Development and use of high-tech systems generate large volumes of data-code, documentation, logs, operational data, and quality records - crucial for system and software maintenance and evolution. However, its complexity makes it challenging for engineers to interpret and use effectively. This track explores how generative AI, combined with traditional techniques, can help analyze diverse data sources, provide actionable insights, improve efficiency, and make maintenance and development more manageable.  

Modernising legacy code remains a persistent challenge in software engineering, especially when migrating between technologies or applying large-scale refactorings. In this talk, we explore the integration of program analysis techniques with generative AI to support such transformations. Rather than viewing AI as a silver bullet, we adopt a critical but open stance. We emphasise the importance of resource-aware generation and the crucial role of rigorous output validation. In particular, we show how static analysis can constrain, guide, and validate AI code generation, enabling safer automation of code transformation tasks and paving the way towards a correct-by-construction approach with emphasis on syntactical correctness. Through initial case studies conducted at ASML and early experimental results, we present evidence on how this synergy improves trust on automated modernisation solutions. We hope these are initial efforts that could move us closer to a future where legacy codebases can evolve more reliably and sustainably.

Philips IGT (Image Guided Therapy) produces medical equipment that saves peoples’ lives – a business in which quality comes first. An important aspect of this is dealing with customer feedback and complaints. Every report that comes in, is scrutinized for risk assessment (how likely is this to happen again?) and root cause analysis (how can we prevent this from happening again?). This is a labor-intensive task since it involves reading, understanding, and evaluating large amounts of text – not only in complaint reports but also related material like technical documentation and system event logs. Generative AI can help to reduce the human effort and to increase consistency and quality of the results, as we will show in this presentation. Since quality processes for medical equipment are subject to strict regulation, an important question is how to introduce Generative AI into these processes while still satisfying regulatory requirements. In particular, trustworthiness of the GenAI’s answers needs to be guaranteed, as will be discussed.