Mainframes still power the world’s most critical systems, but aging skills and rising costs are creating urgency. A composable, AI-led approach offers a faster path forward.

The growing urgency of mainframe modernization

Mainframes form the backbone of global enterprise computing, powering some of the world’s most critical systems. They process over 70% of global credit card transactions and support trillions of dollars in daily payments across banking, insurance, and public sector organizations. Despite the rapid growth of distributed cloud architectures, mainframes remain deeply embedded in enterprise IT landscapes – with 85 of the top 100 banks and 8 of the top 10 insurers still relying on them for mission-critical workloads.

However, while the technology remains highly reliable, the people and expertise sustaining it are becoming increasingly scarce. As experienced professionals retire, organizations are confronting a widening skills gap while managing regulatory requirements, cost pressures, and the demand for digital innovation. This convergence has elevated mainframe modernization from a technical initiative to a strategic priority.

Traditional modernization approaches have struggled to address these issues, and full re-architecture programs can take years to complete. Many enterprises find themselves caught between maintaining expensive legacy infrastructure and pursuing transformations that are too slow to deliver meaningful value – whereas driving more intelligent, automated approaches that combine Artificial Intelligence (AI), automation, and deep domain expertise will reduce risk while accelerating outcomes.

Composable AI: A smarter approach to transformation

That’s where composable modernization comes in. Rather than applying a single, rigid transformation methodology, composable architecture recognizes that each mainframe workload has unique characteristics. For example, a high-volume payments system requires strict latency, security, and regulatory compliance, while an insurance claims platform may prioritize workflow flexibility and auditability.

Composable modernization lets organizations assemble tailored combinations of tools, AI agents, and accelerators to match specific workload needs. Utilizing frameworks such as AWS Transform for mainframe, Capgemini can integrate specialized AI agents into a coordinated multi-agent ecosystem that adapts dynamically to business and technical complexity – reducing modernization timelines from years to months while improving alignment between IT execution and business outcomes.

Capgemini contributes deep industry expertise supported by more than 16,000 mainframe specialists and a structured ‘Assess, Transform, Cloud RunOps’ methodology. It also provides industry-specific accelerators across banking, insurance, manufacturing, and the public sector. Its SmartSDLC suite converts legacy system outputs into structured requirements, architecture designs, Application Programming Interface (API) definitions, and production-ready code.

Amazon Web Services (AWS) complements this by automating discovery, code analysis, business-rule extraction, and AI-driven refactoring across legacy technologies such as Common Business-Oriented Language (COBOL), Programming Language/One (PL/I), and Natural.

Together, these capabilities enable an end-to-end modernization approach that blends deterministic engineering discipline with AI-driven acceleration.

At the core of this approach is an integrated architecture that connects reverse engineering with forward engineering through intelligent orchestration. In the reverse-engineering phase, legacy applications are analyzed to extract business rules, dependencies, technical structures, and operational logic.

AWS Transform for mainframe generates structured Business Rules Engine (BRE) artifacts, which are orchestrated through Amazon Bedrock AgentCore to Capgemini’s SmartSDLC agents to produce cloud-native deliverables. This workflow delivers measurable productivity gains, including 10–20% improvements in early discovery and planning, and a 30–50% acceleration in transformation and testing. The result is a continuous, AI-assisted pipeline that converts legacy mainframe code into cloud-native applications with significantly reduced manual effort and improved consistency.

Real-world impact: From legacy constraints to business value

As a real-world example, consider a mid-size US financial institution working to modernize its core banking and payments platforms. Despite processing tens of millions of transactions daily they faced significant challenges: more than 60% of mainframe developers nearing retirement, increasing regulatory demands for API-based integration, rising infrastructure costs of up to 12% annually, and slower release cycles compared to competitors.

Instead of a risky full-scale migration, the institution adopted a phased, composable modernization strategy:

  • In phase 1, a high-volume transaction platform with nearly one million lines of COBOL was transformed using AWS Transform for mainframe and Capgemini’s SmartSDLC capabilities, resulting in over 100 APIs, 20+ integrations, 100+ modernized User Interfaces (UIs), and multiple Extract, Transform, Load (ETL) pipelines deployed on Amazon EKS, enabling full retirement of the legacy system.
  • Phase 2 extended modernization to core banking systems involving COBOL, Customer Information Control System (CICS), and DB2 workloads, migrating data to Amazon RDS for PostgreSQL and consolidating legacy interfaces into a unified digital platform powered by services such as Amazon ECS Fargate, AWS Lambda, API Gateway, SQS, and EventBridge.

The outcomes highlight the tangible value of composable modernization. The organization achieved a shift from quarterly to bi-weekly releases, reduced transaction costs by 35%, lowered mainframe costs by 25%, improved onboarding time from months to weeks, and increased overall delivery speed by 60%.

For firms embarking on a similar journey, the path forward typically involves three steps:

  1. Assess the application landscape using AWS Transform for mainframe and Capgemini tools, such as CAP360, to identify modernization opportunities.
  2. Execute a focused pilot on a bounded workload to validate the end-to-end transformation approach.
  3. Scale incrementally across domains, leveraging reusable assets and established patterns to accelerate each subsequent phase.

This structured model enables modernization without disruption, while continuously building momentum. Ultimately, mainframe modernization is evolving into a scalable, AI-driven transformation journey that turns legacy complexity into a foundation for long-term innovation and growth.