Report: Data Gaps Hinder AI Adoption in Banking
A new report from Info-Tech Research Group finds that legacy data environments and a lack of real-time data are hindering AI initiatives at scale within the banking industry. Banks are struggling to move beyond pilots in areas like fraud detection and risk analytics due to these structural limitations.
The struggle to scale AI in banking is deeply rooted in decades-old infrastructure. Many core banking systems still run on mainframe technologies and programming languages like COBOL, which were never designed for the high-throughput, real-time data access required by modern AI and machine learning models. Integrating modern APIs and data pipelines with these legacy systems often requires complex and costly custom middleware, significantly slowing down development and deployment. A recent survey highlights the data infrastructure deficit: only 9.5% of financial institutions report being "very prepared" to support AI and machine learning with their existing data infrastructure. This lack of preparedness is a significant bottleneck, with 48% of institutions describing themselves as only "somewhat prepared." The issues range from inconsistent data formats and missing values to information being siloed across disparate, non-integrated systems. Data quality issues are a primary obstacle, with one report indicating that data cleansing and normalization can extend AI project timelines by 25-40% if not planned for properly. For machine learning models, which are only as good as the data they are trained on, this "garbage in, garbage out" problem can lead to inaccurate insights and flawed decision-making, especially in critical areas like risk assessment and fraud detection. While 84% of financial institutions believe AI offers significant benefits, only 12% have a well-defined roadmap for its deployment. A significant portion of AI initiatives in the banking sector are focused on back-office operations and cost savings rather than on developing new products or enhancing customer experiences. This cautious approach is partly due to the challenges of navigating a complex regulatory landscape that demands transparency and explainability in automated decision-making. Despite these hurdles, the banking industry is a major driver of AI research. Over the past five years, the annual AI research output from banks has increased sevenfold. However, this research is heavily concentrated, with just five North American banks—JPMorgan Chase, Capital One, RBC, Wells Fargo, and TD Bank—accounting for nearly 65% of all AI research papers published in the last year.
