Digital twins: the first step in legacy modernisation at banks

It is not just ageing technology bank should be worried about, but the loss of institutional knowledge. When systems are no longer fully understood, they cannot be safely changed, documented, or governed. Enterprise application modernisation stalls, regulatory pressure increases, and even routine incidents become harder to resolve. What banks are facing is not simply a technology problem, but a critical knowledge gap at the heart of their IT landscape.

A digital twin addresses this issue directly. Built using modern AI, a digital twin creates a structured knowledge system of the existing core banking environment that can be queried in natural language.

In most large banks, vast data centres are a growing concern for IT leaders. Millions of transactions run on decades-old systems built in languages such as COBOL, C, or assembler, which form the backbone of account management, lending, and payment processing. These monolithic systems continue to run reliably. However, very few people still understand how they work. As experts retire, the legacy systems they built and developed over their careers are becoming increasingly opaque.

This creates a growing business risk for traditional banks. It is not just the outdated technology that struggles to keep pace with real-time processing and other demands of the modern financial system. A lack of understanding around system logic, structures, and dependencies is equally critical and a new source of operational and regulatory risks. For example, regulations such as DORA (EU Digital Operational Resilience Act) require detailed documentation, which is difficult to produce without a clear comprehension of how these core systems function. At the same time, urgently needed modernisation is becoming a significantly more complex undertaking with each passing month. 

A digital twin addresses the complexity challenge by making unclear processes and structures visible through reverse documentation. It is not a second core banking system, but a semantic representation of the existing one. It brings together millions of lines of code, business logic, data flows, and interfaces into a single knowledge model.

Unlike industrial digital twins, the knowledge model does not represent system behaviour, but system knowledge. This knowledge can be queried in natural language, making it accessible, understandable and searchable to a broader set of stakeholders.

The digital twin can be used to analyse dependencies between systems and uncover hidden threats, enabling a tailored risk reduction strategy. For example, the digital twin can answer questions about which applications are critical for payment processing or securities trading, and explain the dependencies between them. It can also assess the impact of regulatory changes or explain why a batch process failed overnight.

Because each bank’s IT landscape has evolved differently, every digital twin must be created to reflect the organisation it is deployed in. Zühlke combines AI-based code analysis with modernisation expertise to build individual models.

This approach enables a digital twin to be created within one to two months, delivering almost immediate value. Straight away, the bank becomes less reliant on the few remaining experts who are still familiar with legacy systems. A knowledge model eliminates the need to manually review outdated documentation and reduces repetitive tasks, improving both efficiency and job satisfaction for involved teams.

The digital twin also significantly improves collaboration. Business analysts, developers, and architects work on a shared knowledge base, which leads to better alignment and faster decision-making. Moreover, it can be integrated into development environments. For example, developers can receive built-in contextual support directly in their code editor. AI agents can handle tasks semi-autonomously. When a ticket is received, they access the knowledge model to check whether the issue has already been resolved and, if not, they identify the most suitable person to respond. They can also gather relevant information in advance, saving time and effort, and accelerating resolution. 

Most importantly, the digital twin is a starting point for the urgently needed transformation of rigid core banking systems into a modular and interconnected architecture. It enables individual components of monolithic systems to be isolated, analysed, and modernised step by step.

This enables banks to circumvent the risks associated with large-scale, one-off migrations. Parallel systems can be developed, changes introduced incrementally, and decisions made with greater confidence. Crucially, progress no longer depends entirely on a small number of legacy specialists.

Digital twins not only make core banking systems understandable again – they represent a gradual, low-risk path towards modernising mission-critical applications. Initial results are visible within a short timeframe, and accumulate into big wins over time thanks to continuous modernisation.