How To Rapidly Transform Legacy Apps Using Large Language Models

CTO of Softengi with 30 years of experience in software development, business applications implementation and digital strategy creation.

getty

According to a 2nd Watch survey, 80% of enterprises continue to run at least one-quarter of their business processes on legacy apps and systems. Legacy apps can be found in various industries and sectors, including government agencies, financial institutions, healthcare organizations and large enterprises. These organizations may have invested significant resources in developing and maintaining these applications over the years. However, replacing or modernizing them can be a complex and costly process.

According to VentureBeat, 79% of companies believe that legacy applications are hindering their organizations' digital transformation initiatives. These organizations may have invested significant resources in developing and maintaining these applications over the years. However, replacing or modernizing them can be a complex and costly process. Besides hindering the pace of digitalization, there’s a financial factor—transforming legacy apps can enable businesses to cut total costs by up to 50%.

While many organizations strive to upgrade or replace their legacy apps to take advantage of newer technologies and improve efficiency, there are still numerous legacy systems in use due to pain points such as established functionality, integration and interoperability, user familiarity, and critical assets.

Key Assets Of Legacy Apps

A company that has already overcome the above-mentioned considerations on the path to modernizing its legacy app should remember that there are two major assets any legacy app possesses: historical data and business logic.

These assets should be extracted from a legacy app and then transferred to a new app to ensure smooth business operations.

Data

Legacy apps typically store significant amounts of historical data accumulated over years or even decades. They often contain embedded business knowledge that may not be fully documented elsewhere.

Although it is a tedious process, there are dozens of data migration tools available that can aid in data transformation and related tasks (e.g., IBM InfoSphere, Oracle Data Migration, Azure DMS, AWS Database Migration Service).

And while there are plenty of methods and tools that can help transfer various data from legacy apps, business logic, on the other hand, is a more complex matter.

Business Logic

Legacy applications often contain intricate business logic that has been developed and refined over time. This business logic represents the rules, processes and algorithms that govern how the application functions and handles data. It encapsulates the unique workflows, decision-making processes and industry-specific requirements of the organization.

The preservation of business logic is a significant consideration when dealing with legacy apps. Organizations often find it difficult to transfer this logic to modern systems without significant effort. Rewriting or re-implementing the complex business rules and processes in a new application can be time-consuming and expensive, introducing potential risks or errors.

One of the methods of extracting business logic from a legacy app is reverse engineering. Using such a large language model (LLM)-based AI tool such as Copilot, it is possible to translate existing business logic into requirements that could be updated and used to develop a new app.

Transforming Business Logic Into Requirements Using Copilot

Modern LLM-based AI tools (e.g., ChatGPT and Jurassic-2) offer the capability to generate initial requirement artifacts based on the analysis of business logic. By leveraging machine learning algorithms and predefined templates, these tools can automatically draft user stories, use cases or functional specifications.

On the other hand, tools such as Copilot or ChatGPT can generate a narrative description of the current code.

In general, the process of extracting requirements from legacy code will follow these steps:

• Generate detailed comments for the code.
• Enhance the abstraction level of the generated comments.
• Create use cases and enrich them by leveraging historical data analysis.
• Construct actual user flows based on user's activities (logs) analysis.
• Produce user stories by increasing the abstraction level of use cases and user flows.

As a result of these steps, we can retrieve the as-is requirements to the current legacy applications. While the generated requirements may call for refinement, they can serve as a starting point, potentially saving time for the business analyst.

Moreover, it is possible to generate use cases in different annotations—UML diagrams as a code or Gherkin annotation to receive a new asset in the form of business logic.

What’s Next?

Having this new asset, we can update the requirements for business logic and start developing a new app with a new design. It could be performed with traditional development approaches and methodologies or using AI tools such as ChatGPT, Copilot or AWS CodeWhisperer.

It is important to remember that while AI tools can offer efficiency gains, such assistants have their limitations. AI tools heavily rely on the quality and comprehensiveness of the input data, such as code, documentation, and user insights. Human expertise is still crucial for interpreting, validating and refining the output of AI tools to ensure the requirements accurately capture the business logic.

Ultimately, the speed of transforming business logic into requirements with AI tools depends on the complexity of the legacy system, the availability and quality of data, the capabilities of the AI tools being used, and the collaboration between human analysts and AI-driven processes.

Summing Up

When modernizing legacy apps, business logic is one of the most valuable assets. With their well-established and tested business logic, legacy apps enable organizations to maintain continuity in their operations and support specific business needs effectively. However, it's important to strike a balance between the value of the existing business logic and the limitations and drawbacks of the legacy technology stack in terms of security, scalability and maintainability.

Although today it seems painful and time-consuming, preserving existing business logic and transferring it without risk into a modern app is possible with the help of LLMs and AI tools. It is effectively translated into requirements that can be updated, if needed, and implemented in a new app.

---

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

---