The rapid adoption of AI-assisted code generation is transforming how software systems are built, but it also introduces significant risks to software design quality. While large language models (LLMs) can efficiently produce functional code at unprecedented speed, they lack an explicit understanding of the overall intended design structure, often resulting in poor modularity, hidden dependencies, and accumulating design debt. These structural deficiencies can negatively impact key quality attributes of a software system, including maintainability, security, and performance, which are of particular importance in the domain of embedded systems.
We are delighted to share that we are again contributing to the Embedded Software Engineering (ESE) Congress which is German’s leading conference for the embedded software industry. As part of the conference track “Design, Redesign, Refactoring” Dr. Stefan Malich will give a talk about proactively managing the software design quality in the AI era. This talk and the related paper were developed as part of the close collaboration between Dr. Stefan Malich, Prof. Yuanfang Cai (Drexel University, Philadelphia, U.S.), and Prof. Rick Kazman (University of Hawaii, U.S.).
- Title: Managing Software Design Quality in the AI Era – Controlling software design complexity in the context of AI-assisted code generation
- Date: December 2nd, 2026
- Location: Congress Center City Hall Sindelfingen, Germany
- Language: English
In this talk Mr. Malich will present a principled approach for assessing and controlling design structure quality using software design anti-pattern detection, modularity metrics, and Design Structure Matrix (DSM) visualization. Proven metrics such as propagation cost and M-Score provide quantitative measures of structural complexity and modularity. DSM-based visualization reveals the design roles of software artifacts, as well as implicit and cyclic dependencies that are difficult to detect through traditional code-level analysis. Anti-pattern detection exposes recurring structural flaws that degrade software quality over time.
Moreover, Mr. Malich will argue that the speed of AI can be most effectively leveraged only when guided by explicit architectural constraints. In the context of AI-generated software, rapid code synthesis can actually amplify structural problems without such guidance. By integrating these analyses into development workflows, teams can continuously monitor, diagnose, and improve design quality, while also steering AI systems to generate code that conforms to desired design structures, achieving both high productivity and high quality.
You can find the complete program for the Embedded Software Engineering (ESE) Conference here.
Please contact us if you are interested or have any questions related to our approach, tooling, and industry projects. (contact@architecturalchangemanagement.com).
