Industry
Banking
Banking Services Transformation: Unlocking Exponential Value
Highlights
- Mainframe systems are indispensable for BFSI firms but face critical challenges, including an SME skills gap and a two-speed IT divide; adopting an AI-augmented operating model is crucial to enhance agility and business value.
- The transition to an AI-augmented model is built on three core pillars: evolving roles to a human+AI paradigm, democratising talent through AI-powered knowledge hubs, and enabling agile service delivery with integrated AI tools.
- Ultimately, the AI-augmented mainframe operating model must evolve beyond IT productivity gains into a unified business ecosystem, enabling banks and insurers to accelerate innovation and strengthen resilience in an AI-native landscape.
On this page
Why should BFSI firms adopt AI in the mainframe operating model?
In the banking, financial services, and insurance (BFSI) industry, mainframe technologies support mission-critical applications such as policy administration, core banking, and trade processing. Completely replacing legacy platforms is not possible in the BFSI industry because mainframe systems are key to data gravity, security, reliability, regulatory control and transaction scale. Artificial intelligence (AI) and generative AI (GenAI) technologies can unlock value by making mainframe systems more explainable, accessible, and adaptable to changing business priorities.
Completely replacing legacy platforms is not possible in the BFSI industry because mainframe systems are key to data gravity, security, reliability, regulatory control and transaction scale.
AI is no longer only an IT productivity tool, it is driving a fundamental redesign of how mainframe activities can be performed across engineering, operations, governance, and talent enablement.
While business expectations around agility and productivity continue to rise, delivery models remain fragmented, causing friction points around people and processes, affecting agility. Furthermore, sustaining mainframe platforms is becoming increasingly difficult as subject matter experts (SMEs) retire. To address these pain points and realise higher business value, BFSI firms must move towards an AI-augmented mainframe operating model. Besides enhancing the software delivery lifecycle (SDLC), such a model will bridge the gap between stable core technology and the new AI-enabled workforce, facilitating effective human+AI collaboration.
What is holding back BFSI firms from an AI-augmented mainframe?
The shift to a more efficient mainframe operating model is hindered by three primary friction points.
- The SME bottleneck: Deep system knowledge resides in a small pool of experienced mainframe SMEs. As these experts retire, undocumented business logic and operating knowledge create continuity risk, while newer talent is often reluctant to adopt legacy technology stacks.
Force-fitting modern AI capabilities, tools and, governance models, into legacy processes, especially across functions such as insurance policy administration or core banking, may fail to accelerate mainframe service delivery.
- The two-speed IT divide: Certain distributed teams that follow modern agile and DevOps practices are highly agile compared to mainframe teams that follow the waterfall methodology. This dual-speed approach creates a divide in the release cycle, delaying the rollout of new product features because core banking and insurance functions, such as deposits, credit, underwriting, and claims management, still reside on mainframe platforms.
- The tooling trap: Force-fitting modern AI capabilities, tools and, governance models, into legacy processes, especially across functions such as insurance policy administration or core banking, may fail to accelerate mainframe service delivery. While AI can generate code or extract business rules in minutes, human review, validation, and approval, which can take weeks, dilute the value of AI. BFSI firms must strike the right human+AI balance to unlock business value from AI.
How can BFSI firms transition to an AI-augmented operating model?
To implement an AI-augmented operating model, BFSI firms must address the friction points by building the model around three core pillars: role evolution, talent democratisation, and AI-enabled service delivery. Together, these pillars will define the new way of working for an AI-augmented mainframe enterprise across people, process, and technology (see Figure 1).
Figure 1: An operating model for AI-infused mainframe modernisation to grow business value
Let us examine the changes that an AI-augmented mainframe operating model will entail.
The mainframe operating model has historically been constrained by syntax-bound delivery, dependency tracing, and complex, undocumented logic. GenAI drives a transition to an intent-driven approach, changing the role of engineers from manually evaluating code to understanding the impact of a business request on application logic. For example, changes in rules applicable to interest calculation or policy administration or claims adjudication impact business logic across interdependent programmes, data objects, and interfacing processes. AI performs the analysis, identifies the impacted components, and recommends code changes. However, human oversight is a must—we recommend a human+AI model where engineers review and validate AI recommendations and retain accountability. This approach reduces dependence on mainframe SMEs for routine tasks across functions such as billing or trade processing or other core BFSI areas, in turn elevating their role to governing AI-generated outputs and decision-making.
Addressing the talent shortage by hiring or training a new generation in traditional programming languages is unlikely to scale. Building an AI-powered mainframe knowledge hub to store the business logic of legacy core banking or insurance systems greatly improves application explainability, making it easier for developers to discover legacy code and related components. This will enhance the ability of developers with modern technology skills, but limited knowledge of mainframe systems, to effectively support legacy modernisation efforts. In addition, equipping mainframe SMEs with AI skills will increase productivity and improve agility in achieving regulatory compliance, launching new products, rolling out feature upgrades to existing products, and grabbing cross-sell opportunities.
Leveraging AI in mainframe engineering enables developers to easily adopt agile and DevOps practices in the mainframe software development lifecycle (SDLC). By integrating AI tools into the mainframe environment, BFSI firms can facilitate better change management and faster development, accelerating upgrades to core banking and insurance applications in response to evolving business demands. Quality assurance processes also benefit through early identification of defects, test automation, synthetic test data generation, and lower remediation costs, resulting in faster and safer releases. Mainframe source code can be moved into enterprise repositories and be integrated with standard CI/CD pipelines. AI-enabled explainability allows modern and mainframe engineers to collaborate around business value streams rather than technology boundaries, mitigating the two-speed IT divide.
Let us examine the actions banks and insurers must take to move to an AI-augmented mainframe operating model.
- Rewrite the job architecture: Redefine roles and career paths around AI-augmented engineering, architectural approach, domain knowledge, and governance. Enable mainframe engineers to evolve into AI-assisted engineers and modernisation architects. Include skills such as prompt engineering, context engineering, AI output validation and human+AI collaboration.
- Audit human SDLC friction: Evaluate the manual intervention points in the SDLC. Identify opportunities for AI-backed automation and introduce AI into the mainframe software development process. This will help enhance testing agility, reducing the cycle time for application review and analysis.
- Run a pilot to demonstrate proof-of-value: Let us examine the steps that firms must initiate to realise quick wins:
- Select a low-risk, business-relevant value stream such as core banking compliance and reporting applications or insurance claims audit processes.
- Create a balanced team of legacy SMEs, modern engineers, cloud, DevOps practitioners, and AI tooling specialists to validate the future operating model in a controlled environment.
- Establish baseline key performance indicators (KPIs) such as first pass acceptance rate (quality), post-production defects per story point (quality), mean time to restore (speed), cycle time (speed), and total cost of operations, to validate performance.
- Lay down metrics for governance and talent as well as corresponding business KPIs, for example, reduction in claims processing and report generation cycle time.
- Enable continuous training of distributed and mainframe teams as well as business and IT stakeholders to enhance end-to-end productivity through effective human+AI collaboration.
- Establish an AI governance council: A governance team must manage and control AI usage, establishing robust guardrails for human+AI collaboration. Tasks that can be performed by AI and those requiring human approval must be clearly specified. Firms must lay down audit processes to ensure traceability along with stipulations for standardised AI templates and artifacts. In addition, KPIs must be established to track governance, compliance, adoption, and implementation efficiency.
- Monitor performance: Create a dashboard to track AI usage and productivity improvements for various IT teams. Measure progress by reviewing productivity parameters such as delivery velocity, SME effort reduction, output reliability, and the extent of AI use in the SDLC.
Successful pilots will lay the foundation to scale AI across core banking and insurance systems that still run on mainframes. The primary requirement here is to select the right business use cases and not limit AI adoption to IT processes alone—ultimately AI must unlock business value, for example, in areas such as new feature implementation, cost optimisation, and increased customer trust (see Table 1).
Table 1: Non-exhaustive metrics and benefits of AI business impact across BFSI functions
How does an AI-augmented mainframe translate to long-term business value?
For incumbent BFSI firms, completely exiting mainframes may not be an option, given that a large mainframe estate forms the foundation of business-as-usual. The future ecosystem will evolve into a hybrid platform where mainframe systems and AI-augmented models will coexist.
The future ecosystem will evolve into a hybrid platform where mainframe systems and AI-augmented models will coexist.
Three core pillars—the AI-powered knowledge hub, AI-infused mainframe engineering, and agentic AI-driven mainframe transformation—will underpin the new human+AI operating model, overcoming legacy constraints to bring mainframe agility on par with microservices or distributed architectures.
In our view, the AI-augmented mainframe operating model must evolve from delivering IT productivity gains to an ongoing organisation change management (OCM) paradigm, unleashing AI’s potential for exponential business value. This evolution will enable mainframes and modern digital platforms to coexist as a unified business ecosystem, empowering banks and insurers to deliver real-time intelligence, accelerate innovation, strengthen resilience, and create more personalised experiences. Banks and insurers must move quickly to make this structural shift, gain a competitive edge, and emerge as frontrunners in an AI-native landscape.