*by Luis Murillo
In large corporations, the presence of legacy systems is often an obstacle to technological evolution, especially in the banking system, where the programming language traditionally used is Cobol. These are applications built years, even decades, ago that continue to support critical business operations.
To give you an idea, according to research by McKinsey, In the financial sector, for example, core banking platforms and investment management solutions process trillions of dollars in transactions daily on a global scale. In the insurance sector, policy administration systems manage US$1.9 trillion in annual premiums in the United States alone. And social benefit management platforms enable the distribution of over US$830 million annually in services and benefits to millions of American citizens.
It's a vast number of systems, given the extensive use of COBOL for the operation of core systems. To classify these systems simply as "old" would be a dangerous oversimplification. In practice, legacy systems accumulate decades of business knowledge, including fiscal rules, operational logic, business flows, and critical processes that have been refined over time. While they function, they follow a model with low upgradeability and difficulty integrating with modern environments.
Many modernization initiatives fail precisely because they treat legacy systems as something to be discarded. The idea of rewriting everything from scratch may seem appealing, but it often ignores an essential factor: the complexity of the knowledge embedded in these systems.
Each line of code represents decisions made over years of operation. Recreating everything manually not only demands significant time and investment, but also increases the risk of losing critical business rules that, in many cases, are not documented anywhere else.
The scenario raises a question: "what got you here isn't what will take you forward," paraphrasing leadership expert Marshall Goldsmith. In other words, the importance of legacy is undeniable, but without the necessary updates, a company will not be prepared for a constantly changing market that demands innovation. The real challenge is preserving this knowledge while simultaneously evolving technologically.
An increasingly adopted approach in the market starts from a different principle: preserving system knowledge at a higher level of abstraction, separating business logic from the programming language. The process begins by extracting the specifications of the legacy system, capturing its rules, entities, and processes. This information then forms a structured knowledge base that describes the application's operation at a higher conceptual level.
Artificial intelligence has been key to supporting these modernization initiatives, especially for code analysis and the generation of new applications. However, the use of generative models requires caution in critical corporate environments, such as financial, industrial, or infrastructure systems.
This is because these tools, based exclusively on code generation, can exhibit hallucinations, meaning that the AI produces plausible but technically incorrect results. More structured approaches, which combine formal knowledge bases with automated generation mechanisms—that is, uniting the use of Generative Artificial Intelligence and Symbolic AI models, which are rule-based—reduce the reliance on probabilistic inferences while increasing the reliability of the final result.
In this model, this set of AIs acts as a mechanism for controlled transformation and generation, and not as a substitute for the structured logic of the system.
This doesn't mean the challenge of legacy systems will disappear. On the contrary, as companies increasingly digitize their operations, the amount of critical software in production tends to grow. The difference lies in how organizations deal with this reality. Instead of seeing legacy as an obstacle, more advanced companies have begun to treat it as a valuable source of knowledge that needs to be preserved and evolved.
Modernizing legacy systems means ensuring that this knowledge is not retained in low-level code – even in a new language – but is stored in a knowledge base, allowing code to always be automatically generated from it.
*Luis Murillo is Technical Support Manager at GeneXus by Globant, a company specializing in Enterprise Low-Code platforms that simplify software development and evolution through Artificial Intelligence.
Notice: The opinion presented in this article is the responsibility of its author and not of ABES - Brazilian Association of Software Companies
