Enter a few lines into a prompt and generative AI will instantly deliver requirements, test cases, architectural designs or even finished code. In light of this rapid development, one question inevitably arises for requirements engineers, software architects and testers: Have visual modelling languages such as UML and SysML already become obsolete? Why should anyone invest time in learning, creating and maintaining complex diagram types when AI can handle everything at the push of a button?
The answer may come as a surprise: No, they have not become obsolete. In fact, quite the opposite is true. In the age of AI, model-based methods are becoming even more important. This article explores why this is the case, and what the perfect synergy between artificial intelligence and precise modelling looks like.
The Illusion of Carefree AI: Why Text-Based Requirements Alone Lead to a Dead End
Generative AI is a powerful text and code assistant. It can generate pages of user stories, functional requirements, and well-structured code components in an instant. However, this is precisely where the danger lies. It creates a flood of isolated texts and code snippets that lack structural and logical context.
Those who rely solely on AI-generated text-based requirements and dispense entirely with visual models will quickly reach critical limits.
- Missing Connections (The “Silo Trap”): AI can formulate a single user story perfectly. What it still struggles with today are cross-references. Without use case or requirements diagrams, it’s unclear how requirements influence one another, which actors are involved, and where functional dependencies exist. There is no visual map that makes the system comprehensible as a whole.
- Illusion of Dynamism and “Hidden Debt”: Requirements often describe static states. However, it is difficult to accurately capture how a system reacts when different events occur simultaneously in plain text. Without a state machine, critical system states and logical dead ends can be overlooked. “Technical debt” arises not because the generated code is difficult to read at the micro level, but because the system’s behavior at the macro level exhibits conceptual gaps. As a result, one quickly and easily builds beyond the architectural reality.
- Lack of “Single Source of Truth”: Text is flexible and open to interpretation. This applies to human prose as much as it does to prompts. Seamless and unambiguous traceability is mandatory, especially in regulated industries such as automotive, aerospace, and medical technology. A UML or SysML model provides this type of graphically verifiable source of truth on which both humans and AI can rely.
The New Ease of Modeling: Systems Thinking as a Core Competency
So, do experienced practitioners and students still need to spend a lot of time working with class diagrams, sequence diagrams, and state machines? Yes, but not to simply connect boxes with arrows. For requirements engineers, mastering UML and SysML has never merely been a technical drawing skill. Rather, it has been the central tool for structuring and analyzing.
This is where the requirements engineer’s true core competency has always been – and continues to be – systemic thinking:
- How are the domains logically related?
- Which system boundaries must be strictly adhered to?
- Are all interfaces and data flows consistent?
UML and SysML are visual grammars that facilitate this structured way of thinking. Those who master this grammar retain control over the architecture. This deep understanding is the decisive advantage in the age of AI. Only those who have logically penetrated a system can provide precise instructions to an external generative AI and confidently evaluate its results (keyword: context engineering). Requirements engineers, system architects, and developers are the strategists who design systems and establish guidelines.
Synergistic Interaction: AI and Models in Everyday Methodological Practice
The relationship between generative AI and traditional modeling is not an “either/or” situation. In practice, it’s evident that AI doesn’t replace modeling; it benefits from it. When requirements engineers skillfully combine general-purpose AI assistants with model-based methods, powerful synergies emerge in their daily work, regardless of the tools they use. This collaborative approach is evident in three key areas of daily practice.
- Support for Text-Based Work: Generative AI excels at drafting, refining, and supplementing unstructured requirements from customer discussions with initial test cases. Using these refined texts, requirements engineers can create appropriate UML or SysML diagrams in the modeling tool in a much more targeted manner.
- Models as AI Guidelines for Code Generation: Those who use development AIs (such as GitHub Copilot or similar tools) for code generation quickly realize that, without context, the AI “hallucinates.” However, if you provide the AI with the logical structures of a precise, self-created UML model in advance (e.g., as a text export), it will generate clean source code that adheres closely to the architecture. The model serves as a guide for the AI.
- AI-Assisted Reviews: AI systems are ideal sparring partners for quality assurance. You can task AI assistants with checking exported requirement catalogs or text-based model structures for logical contradictions, unclear wording, or gaps in the system design. This significantly reduces the workload of manual reviews.
Conclusion: The Method Remains and the Tool Evolves
Generative AI does not replace the methodological foundation of software and systems development; rather, it provides significant support in the areas of specification, design, and implementation. Those who have mastered the UML and SysML diagram types possess the rare ability to make complex worlds understandable. This applies to both people and machines.
A Clear Message to the Tech Talents of Tomorrow
At microTOOL, we have been training students in the dual study programm for many years. We understand that visual modeling may initially appear outdated to students. Why bother with functional dependencies when a few prompts seem to get you there faster?
The answer: UML and SysML have not gone extinct like dinosaurs – they’ve evolved. Today, they’re not a rigid requirement but rather a decisive upgrade for your career.
The reason is simple. In the age of AI, anyone who can only generate code is replaceable. However, those who can grasp systems in their entirety, identify logical dead ends in state diagrams, and guide AI as a methodological pacemaker will be key players in every development team. Modeling is the foundation, and AI is the accelerator.
Putting Requirements and Models into Practice
No matter if this methodological foundation was recently acquired in college or has been refined over time through hands-on experience, it requires the right environment to thrive. After all, even the best methodology is of little use if you have to struggle with tools that isolate requirements and models in separate silos in everyday work. For the synergistic interaction between people, methods, and AI to succeed in practice, requirements and system architecture must be linked seamlessly.
This is precisely where objectiF RPM is useful.
As a holistic platform for requirements engineering, objectiF RPM ensures that goals, requirements, and visual models are integrated. Whether you’re refining user stories as a team, mapping complex system architectures using UML/SysML, or ensuring seamless traceability for critical audits, objectiF RPM provides a stable, modern foundation on which human expertise, methodological precision, and forward-looking AI work together perfectly.