Abstract

Generative AI (GenAI) has moved into professional work at remarkable speed, yet the gap between how it is used, imagined, governed by organisations, and how it unfolds in practice lacks research. This study examines how GenAI is integrated into the daily workflows of professionals in information technology (IT) and management, and what risks arise as a result. The research takes a qualitative approach and was conducted in collaboration with the Nordic IT infrastructure company ATEA. The material was analysed combining a categorised risk taxonomy derived from the literature on recent GenAI research with Steven Shorrock’s framework The Varieties of Human Work.

The findings show that GenAI is broadly integrated across both domains, with substantial overlap in activities such as drafting, information retrieval, transcription, and as a sounding board. The study shows where the validation occurs, between professionals in IT and management. While IT respondents tend to validate output against their own expertise, management respondents more often validate partially, through a second tool, or not at all. The reported risks map largely onto the four categories, with an additional concern, the erosion of individual and organisational competence, recurring across the material. Applying Shorrock’s framework to activities rather than
persons relocates these risks to the gaps between imagined, prescribed, done, and disclosed work, and surfaces four limitations of the framework when applied to cognitive GenAI work.

Keywords: Generative AI, AI literacy, sociotechnical risk, professional workflows, Work-as-
Imagined, Work-as-Done, Work-as-Prescribed, Work-as-Disclosed, risk management,
information technology, management