AI as relief or a new source of stress? Why humans remain in the loop despite GenAI

How is generative AI changing the work of software developers, and does it actually reduce the strain on them? Researchers at the Institute Digital Technology Management have investigated this question as part of our SNSF project. The study is based on 26 interviews with DevOps professionals from two Swiss companies in the finance and insurance sector. The findings show that GenAI can relieve employees, but at the same time it creates new demands and new forms of technostress.

Generative artificial intelligence is rapidly changing the world of work. It writes texts, generates program code or summarises documents, and promises to make work more efficient (Merkel & Dörpinghaus, 2025; Nguyen-Duc et al., 2025). But does more automation actually mean less strain? Our research shows that AI can make everyday work easier while at the same time creating new forms of stress caused by digital technologies – also known as technostress. This becomes apparent, for example, when employees have to continuously learn new tools, operate several systems in parallel or deal with a flood of information (Jeyam et al., 2026).

This is particularly evident in software development. Here, continuous technological change is part of everyday work – and so is working with generative AI. To better understand how employees experience this development, we conducted 26 interviews with DevOps professionals from two Swiss companies in the finance and insurance sector. The respondents see major advantages in GenAI: AI supports them in writing code, summarises documentation or helps to analyse technical problems. This means routine tasks can be completed more quickly. At the same time, new challenges arise: AI does not always deliver correct results and can produce so-called hallucinations. Hallucinations are contents generated by an AI model that appear realistic but deviate from the given source inputs. Its suggestions therefore have to be checked, put into context and, where necessary, corrected. Responsibility does not disappear – it shifts.

«Still in the loop»

Our interviews therefore gave rise to the concept of “Still in the Loop”. Unlike the familiar term “human in the loop”, it describes the fact that professionals do not merely check individual AI results, but remain responsible throughout the entire work process – from formulating the inputs to evaluating and implementing the results. The more powerful AI becomes, the more important human capabilities such as critical thinking, domain expertise and judgement become. The work does not disappear; it changes.

Technostress cannot be managed through individual resilience alone. Our findings show that organisations play a central role. Knowledge sharing, continuous further training, standardised tools and an open culture around mistakes help to reduce strain and support the productive use of AI. The introduction of generative AI should therefore not be understood as a purely technical project. It is equally important to prepare employees for the changed requirements and to design work processes in such a way that AI does in fact reduce the strain.

Generative AI opens up major opportunities for knowledge-intensive work. At the same time, new forms of technostress arise because employees have to understand, verify and take responsibility for the results produced by AI. People therefore remain “still in the loop”. The sustainable success of GenAI depends less on the performance of the technology than on how well companies design the collaboration between people and AI.

What comes next?

But do all employees experience working with AI in the same way? In our further research, we are examining the role that age and gender play in the experience of technostress. We are particularly interested in whether different groups perceive and use GenAI differently, and how they deal with the associated strain.

Because the next question after “Still in the Loop” is: who experiences collaboration with AI, and how?

This article is based on the following paper:

Dharneeka Jeyam, Anna Wiedemann, Gerhard Schwabe, and Kadircan Güney. 2026. “Still in the Loop”: Coping with Technostress in DevOps Teams and the Impact of GenAI. In Proceedings of the IEEE/ACM 48th International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP ’26). Association for Computing Machinery, New York, NY, USA, 590–600. https://doi.org/10.1145/3786583.3786899

References

Merkel, M., Dörpinghaus, J. The transformative potential of AI in software engineering: a case study on LeetCode and ChatGPT. Empir Software Eng 31, 180 (2026). https://doi.org/10.1007/s10664-026-10912-5

Anh Nguyen-Duc, Beatriz Cabrero-Daniel, Adam Przybylek, Chetan Arora, Dron Khanna,TomasHerda,UsmanRafiq,JorgeMelegati,EduardoGuerra,Kai-Kristian Kemell, Mika Saari, Zheying Zhang, Huy Le, Tho Quan, and Pekka Abrahamsson. 2025. Generative Artificial Intelligence for Software Engineering—A Research Agenda. Software: Practice and Experience n/a, n/a (2025). doi:10.1002/spe.70005 _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/spe.70005.