Legacy Data Gaps Are Stalling AI in Banking
A new report from Info-Tech Research Group finds that while banks are eager to deploy AI for fraud and risk, they are being held back by legacy data environments. The inability to access and process real-time data is a major structural barrier preventing many institutions from scaling their AI initiatives effectively.
The financial services sector is a leader in AI adoption, with roughly 50% of firms actively using it. However, a significant portion, around 75%, are hindered by outdated infrastructure when trying to implement new solutions. This is compounded by the fact that only about 12% of financial institutions have a well-defined and adequately resourced AI strategy. A major part of the problem lies in the continued reliance on legacy systems, with some banks still using COBOL, a programming language from 1959. Shockingly, 43% of banking systems are still built on these older technologies. These legacy systems often create data silos, making it difficult for AI to access and analyze the comprehensive, high-quality data it needs to be effective. In fact, 54% of financial institution leaders see these data silos as a major barrier to innovation. The cost of maintaining these outdated systems is substantial, consuming up to 75% of a bank's IT budget. This leaves little room for investment in new technologies. The financial impact of poor data quality alone is estimated to cost financial sector organizations an average of $15 million annually. Despite these challenges, the push to adopt AI is strong, with financial services expected to account for over 20% of all AI spending between 2024 and 2028. Many institutions are turning to AI for transactional purposes like credit analysis and risk assessment. However, the lack of readiness in their data infrastructure is a critical bottleneck, with only 9.5% of financial institutions feeling "very prepared" to support AI and machine learning.
