explained by LLMs

Legacy code often poses a significant challenge for businesses as it is typically difficult to understand, maintain, and adapt to changing requirements. LLMs are now capable of explaining these complex codebases, breaking them down into more manageable components. This explanation process not only provides valuable insights into the inner workings of the legacy system, but also lays the foundation for a more efficient and effective application rewrite.

Application and data decomposition with LLMs

While LLMs can provide invaluable assistance in explaining legacy code, they still require input from business domain experts to fully understand the application and data decomposition. This collaboration between AI and human expertise helps ensure that the decomposition process accurately reflects the needs and requirements of the business. The resulting explanation serves as the basis for creating user stories, which will guide the development of the target-state architecture.

Target-state architecture: Unconfined from legacy constraints

One of the most significant benefits of AI-fueled application rewrite is the ability to model the target-state architecture without the limitations imposed by the legacy environment. This freedom allows developers to design a modern, scalable, and maintainable system that can better serve the needs of the business. The user stories derived from the explanation process serve as the blueprint for this new architecture, confirming that it accurately reflects the desired functionality and performance characteristics.

User stories: Generating code and unit tests for the future-state architecture

With the target-state architecture in place, AI-powered tools can use the user stories to generate code and unit tests for the future-state system. This automation can accelerate the development process and help ensure that the resulting code adheres to established coding conventions. Moreover, the generated unit tests provide a robust foundation for confirming the quality and reliability of the new application.

The metaphor: From assembler to code to business functionality

The metaphor of assembler code to modern programming languages is an apt one for this transformation. Just as assembler code was once the primary means of translating human-readable instructions into machine code, modern programming languages have become the new abstraction for translating business requirements into functional applications. With the introduction of LLMs, organizations are on the cusp of the next leap in abstraction in which business requirements are explained or translated into code by AI-powered systems.

Orchestrating functionality: The future of software development

As AI-powered application rewrite becomes more prevalent, the role of software developers will likely shift from writing code to orchestrating functionality. This evolution will require a new set of skills and expertise, as developers will need to understand how to leverage AI tools effectively and collaborate with business domain experts to ensure that the resulting applications meet the needs of the organization. This new paradigm will likely lead to more efficient, reliable, and maintainable software, ultimately benefiting businesses and end users alike.

AI-fueled application rewrite is set to revolutionize the way legacy code transformation is approached, bringing a new level of abstraction to the translation of business requirements into code. By leveraging the power of LLMs, developers can now efficiently decompose, understand, and rewrite legacy applications, resulting in modern, maintainable systems that better serve the needs of businesses. As we look to the future, the focus will likely shift from writing code to orchestrating functionality, with AI-powered tools playing an increasingly vital role in the software development process. This transformative technology promises to reshape the IT landscape, ushering in a new era of innovation and efficiency.