Why Data Fragmentation is Hiding AI’s Potential in Banking — and How Semantic Graphs Can Fix It

Banking institutions are grappling with a paradox: while generative AI (GenAI) promises transformative capabilities, many organizations report underwhelming results from pilot projects. The root cause? Data silos, not the technology itself.

The AI-Data Paradox: Quality Over Quantity

Artificial intelligence systems, including GenAI, rely on high-quality, context-rich data to deliver value. A machine learning model is only as good as the data it’s trained on. When data is fragmented across departments, the AI loses its ability to discern meaningful patterns.

Banking data often exists in isolated systems. This fragmentation leads to “AI landscapes that are more disjointed than the legacy IT systems,” which have long been a source of complaint.

Semantic Graphs: A 1950s Solution for Modern Problems

One promising approach to overcoming data silos is semantic graphs, a concept dating back to the 1950s in computer linguistics. These structures—essentially networks of interconnected data nodes—allow organizations to map relationships between structured and unstructured data without physically moving information.

Semantic graphs act as a way to link disparate data sources through domain-specific ontologies. This enables AI models to access contextually rich data while preserving the integrity of existing systems.

The European Data Warehouse, established in 2012 following the financial crisis on the initiative of the European Central Bank, exemplifies this approach. By collecting loan-level data for asset-backed securities transactions in Europe, the European Data Warehouse provides a structured dataset for analysis.

Challenges in Implementation

Technical complexity is a factor. Unlike traditional data lakes, semantic graphs require domain experts to define relationships between data points. It is about encoding business logic into the structure.

The Human Factor: Culture Over Code

Even the most advanced technical solutions fail without organizational buy-in. Technology alone cannot bridge silos; leaders must foster a culture of cross-departmental collaboration and interdisciplinary teamwork.

Looking Ahead: The Future of AI in Banking

As generative AI evolves, the need for integrated data strategies will grow. The future belongs to institutions that treat data as a strategic asset. Semantic graphs offer a path forward, but success will depend on technical innovation and cultural transformation.