2023 was the year in which generative AI broke into the public consciousness, with 59 per cent of consumer internet users using Open AI's ChatGPT. As developers, we've likely been experimenting with AI code generators outside of work, even if we're not using them in earnest in the corporate environment. However, in 2024, generative AI will start to create a significant impression in software development across all programming languages, as enterprises evaluate and add specific AI tools to their approved lists. One that is likely to be a game changer in the IBM Z universe is watsonx Code Assistant for Z which is underpinned by IBM's watsonx.ai code model.
Leaning on AI for COBOL to Java translation
If you've not come across it already, the premise behind watsonx Code Assistant for Z, is all about using generative AI to streamline the modernization of COBOL applications. Specifically, it aims to make it easier for developers to lean on AI to selectively and incrementally transform COBOL into well-architected high-quality Java code. In doing so, it's going to fast-track application modernization initiatives for IBM mainframe customers, cutting costs and risks, and reducing the reliance on the shrinking pool of COBOL developers.
If you're part of the mainframe community, you're well aware that COBOL represents a double-edged sword. On the one hand, as a 60 plus year old language it has really stood the test of time. There are 220 billion lines of COBOL code still in use today underpinning 43 per cent of all banking systems, $3 trillion of daily commerce transactions and handling 95 per cent of all ATM card swipes.
On the other hand, many of these applications were written decades ago and are in dire need of modernization. Enterprises want to build new service offerings, add new functionality and integrate with new applications written in modern languages and cloud applications.
Enterprises choosing to upgrade existing mainframe assets
In fact, around 70 per cent of executives admit that the mainframe applications in their organizations need to be modernized - and they are twelve times more likely to upgrade existing mainframe assets rather than rebuild their applications from scratch in the next two years, which they fear could be too costly, time-consuming or risky.
So, most enterprises would rather retain and build on existing applications than create new applications or migrate everything to the public cloud and risk losing the primary advantages of IBM Z: high levels of security, resilience and performance.
However, with the chronic shortage of skilled COBOL developers, this is easier said than done. Hence the introduction of watsonx Code Assistant for Z. Expected to be available from the end of 2023, IBM says it is being developed to help businesses use generative AI and automated tooling to accelerate mainframe application modernization - while preserving performance, security and resilience of the mainframe platform.
Taking a targeted approach to mainframe modernization
Importantly, the watsonx Code Assistant will allow companies to take a targeted approach. Rather than modernizing entire applications, customers will be able to select specific parts of a COBOL application and use watsonx Code Assistant to transform it into Java. The resulting code will be optimized to interoperate with the rest of the COBOL application, IBM says.
The Code Assistant will be part of an end-to-end solution with a set of complementary tools. There will be tools for application discovery and analysis to help understand the target application dependencies, for selecting and extracting COBOL code that delivers particular business functions; and for performing automated code refactoring and COBOL to Java translations. Also included will be functionality for auto-generation of test data so that developers can validate the translated code.
Trust in AI models is going to be key
A key factor in the adoption of any AI-assisted code generation solution will be trust. Are enterprises and their development teams going to trust the code that's automatically generated? Here, of course, IBM has major advantages because of its strong foundation in governance and innovation - and its commitment to transparency. People are also more likely to put their trust in an IBM AI offering for COBOL because it has access to the largest private data set of COBOL data with which to train the most state-of-the-art large language models.
On top of this, the watsonx.ai code model, which underpins the Code Assistant for Z is a highly sophisticated system that has learned from 1.5 trillion tokens across 115 coding languages, demonstrating extensive generative AI capabilities.
Nevertheless, customers will likely begin with a cautious and iterative approach. Rather than upgrading the most critical parts of their applications, they will start with parts that involve the lowest risk. This will allow them to build confidence and trust. And IBM will be on hand to provide consultancy to help customers identify the right application areas to modernize to get the most from watsonx Code Assistant for Z.
In summary, the automation and AI-powered recommendations delivered by watsonx Code Assistant are designed to boost developer productivity, lower the total cost and complexity of mainframe application modernization initiatives and reduce their time-to-value. Beyond the technical and cost advantages, the product also strengthens IBM's relationship with its mainframe customers, providing a mutually beneficial partnership for modernization efforts.