Agentic artificial intelligence is rapidly reshaping how organisations modernise mainframe systems, offering a path to translate legacy Cobol code into modern languages, reduce operational risks, and cut costs. However, industry experts insist that human judgment and expertise are still critical to ensure these transformations succeed.
Drivers of adoption
Three main factors are pushing enterprises toward AI-assisted code modernisation: skills shortages, cost efficiency, and the need for greater agility. The chronic lack of Cobol and mainframe specialists has long been a headache for organisations still running legacy applications. As experienced professionals retire, the risk of losing institutional knowledge grows. Agentic AI tools, such as Anthropic's Claude Code, promise to automate much of the language translation process, potentially easing the skills gap.
Michael Vincetic, a practice leader at a major IT services firm, notes that the ability to understand decades-old business logic and interdependencies has 'flipped the script' on modernisation costs. Previously, reverse-engineering a legacy system could be more expensive than rebuilding from scratch. AI now makes it feasible to decipher that complex logic at a fraction of the time and cost.
Beyond simple translation
Simply porting code from one language to another offers limited value, warns Manjunath Bhat, a distinguished vice-president analyst at Gartner. 'There is very little value merely in porting code from one language to another without modernising the architecture and infrastructure,' he says. True modernisation requires adopting composable, modular architectures that allow independent testability and deployment, reducing the 'blast radius' of changes.
Anthropic claims Claude Code goes beyond syntax conversion. The AI understands dependencies, preserves business logic while updating to current frameworks, and even generates test units and modern documentation. Bhat adds that AI's real power lies in semantic conversion—mapping underlying data flows and explaining what code does, which parts are risky, and what interdependencies exist. This 'discovery' phase prepares organisations for smoother downstream work.
Agility and hybrid strategies
For companies seeking agility, moving parts of their systems to public clouds allows them to tap into advanced data management and analytics. But running both legacy and modern infrastructure can create a 'two-speed IT' environment, where the old operates in rigid waterfall cycles and the new in agile sprints. This often increases costs and slows time-to-market.
Banking and government sectors are leading the charge. Banks face ongoing digitisation pressures, while governments aim to improve citizen services. A recent survey by the IT services firm found that 80% of organisations have shifted their mainframe modernisation strategies in the past year. Among them, 43% are modernising more on the mainframe itself, 34% are integrating more with cloud, and only 16% are moving more applications off the mainframe. Remarkably, only one in 500 respondents plans to abandon the mainframe entirely.
However, the survey also indicates that 98% of respondents are moving some applications off the mainframe, averaging 28% workload migration. At the same time, 56% are increasing overall mainframe use, often positioning it as the centrepiece of a hybrid environment.
Three pillars of modernisation
Vincetic outlines three elements for successful modernisation: updating infrastructure (e.g., from mainframe to cloud), modernising operations to suit the new environment, and revamping the overarching operating model. The real challenge is understanding the context and interdependencies within a complex mainframe ecosystem. Agentic AI can untangle these systems to reveal business rules and data flows—often in half the time and with high quality.
Nevertheless, 'the expert in the loop is still very critical,' Vincetic emphasises. Strict controls around regulatory compliance, availability, and disaster recovery require human oversight. Mainframes have endured because of their security, high-volume transaction processing, and reliability. 'There are certain workloads that absolutely are best placed to reside on a mainframe,' he explains. The goal is to balance those intrinsic strengths with the market's demand for agility and digital channels.
Modernisation is not an all-or-nothing proposition. Tackling one workload at a time is practical, but understanding the context and overarching drivers is essential. AI can accelerate the process, but expert guidance remains indispensable for achieving equal or better capabilities than before.
Source: ComputerWeekly.com News