The Future of GenAI Will Drive the Modernization of Legacy Core Systems

Our research shows that companies focused on innovation are the driving force behind the implementation of new technologies.

Specifically, 87% of business and technology professionals with some knowledge of emerging technologies are experimenting, testing, or implementing AI.

The third most common use case for generative AI is software development.

Many users are enthusiastic about the ability to write code in natural language or use an autocomplete editor or pair programming to help them write code quickly.

Financial services organisations such as ANZ and Westpac have conducted internal experiments to prove that software coding can increase productivity by 42-45 per cent.

However, GenAI, as well as AI and machine learning, can collectively achieve much more in the field of software development and maintenance.

Forrester has named the field of AI-assisted software development Turing Bots, after British AI scientist Alan Turing. We see Turing Bots as a new exponential technology that will have a positive impact on the financial services industry in the short and long term.

Turing bots speed up software lifecycle

Turing Bots can support all phases of the Software Development Life Cycle (SDLC), from the initial planning, analysis, and requirements phases, through design, coding, testing, and delivery. There is a TuringBot type for each phase of the life cycle.

Indeed, a rich ecosystem of vendor-driven tools is rapidly emerging in the market, with established software and platform providers like GitHub, Atlassian, Google, and AWS, along with startups and smaller players like Applitools, Coedium, Tabnine, and the well-known OpenAI, all racing to build and integrate Turing Bots into current and future development tools.

More advanced Turing Bot approaches use open large language models (LLMs) specializing in software as data, including Codex, Code Llama, and StarCoder. This technology is on its way, but are we ready? No.

As technology advances at an ever-increasing rate, we, as individuals and businesses, struggle to keep up and change the way we work. Given the resistance of businesses to change, we believe that in less than five years, this technology will completely change the way software, applications, and products are built and maintained.

What about heritage?

When discussing legacy technology, let’s take ATMs as an example. Interestingly, 95 percent of ATM card swipe transactions to withdraw cash are done on old Cobol programs. Cobol stands for common business oriented language and is a computer programming language that has been used primarily in business, financial, and administrative systems since 2002.

It is estimated that more than 220 billion lines of Cobol code are still used in production worldwide.

This is the beginning of a new world in which the costs of migrating legacy systems to modern technology stacks will no longer be prohibitive.

Legacy applications process, execute, and manage customer policies for insurance products and store most of the data that is critical to maintaining the operational efficiency of the business. This data is also critical to modern applications, but it is not easy to access through legacy applications.

Legacy doesn’t just mean Cobol. Even applications built in newer programming languages like Java—including older versions like Java6—are now legacy. Dealing with legacy is a challenge that has slowed modernization for years. The attitude toward legacy has always been, “We have no idea how, but it works, so let’s leave it alone.”

TuringBot application examples can also apply to older versions

This is all about to change as GenAI offers great capabilities to help modernize any type of legacy code, provided the LLMs have been trained or refined on that specific legacy code.

The good news is that any Turing bot that has learned to write and understand code in a different programming language can be improved and adapted to understand code in a new language.

Turing bots – AI-powered assistants that generate code based on English input – work with 10 (and more) of the most popular programming languages, including Java, C, C++, Python, etc. They can help you do many things that we couldn’t do in the past using older versions, at least not without extraordinary effort and time.

Here are some of the most important use cases:

Automatically generate documentation for the legacy program file. Legacy documentation is the first step to unlocking the power of modernizing legacy systems. While this doesn’t and can’t happen at the push of a button, real experiments are happening at scale in large global organizations.
Automatic translation from one programming language to another. Most Turing Bots coders can translate old Java code files to newer Java code, but they can also translate Cobol to Java. For example, IBM enables WatsonX.ai with a 20 billion LLM parameter for Cobol-enabled code.
Automatic test case generation and optimization based on requirements. Test cases are the basis of testing so that the original requirements implemented by the legacy system are also met by the system that has been migrated to the modern language. Tester Turing Bots help in this use case.

Example of automatic migration from Cobol to Turing Bots

The above use cases are already very helpful in the experiments of a large system integrator in migrating from older Cobol versions to modern architectures. In short, LLM-enabled Turing Bots code is used to generate documentation from Cobol (and JCL) programs.

The documents are reworked to define requirements and provide a general understanding of the application architecture, which is an essential starting point for working with any legacy code.

In addition, GenAI woven into the Turing Bots tester generates test cases from documented requirements. In parallel, WatsonX.ai translates Cobol and JCL to modern Java code.

Then, the newly generated Java code is tested manually and automatically using test cases. This process ensures that the new code meets the functional requirements extracted from Cobol.

This semi-automated process is a starting point; development teams can further enhance it with additional steps, such as adding new requirements or capabilities during the document generation process. In other words, further redesign of the old application or business process is possible throughout the process.

This is the beginning of a new world in which the costs of migrating legacy systems to modern technology stacks will no longer be prohibitive.

In the near future, organizations will stop dealing with legacy solutions and start migrating or regenerating them, transforming them into new, more modern applications.

So if you are one of the many companies stuck with legacy systems that you can’t afford to migrate or replace, don’t lose hope. GenAI has the potential to radically reduce migration costs.

Diego Lo Giudice is a vice president and principal analyst at Forrester