Industry
Banking
Banking Services Transformation: Unlocking Exponential Value
Highlights
- Mainframe modernisation in banking, financial services, and insurance requires a strategic, workload-by-workload evaluation, classifying applications into golden core, adaptive core, and peripheral categories to determine the most appropriate transformation path.
- A structured six-phase modernisation methodology powered by artificial intelligence addresses key industry challenges across strategy formulation, target architecture design, and execution assurance.
- Integrating artificial intelligence with human expertise accelerates the delivery lifecycle by up to 40% while maintaining robust governance, enabling financial institutions to transition from legacy technology refreshes to sustainable business innovation.
On this page
What are the key challenges and strategies for modernising legacy mainframe applications?
In the banking, financial services, and insurance (BFSI) industry, mainframes will continue to power business-as-usual and change initiatives. Even as firms undertake modernisation initiatives, they recognise that it is neither a simple technology refresh nor a complete shift from the mainframe. For BFSI firms, legacy modernisation is a workload-by-workload decision: they must evaluate their application landscape and decide which should be retained, optimised, re-platformed, refactored, replaced, retired or reimagined.
For BFSI firms, legacy modernisation is a workload-by-workload decision: they must evaluate their application landscape and decide which should be retained, optimised, re-platformed, refactored, replaced, retired or reimagined.
BFSI applications can be classified into three categories (see Figure 1)—golden core (differentiated capabilities), adaptive core (configurable capabilities), and peripheral (routine capabilities). Depending on organisation, business, and IT drivers and the type of workloads, firms typically reimagine or retain the golden core, reimagine or refactor the adaptive core, and refactor or re-platform peripheral capabilities. Some BFSI firms consider ‘replace’ as a viable option even for the golden core and refactor or reimagine only differentiated or business confidential capabilities and features that are not supported by the commercially available off-the-shelf products.
Figure 1: Classification of banking and insurance functions into core, adaptive, and peripheral layers to guide mainframe modernisation
However, the legacy modernisation process comes with several challenges.
Selecting the right strategy: BFSI firms often choose a re-platform or refactor approach without fully understanding the mapping complexities and interdependencies across applications and business capabilities, business criticality, cost of operations, risk, technical debt, and future-state operating models. Consequently, modernisation projects experience misaligned business outcomes, schedule delays, and budget overruns.
Designing the target state: Transitioning to distributed or cloud-native platforms from tightly coupled legacy environments requires firms to design the target architecture, factoring in business domain boundaries, upstream and downstream application integration, transaction handling, security, observability, data ownership, and the operating model.
Managing execution risk: Modernisation programs come with a high degree of execution risk as banks and insurers must ensure that there is no negative business impact across customer experience, product performance, and operational costs. Database migration also needs careful planning for bi-directional real-time replication.
In our view, BFSI firms must adopt a mainframe modernisation strategy underpinned by artificial intelligence (AI) and generative artificial intelligence (GenAI) technologies. The core approach must centre on target state design and execution assurance—while AI adoption will accelerate analysis and artefact development in both areas, human judgment must guide implementation decisions, mitigating the risk of business disruption.
BFSI firms must adopt a mainframe modernisation strategy underpinned by AI and GenAI technologies with human judgment guiding implementation decisions in order to mitigate the risk of business disruption.
The third of a three-part thought paper series on legacy modernisation in BFSI, this paper offers a phased, AI-powered modernisation approach centred on refactoring and reimagining modernisation paths. The first paper explains the approach to building an AI-powered knowledge hub, while the second one dwells on the role of Al in mainframe engineering.
How can an AI-powered, six-phase strategy accelerate mainframe transformation for BFSI firms?
A practical, agentic AI-enabled modernisation program can be organised into six connected phases: modernisation assessment, system discovery, target architecture and operating model, data migration, application code engineering, and testing (see Figure 2). These phases are iterative rather than sequential because evidence discovered during engineering, migration or testing may refine the architecture, roadmap, or scope. While AI expedites the implementation, human validation is a must.
Incorporating agentic engineering into mainframe modernisation can potentially reduce the cycle time by 30-40% across the six phases. Success will depend on combining these capabilities with strong architecture governance, engineering discipline, and business alignment.
Figure 2: A six-phase, AI-powered strategy to mainframe modernisation from assessment to deployment
Let us examine each phase of the modernisation lifecycle in more detail.
BFSI firms must adopt a structured approach to determine the best modernisation path for each application. Every application must be assessed by leveraging AI tools augmented by insights from subject matter experts (SMEs). Dimensions such as business, technology, architecture, cost, resilience, and risk must also be factored into the decision-making. For example, deposit processing and commercial lending are core banking capabilities and may demand a rearchitect to unlock agility and business value. For peripheral capabilities such as statements and reporting, re-platform or refactor may be adequate instead of a complete rebuild.
Mainframe applications in banks and insurance firms have evolved over decades and documentation is frequently missing or incomplete. Consequently, firms lack a clear understanding of functionalities, business rules, and interdependencies across applications. AI speeds up the discovery process and quickly captures application related information and the business processes they support. These insights are stored in a mainframe knowledge hub, enabling infrastructure and business teams to search and trace the findings to the source. Application owners and production support teams validate inferred boundaries, exceptions, and undocumented operating procedures, contributing to the selection of the right R-path, design, migration sequencing, and test scope.
Architecture teams must leverage AI to analyse application characteristics, dependencies, and business processes to arrive at design options for the operating model. AI can help identify business domains and technology boundaries as well as architecture patterns. In addition, AI can help evaluate the various architecture options and offer recommendations. However, architects, platform owners, security and operations teams, and business stakeholders must consider performance, cost, and long-term operations and maintainability while choosing the target architecture and operating model.
A clear understanding of how databases, files, copybooks, batch feeds, and downstream applications use business information forms the foundation of data migration. Given that business rules and entity relationships are rarely well-documented, firms must use AI to extract embedded semantics, pinpoint data quality anomalies, and draft source-to-target schema mappings for modern cloud databases. Domain experts must validate these mappings and execute them through automated, version-controlled pipelines. To mitigate the high degree of execution risk, firms must run the new and legacy systems simultaneously and ensure that the same business outcomes are delivered by both, before completely switching to the target model. This may involve comparing totals of account balances, transaction counts, premiums, claims or payments across both systems to identify errors, prevent duplication or sync conflicts, and maintain stringent system-of-record ownership.
AI can expedite software development by generating code, documentation, and design artefacts. However, architects and engineering teams must define business requirements, user needs, application functionalities, and architectural principles to guide AI agents in engineering code components, service interfaces, and unit tests. The real challenge lies in demonstrating that the target state behaves in the same way as the legacy system, delivering identical outcomes under real-world conditions. For example, functions such as interest and premium calculations and account balances must remain identical; similarly, the same business rules must underpin transaction processing—even a minor discrepancy can adversely impact business-as-usual and cause compliance issues. In other words, the new model’s ability to deliver the same business outcomes, efficiently manage transaction volumes, maintain service quality, and quickly recover from failures will determine if it will be deployed into production.
AI can accelerate overall testing through faster test case generation, regression suite optimisation, synthetic test data creation, and validation of outputs against the legacy environment. AI can also assist in reconciling data across legacy and target environments and identifying discrepancies that would otherwise require significant manual effort. In our experience, successful migration to the target state demands meeting acceptance criteria across accurate functional performance, data integrity, security, batch completion, and system integration. Firms must take the decision to migrate to the target state only when there is adequate evidence that it meets predefined quality, performance, and operational standards. This evidence-based approach improves confidence in the target state.
In our experience, incorporating agentic engineering into mainframe modernisation can potentially reduce the cycle time by 30-40% across the six phases. Success will depend on combining these capabilities with strong architecture governance, engineering discipline, and business alignment.
What does AI-assisted mainframe modernisation look like for a global investment firm?
A global investment management firm embarked on a modernisation initiative to migrate critical retirement and investment administration batch workloads from a legacy mainframe platform to cloud-native architecture. Using an AI-assisted modernisation strategy, we transformed the complex legacy platform, ensuring that the new solution performed exactly like the legacy platform. With this implementation, the firm realised several benefits:
- Accelerated processing performance by up to 300%.
- Reduced infrastructure and operational costs through cloud adoption.
- Minimised migration risk through automated validation and testing.
- Established a scalable modernisation foundation for future transformation programs.
What is the future of mainframe modernisation and how can BFSI firms begin their transformation journey?
Mainframes will continue to play a critical role in the technology landscape of banks and insurers for years to come. The future lies not in full-scale exit, but in adopting the right modernisation path for each mainframe workload. As the AI landscape matures, the industry will see the rapid advent of more capable models. With the use of AI and agentic IDEs, the cost, risk and time constraints across modernisation options are reducing swiftly. It is only a matter of time before BFSI firms will be able to leverage the full potential of AI to accelerate their modernisation journey.
In the next few years, we expect banks and insurers to have incrementally modernised their legacy estate—while small firms may completely exit mainframes, others may adopt hybrid modernisation.
In our view, BFSI firms must start their modernisation initiative with low-risk peripheral applications and slowly move to applications in the adaptive and golden core. In the next few years, we expect banks and insurers to have incrementally modernised their legacy estate—while small firms may completely exit mainframes, others may adopt hybrid modernisation. The ultimate objective will shift from technology migration to business transformation, enabling banks and insurers to innovate faster, respond more rapidly to regulatory change, and deliver increasingly personalised, AI-enabled customer experience.