Artificial intelligence (AI) is opening a new path for banks seeking to modernise ageing software without disrupting the systems that keep their operations running. From processing payments to managing customer accounts, decades-old applications remain central to financial services, even as institutions face growing pressure to deliver faster, more flexible digital experiences.
The challenge is not simply replacing old code with newer technology. Banks must preserve the business rules, transaction records and interconnected processes that have supported their operations for years. An error during this transition could affect payments, account balances and customer access to essential services.
According to recent reporting on OpenAI’s efforts to improve legacy software modernisation, advances in AI are creating fresh opportunities to understand and maintain older programming languages, including COBOL, Fortran and Visual Basic.
For financial institutions, this development could make modernisation less expensive and time-consuming. However, the priority remains clear: improve the technology without compromising the reliability customers expect from their banks.
Also Read: African Banks Urged to Unlock More Value From Existing Payment Technology
AI Tools Could Reduce the Risks of Banking Software Modernisation
OpenAI says its latest models have improved at writing and translating software in older programming languages, potentially helping banks tackle systems that have become difficult to maintain. In an October 19 announcement reported by The Economic Times, the company highlighted progress in handling legacy code and preserving existing software behaviour during migration.
The implications are significant. A bank does not necessarily need to discard an entire platform to improve its capabilities. AI can help engineers examine existing applications, document their logic, identify dependencies and develop replacement components in stages.
For example, a bank could begin by modernising a reporting application or introducing an interface that allows older software to communicate with newer digital services. Engineers could then compare the new component’s output with that of the existing system before expanding the changes.
Google Cloud has outlined a similar approach in its guide to AI-powered mainframe modernisation, emphasising assessment, incremental transformation, risk reduction and data migration.
This approach offers a more controlled alternative to replacing an entire banking platform in one major exercise. It also gives financial institutions an opportunity to identify problems early, before changes affect a wider range of services.
Why Banks Cannot Afford to Get It Wrong
The difficulty lies in the intricate relationships between banking applications. A seemingly minor software change can affect transaction processing, customer records, reporting systems and other services that depend on the same underlying data.
According to the Federal Reserve Bank of Kansas City, many financial institutions continue to operate core banking systems that are decades old. These platforms remain essential because they support fundamental activities such as maintaining accounts and processing financial transactions.
AI-generated code therefore requires more than a successful technical review. Banks must establish that the modernised application produces the correct results under different operating conditions, including unusual transactions and unexpected system failures.
Security, regulatory compliance, data protection and disaster recovery must also remain central to the process. Independent testing and human approval are necessary before new software is allowed to handle live transactions.
For Nigerian banks, where customers increasingly depend on mobile applications, transfers and digital payment channels, an unsuccessful migration could quickly undermine confidence. Modernisation must improve service delivery without introducing avoidable interruptions.
Back Story: The Legacy Systems Behind Modern Banking
Legacy banking software is not necessarily ineffective simply because it is old. Many systems have proved dependable over decades, and replacing them can be expensive, particularly when their original documentation is incomplete, or the people who developed them are no longer available.
COBOL, a programming language developed in the late 1950s, remains associated with major transaction-processing environments. Over time, banks have added new features and business rules to existing applications, creating complex systems that can be difficult to understand without detailed technical knowledge.
AI could help bridge this knowledge gap by translating unfamiliar code into plain language, generating documentation and assisting developers with carefully controlled changes. However, understanding what a programme does is only the beginning. The real test is whether a modernised system can preserve every important calculation, decision and transaction outcome.
Ultimately, AI offers banks a practical opportunity to modernise essential software without abandoning systems that continue to serve them. Success will depend not on how quickly AI can rewrite old code, but on how effectively financial institutions combine automation, expert oversight and rigorous testing to protect the services customers rely on every day.
Also Read: Artificial Intelligence Is Reshaping South African Banking as Banks Move Beyond Pilot Projects
