Translate

Thursday, January 15, 2026

Will AI be a Job Killer? Or Creator?

AI Systems and Steam Engines

AI is a really powerful technology. 

Especially GenAI for software. 

It reminds me of the first steam engines. Around the year 1800.

Gigantic machines. But minimal efficiency, compared to the standards of today (2026). 

Though back in 1800, they were like the eighth wonder of the world. Efficiency? Less than 1%. Not unlike the AI systems of today. But still impressive at the time.

And people had to find out how to use them efficiently.

Which they did.

And so they will for AI systems. We will find ways to put them to good use.

Some jobs will be replaced by AI. But a lot more will be created.

See my Article: "Requirements for Measurability and New Jobs in Software Development"

The article in IT Spektrum 1/26

Here's an overview, generated by Google Gemini:

The article itself is written in German. If you'd like to have an english translation, please let me know. I'll provide it.

Overview of the Article 

"The article "Requirements for Measurability and New Jobs in Software Development" by Martin Rösch argues that Requirements Engineering (RE) and Architecture will remain the most important components of software development in the long term, especially in the age of Artificial Intelligence (AI).

Overview of Key Statements

  • Requirements provide Measurability

  • Zero-Defect Software Development

  • Connection to Cognitive Psychology

  • Dealing with VUCA

  • Four Types of Knowledge Areas

  • Process Adaptation

  • Suitability for AI

  • Consequences of GenAI (Generative AI)

  • Conclusion - New Jobs

Key Statements

Requirements provide Measurability: The author describes how requirements (Anforderungen) in conjunction with acceptance criteria (AKs = test cases) already introduced objective measurability into software development in 1994 and saved projects by helping to resolve disputes about the "correct" object model.

Zero-Defect Software Development: Defect-free software could (beginning in 1994) be developed through the precise documentation of requirements and Acceptance Criteria (AKs), as well as the automated verification of object models. The questions about concrete AKs helped to identify the real experts.

Connection to Cognitive Psychology: Object models describe not only the business core of the software but also the stakeholders' understanding. According to ISO 30173, the "Real Twin" of the Digital Twin is the perception / imagination in the minds of the stakeholders.

Dealing with VUCA: Problems arise in RE when more than one knowledge area (KA, German: WG) is considered. The resulting frustration is described with the acronym VUCA (Volatility, Uncertainty, Complexity, Ambiguity). The solution lies in treating each knowledge area separately and connecting them precisely instead of creating a single large collection of requirements.

Four Types of Knowledge Areas: A distinction is made between (1) the business KAs of the users, (2) the technical KAs of the system, (3) the KA for process for creating the system, and (4) the KAs for each usage process of the system.

Process Adaptation: Large systems consist of n knowledge areas. The processes must be adapted to document KAs individually (requirements, AKs, object models) and to perform the integration not at the requirements level, but at the object model level (analogous to Amazon's approach to enterprise IT architecture).

Suitability for AI: Well-defined requirements and AKs (as automated tests) minimize AI hallucinations. Automated code generation by AI is expected.

Consequences of GenAI (Generative AI): Copy & Paste coding will disappear because AI is faster and better at that kind of work. The author draws a historical comparison with the introduction of the steam engine, where automatable parts of the work were automated, jobs were lost, prices fell, demand rose, and ultimately more new jobs were created.

Conclusion - New Jobs: In the long term, only RE (future knowledge documentation) and Architecture will remain from today's software development. More jobs will be created for knowledge documentation and for architects who design and connect all four types of knowledge areas. The reason for this is that RE brings measurability, precision, and quality, which leads to faster production, lower prices, and thus to a rising demand for individual software. "

If you ask for the full text (plus graphics) in English, I'll provide it.

Happy Reading !


The Time Has Come, For Creating Software Straight from Knowledge

After a break of 18 years, ...

... I now continue this blog.

Because AI has made it possible to generate software from descriptions of knowledge.

This is a game changer.

Now it gets interesting. 

Now AI can take care of the ugly parts of software development. 

The title I gave this blog in 2008 was "Software-Entwicklung ist Wissensverarbeitung". This is German and translates to "Developing Software is a Kind of Knowledge Processing". 

Maybe I was a bit early. 18 years early, to be precise. 

But the world has turned and now a few more people seem to agree.

Watch out for more to come.