Abstract
<p>Generative AI is rapidly transforming white-collar work; however, little is known about its uptake in research administration, a specialized field of non-routine, judgement-dependent labour. This study reports the findings from a cross-sectional survey of researchers and research administrative staff at a large academic research hospital in Toronto, Canada (n = 119). Almost three-quarters of respondents (73%) reported using generative AI tools for research administrative work. Four outcome variables, Frequency, Effectiveness, Impact, and Quality of generative AI use, were strongly intercorrelated, indicating that respondents who used the tools more frequently also rated them more favourably across all three evaluative dimensions. Self-assessed confidence positively predicted evaluations of the tools, whereas job security concern was negatively associated with evaluations of generative AI. Respondents who iteratively refined outputs rated the tools more favourably than those who abandoned their attempts at using them, while heavier workloads were associated with less favourable quality perceptions. These findings suggest that realizing generative AI’s potential in administrative work depends on the user’s skills, confidence, and working conditions, rather than the tools themselves.</p>