Legacy Data Systems Cripple Bank AI
Banks are rushing to implement AI for fraud detection and risk analytics, but are being held back by legacy data environments. A new report from Info-Tech Research Group finds that real-time data gaps are a primary structural limitation hindering the ability to scale AI initiatives effectively.
Many of the world's largest financial institutions still run on core banking systems built with COBOL, a programming language over 60 years old. This technology underpins an estimated 95% of ATM transactions and 80% of in-person banking activities, processing a vast number of the globe's business transactions. These legacy systems, often mainframe-based, create significant data silos. Information is stored in fragmented, incompatible formats across different parts of the bank, making it incredibly difficult for modern AI algorithms to access the comprehensive, real-time data needed for effective analysis. The problem isn't just data access but also processing power. Legacy architectures were designed for batch processing and structured transactions, not the real-time, dynamic data streams and unstructured digital interactions that power advanced AI and machine learning models for fraud detection or risk analytics. Nearly three-quarters of banks globally continue to operate on these legacy core systems, with 59% of bankers viewing them as a major business challenge. This reliance creates a significant hurdle, as nearly 60% of banking leaders identify outdated infrastructure as the primary obstacle to business growth and AI adoption. To overcome this, the financial services sector is expected to be the biggest spender on AI solutions through 2028, accounting for over 20% of all AI spending. A key focus of this investment is modernizing data infrastructure by moving away from siloed systems toward unified data lakes and cloud-based solutions that provide the necessary scalability. The talent pool is another critical issue. The average age of a COBOL developer is over 55, creating a looming skills gap as this workforce approaches retirement. This shortage of experts capable of maintaining and integrating these older systems with new AI technologies adds another layer of risk and complexity to modernization projects.
