Banks' Legacy Systems Hinder AI Adoption

The heavy reliance on legacy systems means banks spend up to 70-75% of their IT budgets just on maintenance, leaving little for innovation. This spending is projected to grow from $36.7 billion in 2022 to $57 billion by 2028. This technical debt not only drains resources but also directly hinders the launch of new digital products and services. Many of these core systems, often running on COBOL, were not designed for the data volume and speed required for modern AI applications like real-time payment analysis and mobile banking analytics. Integrating modern AI tools with these decades-old architectures requires complex, costly, and time-consuming customizations that can delay projects by an additional 3-6 months compared to modern infrastructures. The talent pool required to maintain these legacy systems is shrinking and becoming more expensive. Banks report paying 2-3 times more for COBOL engineers than for developers working on modern systems. This creates a two-fold talent problem: a shortage of staff to manage old infrastructure and a separate, intense competition for AI and data science experts who can build new solutions. This skills gap directly impacts a bank's ability to leverage AI for crucial functions. AI engineers often lack a deep understanding of complex financial regulations, while compliance experts may not be familiar with AI model training. This disconnect can slow the adoption of AI in areas like anti-money laundering (AML) and sanctions screening, where deep domain knowledge is critical. The data itself is a major roadblock. Legacy systems often create data silos, where valuable information is trapped in disparate formats across the organization. This lack of high-quality, centralized data makes it incredibly difficult to train and deploy effective AI models for fraud detection or customer personalization, which rely on analyzing vast, unified datasets in real-time. As a result, many institutions are pursuing a strategy of phased modernization, using AI-powered "orchestration layers" to enhance legacy systems without a full, disruptive overhaul. Others are investing in AI overlays, such as those using natural language processing (NLP) for sanctions screening, to improve specific functions while preserving their current IT investments. The goal is to bridge the gap between old and new, enabling more immediate AI adoption while planning for long-term infrastructure replacement.