Artificial intelligence in software modernisation

Our six-part series of articles covers all important aspects of software modernisation. This is the sixth and final article in the series. An overview of the topics is provided at the end of this article.

Legacy systems that have been in operation for decades are often expensive to maintain, inflexible and make it more difficult to integrate modern technologies. As demands on agility increase, software modernisation is becoming mandatory.

The use of generative artificial intelligence can be a useful tool in this respect. It opens up new possibilities for optimising and transforming systems, potentially making the entire modernisation process faster and more cost-effective. AI can be used for various purposes.

Code analysis and refactoring

Even today, generative AI can analyse a considerable amount of code at once. This allows for a comprehensive code analysis that goes beyond the capabilities of static analysis tools alone. As a result, inefficient structures can be identified, suggestions for changes can be generated and concepts used can be collected in order to speed up code analysis and carry out refactoring based on previously proposed improvements. Readability and maintainability can be optimised in this way.

This allows the code to be transformed into a more understandable form using AI tools.

Migration of legacy systems

AI can also provide support for migrating legacy software systems to new technologies, platforms or architectures. For example, code can be translated into another programming language. Although this is possible, the risk of errors in translation or incomplete translation is high and should be backed up by extensive testing or adaptation of the existing test suite to the new system.

Moreover, only parts of systems can be migrated to newer technologies using generative AI. For example, interfaces to databases or bus systems can be migrated more efficiently, allowing more modern solutions to be implemented. This can prove useful, for example, in the event of cloud migration. Generative AI can also provide support for upgrading to a new interface technology. However, if usability is to be improved, the expertise of UX specialists is essential.

Automated testing

Legacy systems generally do not offer automated tests, which are one of the essential prerequisites for maintainable systems. Generative AI supports the creation of automated test cases for these systems. This allows engineers to build an understanding of the system while at the same time creating a suite of automated tests to support them with future modernisation.

Documentation and understandability

Lack of documentation is often a challenge when modernising legacy systems. Generative AI can analyse existing code and independently create appropriate documentation to make the system easier to understand.

Generated specifications with limit values can increase understanding of existing source code so that developers can continue to develop more effectively and efficiently. The system can therefore be analysed using chatbots and better understood by asking questions. For example, prevailing concepts or design patterns can be identified more quickly.

AI-enhanced processes

In addition to supporting modernisation, AI can also be integrated into everyday work. For example, AI can act as the first pull request review instance, providing early feedback.

Conclusion: Revolutionary – but not without people

Generative AI offers a revolutionary way to modernise legacy systems efficiently and more cost-effectively. Conventional modernisation approaches often require a lot of effort and large development teams to perform analysis and refactoring tasks in manual processes. Generative AI can perform some of these tasks automatically and autonomously, reducing project costs and accelerating the modernisation process.

AI will play an even greater role in the future, potentially enabling self-optimising systems, AI-supported DevOps processes and the integration of new features and technologies with equal efficiency.

Experiments with generative AI in software modernisation should be conducted and examined for applicability in the specific situation. Human involvement in the system is essential to complete the process. An experienced software engineer is needed to guide the AI. Creating the right prompts, asking the chatbot the right questions, evaluating AI output and refining it by asking the right questions – this requires experienced software engineers. A machine cannot assume responsibility.

Further articles on software modernisation in the bbv blog:

• Blog #1: Software modernisation: Everything you need to know?
• Blog #2: Software Modernisation Canvas: Structured for success
• Blog #3: Strategies for successful software modernisation
• Blog #4: The five most important factors for successful software modernisation
• Blog #5: The future of legacy systems: from modernisation to evolvability
• Blog #6: Artificial intelligence in software modernisation