Software engineering of tomorrow

AI-generated code promises productivity, but could become the most expensive problem in software development - because no one is footing the bill for poor quality. While low-code and no-code are finally making model-based approaches acceptable, software testing is faced with the question: how do we ensure quality when machines generate code that cries out for appraisal? The answer lies not in more automation, but in a new generation of software testers with practical experience - and tools that finally bring the constructive and analytical sides together.

Podcast Episode: Software engineering of tomorrow

In this episode, I talk to Ina Schieferdecker about the future of software engineering. The conversation is about how artificial intelligence, low code and model-based approaches could affect our work. What will change in training? What skills do we need?

“I’m not a fan of vibecoding. I think we’re bringing in a big problem through the back door, because bad code that’s waved through too quickly becomes very, very, very expensive in the long run.” - Ina Schieferdecker

Prof. Dr.-Ing. Ina Schieferdecker is an independent researcher and honorary professor for software-based innovations at the Technical University of Berlin. She is a member of the German Academy of Science and Engineering (acatech), an honorary member of the German Testing Board and an active member of the International Software Testing Qualifications Board. She is also a member of the executive committee of the German Informatics Society and a board member of Informatics Europe. Her research interests include software quality engineering, open data platforms and the twin transformation of digitalization and sustainability. She is a winner of the German Award for Software Quality of the ASQF, the GI-TAV and the GTB, among others.

Highlights der Episode

• AI-generated code is cheap - bad code that runs for 10 years is expensive.
• Testing starts with testable requirements, or you’re developing for the garbage can.
• Model-based development failed due to maintenance - AI could enable bottom-up pattern recognition.
• Theory is not enough: testers need hands-on experience with real, critical challenges.
• Automation is on the rise, but real quality assurance needs experts - not automatons.

The future of software engineering - quality between AI, models and practical experience

We live in a time in which software shapes our lives on many levels. Applications control energy, vehicles, even medicine. At the same time, new tools, AI and approaches are emerging that are changing the way software is built and tested. What remains important in this growing complexity? And how do we prepare for the future? A conversation between Richie and Ina Schieferdecker provides insights and clear answers.

What is behind Next Gen Software Engineering?

For many years, the focus was on testing and quality control of software. Ina Schieferdecker has witnessed and shaped this development. But testing alone is no longer enough. She demands it: The close connection between building the software and testing must become stronger.

Traditional models such as the V-model already show that there are clear structures on both sides - development and testing. Software always needs a test system, and every specification also needs a test specification. This applies not only at the beginning, but throughout the entire process, right down to the requirements. It is precisely this interaction that will characterize this new era of software engineering.

AI, low code and model-based work - opportunities and risks

AI-based approaches are booming. Many are talking about large language models, agents, low-code or no-code solutions. There is a great temptation to use these tools to achieve goals more quickly. However, Ina Schieferdecker remains realistic - and warns that it will take time before we really become productive and the technology brings the desired benefits. She speaks of five to ten years until the new methods are really mature.

The biggest problem: poor, automatically generated code remains expensive and risky in the end. Quality deteriorates if we recognize errors too late or if they creep in due to poor AI results. It becomes particularly critical when software controls systems in which errors are not just expensive, but dangerous. That’s why automation should only take over what it can really do better and more reliably.

From theory to practice - what we need to (and can) learn

Currently, training in testing often consists of a lot of theory. Certifications and curricula provide an overview and precisely defined terms. But this is not enough when software is becoming increasingly complex and critical.

Ina Schieferdecker explains: “We need to put people more into practice. In the courses she has supervised, participants learn the most when they work on real examples. Not made-up requirements, but cases that come directly from their work. This creates real “aha” moments - and that’s what counts later in practice. Only those who really try things out and apply them will become experts in quality assurance.

AI can help to lower this threshold. It can make tasks and exercises accessible to many and offer new ways to practise and apply knowledge. This makes the practical learning effect even stronger.

Why testers will still be needed tomorrow

Many fear that automation and AI will make the testing profession obsolete. The opposite is true, says Ina Schieferdecker. The more software determines our lives and the more this software takes on tasks with great responsibility, the more important quality assurance remains. AI can support testing, make it faster and more targeted. But it can’t do everything.

The limited testing resources must be used as effectively as possible in the future - experience and skills are key. Smart testers bring these skills to the table, ensuring that software for critical systems remains secure in the future.

Next-gen software engineering means: more linking of development and testing, more courage to practice, more responsibility for quality. AI and new tools help, but do not replace the knowledge and experience of experts. It remains important to expand your own skills - and to actively shape the future of testing.