Original text here from Patrice Bernard (LinkedIn)
IBM has been a favored provider to financial institutions for over half a century, but today finds itself caught between the legacy of its past, especially its iconic mainframes, and the potential revenue sources of tomorrow, stemming from its acquisition of Red Hat. The big question is whether artificial intelligence can bridge these two worlds.
IBM's latest initiative is to offer a solution for clients burdened with legacy Cobol software, a language from the 80s, by creating an assistant to help them convert this software into Java, a language of the 2000s (though not the 2020s, unfortunately). Set to launch in the last quarter of 2023, this assistant will join a suite of AI-based solutions designed to streamline code reorganization and automate testing, covering all phases of modernization.
The official goal is to enable the majority of organizations, which are hesitant to embark on a costly and massive migration of their applications through complete rewriting, to instead take a selective and gradual approach. According to IBM, with the help of watsonx, these organizations could save about 30% of the time needed for conversion and deploy the generated components in the same environments as the original ones, thus minimizing disruption.
Subscribe to our newsletter:
Besides this smooth adaptability, including compatibility with CICS, IMS, DB/2 infrastructure tools, and the continued benefit of their hardware's exceptional performance (as claimed by the manufacturer), the generated programs will, we are assured, be optimized for the target language, which is quite different from the procedural model of its distant ancestor. However, I have doubts about this promise, as it seems almost unachievable: even though Java is inherently object-oriented, this can be superficial.
With such capabilities, watsonx stands a good chance of appealing to IT directors who wish to shed the burden of their outdated systems without incurring huge costs and risks of uncontrolled setbacks. This is because using AI in this evolutionary strategy significantly reduces the need for scarce Cobol skills, often a major obstacle in such endeavors.
I wouldn't be surprised if many have serious reservations about IBM's suggested approach. Transitioning to a new language might allow for maintenance and minor adjustments to currently static systems, but for systems that are typically critical, it does little to contribute to the deep transformation required for the digitalization of businesses, such as in terms of operational and management flexibility.
Moreover, even though it appears to spread out costs over time, the long-term economic equation of this option is not favorable. Translating obsolete software one by one only delays the inevitable need for a complete overhaul, especially when IBM eventually phases out its mainframes to focus on more profitable ventures. The Java components inherited from the watsonx assistant will then become orphaned and useless.