GenAI and the modernisation of mainframes and legacy systems

By Gerhardt Scriven, Anil Kumar Mallanna e Sanjay Rao

In an era when mainframes continue to be the backbone of many global companies, the challenges of maintaining, optimising and modernising outdated legacy systems have become a priority. These systems, often burdened with technical weaknesses from years of patching, are not only expensive to maintain, but also hamper business agility.

The scenario is further complicated by the imminent retirement of a significant portion of the skilled labour force that has the knowledge to maintain these systems, as well as the often inadequate technical documentation, which makes knowledge transfer expensive and inefficient.

Demand for mainframe modernisation is therefore growing, driven by the need for simplified applications that increase agility and mitigate continuity risks. However, these projects are notoriously complex, with a failure rate of 74% among the organisations that undertake them. Fortunately, recent advances in technology, partnerships with Cloud Service Providers, the use of hybrid cloud environments, as well as incremental modernisation strategies, offer more viable and less risky paths to modernisation.

Despite this, modernising COBOL continues to pose significant challenges. Decades of business logic coded in this language make it difficult to extract, document and translate into more modern languages. Automated code conversion often produces Java correctly, but there is still a difficulty in maintaining and scaling, hampering the use of modern programming practices. In addition, traditional one-size-fits-all tools do not fully take into account the nuances of different legacy systems, requiring a great deal of manual effort in debugging, testing and restructuring.

It’s worth noting: there is growing recognition of AI’s potential to address these challenges, offering promising solutions for automated code conversion, back-documentation and testing processes in mainframe modernisation.

Can Artificial Intelligence and Generative AI meet the challenges?

Generative AI (GenAI) represents a leap beyond traditional mainframe technology and AI applications. It’s not just transforming business problem solving with new human-like content, but actually accelerating mainframe modernisation. GenAI’s deep understanding of legacy code semantics, as well as heuristic interpretation of corporate engineering standards, allows for more effective capture of business logic and intent, enabling precise code transformation and knowledge encapsulation.

Several consulting companies, IT services and cloud service providers are exploring AI applications for refactoring code, generating visuals to demystify complex systems and improving data migration. These innovations aim to bridge the gap between legacy systems and modern technologies, although many are still under development.

At the heart of this vision is the synergy of knowledge management (one of the most important critical business and continuity risks for companies in the most diverse sectors), the optimisation of existing COBOL code and the transformation of legacy applications into modern, sustainable systems.

This is not to say that we are advocating moving away from the mainframe using GenAI: instead, by leveraging GenAI for a more systematic and controlled modernisation process, we simplify systems in terms of flexibility, performance and maintainability.

Accelerators are able to create virtual replicas of legacy applications by analysing millions of lines of legacy code. Feeding GenAI extractions from these virtual replicas makes it possible to accelerate modernisation results in knowledge management and code optimisation, application modernisation and testing, and application maintenance and support.

It’s not about promoting one LLM (Large Language Model) over another. Generative AI models are becoming more sophisticated at an unprecedented rate, and many of the proprietary as well as open source models perform exceptionally well for mainframe modernisation purposes. Here are a few examples.

1. Knowledge Management and Code Optimisation

• For static content, it is recommended to develop domain-specific knowledge portals that are adapted to different types of users and usage scenarios. This eliminates a tonne of cumbersome technical reports that are usually created by current market tools, and which are often not used at all.
• In addition, “Ask AI Anything” chatbot features allow users to extract insightful and meaningful answers from GenAI for complex technical queries, including business logic and business rule extraction, copilot for COBOL or intelligent system refactoring recommendations to simplify and optimise existing legacy code.
• Code analysis features can also be used to perform deep semantic analyses to discover security vulnerabilities, patterns or performance bottlenecks in the source code. Unlike traditional tools, GenAI can understand the context of the code, making it better at detecting complex vulnerabilities and suggesting more nuanced solutions.

2. Application modernisation and testing

• Knowledge management resources are ideal for reverse engineering legacy source code for user stories and acceptance criteria, and then creating functional test cases based on these criteria. This, together with detailed knowledge of how data exists in the legacy system, allows you to create synthetic data.
• Classical Machine Learning (ML) techniques should be used to design sustainable and isolated services to improve the performance of the target system and reduce costs, as well as ensuring that the modernised system is robust and sustainable.
• In addition, taking advantage of GenAI guarantees that the final code is fully compliant with the quality and security standards of corporate engineering, automating the creation of the target system’s documentation as well as the Unit Tests.

3. Application Maintenance and Support

• Don’t forget Incident Management: GenAI can also be used for Root Cause Analysis, identifying underlying problems in the source code and then suggesting solutions based on the resolution of past incidents or generating new solutions through contextual understanding of the AI application’s code base. As an extension, you can automatically update the knowledge base by documenting such incidents and the corresponding Root Cause Analysis and Resolution.
• Code scans can also turn into predictive mode based on the trends and anomalies found in the code being scanned and comparing it with the Root Cause Analysis of past incidents. These incident management opportunities can be automatically documented in the chosen tool, recording, tracking and auditing automatically without any manual intervention, while workflow approvals are directed to the responsible authority.
• You can also take advantage of GenAI to improve developer productivity in application development and maintenance tasks by identifying opportunities to improve existing problems in the source code or even best coding practices, making maintenance easier. For example, a large scope of analysing specific patterns in the application’s code base and applying fixes for similar patterns is a good example of POC for a customer use case.

On a broader level, the implementation of GenAI makes it possible to manage complex dependencies in legacy systems, ensuring that upgrades or migrations don’t interrupt functionality and, at the same time, reducing integration problems in the process. In addition, dynamic knowledge updates ensure that system documentation remains up-to-date with production code.

As we embrace a new era of mainframe optimisation and modernisation enhanced by GenAI, it’s time to move beyond obsolete ways of working to thrive in a modern digital landscape. You need to transform legacy systems and propel your company to the forefront of agility and innovation.

About the authors

Anil Kumar Mallanna – Managing Partner, Legacy Application Modernization and Platform Services (LAMPS) Wipro FullStride Cloud Services

Anil brings 25+ years of IT experience with extensive knowledge of enterprise-wide application businesses and expertise in designing and implementing mission-critical applications. He is responsible for Wipro’s Americas legacy modernisation charter and has led successful sales, pre-sales, consulting and IT delivery organisations serving global leaders in large financial services industries.

Sanjay Rao – Diretor de Serviços de Plataforma e Modernização de Aplicativos Legados (LAMPS) Wipro FullStride Cloud Services

Sanjay is a legacy modernisation consultant and cloud architect with 25 years of experience in simplifying, augmenting, migrating and modernising mainframe applications. He leads pre-sales, consulting and delivery for Wipro’s Americas 1, which includes the healthcare, life sciences, communications, retail and Brazil sectors.

Gerhardt Scriven – Executive Director, CAPCO

Gerhardt Scriven has more than 20 years’ experience in IT, with a focus on eliminating risks and solving complex problems in the delivery of mission-critical projects, particularly through early risk discovery and mitigation. He specialises in Legacy Application Optimisation and Modernisation.