Where administration is often seen as sluggish, a local initiative shows how quickly testing can become practice. With MUCGPT and the Sherlock assistant, standard-compliant test cases in accordance with ISO 29119 are created from precise prompts and can be transferred directly to TestLink and Xray. Humans curate and decide. At the same time, Watson is created as a generator for test reports.
In this episode, I talk to Mark Menzel from the City of Munich about MUCGPT and Sherlock, the software testing AI assistant. Munich shows: administration can innovate. Sherlock helps departments to create standard-compliant test cases in accordance with ISO 29119 without a training marathon. MUCGPT generates test cases that can be exported to TestLink and X-Ray. The human remains the decision-maker.
"AI is here to stay. We won't be able to avoid it. And it is also being used and applied in the city of Munich." - Dr. Mark Menzel
Dr. Mark Menzel has more than 25 years of experience in software quality and test management. After completing his doctorate in Business Administration, he specialized in AI-based test techniques. As a test manager at the City of Munich, he manages test projects for the City of Munich. He also works as a university lecturer and mentor and researches how AI can revolutionize testing.
Munich is not only the Bavarian capital, but is now also a pioneer when it comes to the use of artificial intelligence (AI) in public administration. This is demonstrated by the conversation between Richie and Mark Menzel in the "Software Testing" podcast. What might surprise many: The city has developed its own GPT, called MUCGPT, and is already using it in many areas. The goal is clear - practical support for employees without them having to have in-depth technical knowledge.
The highlight: in addition to standard functions such as summarizing texts or formulation aids, there is now "Sherlock", a dedicated AI tool for creating test cases - tailored to the challenges of the city.
Anyone who has ever worked in a public authority knows that a lot of it revolves around suitable forms, systems and processes. When laws change or new requirements are introduced, these not only have to be programmed, but also tested. Test cases are the backbone of this. The problem is that the experts live and breathe their topics, but often lack specialist testing knowledge.
This is where "Sherlock" comes in. Mark Menzel describes how many in the specialist departments have the necessary knowledge of their own system, but writing test cases is difficult. You are often faced with a blank sheet of paper and don't know where to start. This is exactly where AI as an assistant has found its place.
"Sherlock" is not just a nice name, it is a real assistant that helps employees to create test cases in accordance with ISO standard 29019. The highlight: the interaction is simple and dialog-based. You ask Sherlock: "Who are you?" - He answers that he helps to create test cases and provides an example directly. This creates a dialog in which test cases are built up piece by piece.
The special advantage: test cases can be exported directly into the required format for the various test management tools (Testlink or Xray). This not only saves time, but also takes away users' fear of having to deal with new software.
How has the service been received? According to Mark Menzel, there is positive feedback, which is mainly due to the fact that it makes work much easier. In the training courses, they place particular emphasis on "prompt engineering" - in other words, how to ask the AI the right questions and commands. This is the only way to obtain test cases that are technically and methodically sound.
Of course, there are teething problems: The AI has to rely on the information provided, as it does not know "the city of Munich" on its own. And there are often requests from the specialist departments such as: "I need ten test cases, not just one!" or "Can defect reports also be generated automatically?" The team is already planning the next step.
Whoever says Sherlock must also say Watson. That's why a student is now working on creating "Watson", an assistant that automatically creates test reports. This includes the most important data: How many test cases were run? Which tests failed? Which bugs were found? Such automation takes the pressure off everyone involved and frees up time for the really important tasks.
Despite all the innovations, Richie notes in the interview that administration is by no means always outdated. People like Mark Menzel and the City of Munich's willingness to invest in AI expertise show that change is possible. It is particularly important to be open to fresh ideas, to work closely with young people and to be willing to grow with mistakes and feedback.
AI-based assistants such as Sherlock and Watson ensure that software tests in Munich become faster, simpler and more efficient. Their use makes specialists more independent, relieves the burden on IT teams and takes administration a big step forward - all without losing sight of people.