IRAIS 2026 Proceedings

Abstract

Generative AI has become deeply embedded in everyday knowledge work, with organizations increasingly deploying AI tools to improve efficiency by streamlining workflows, reducing task friction, and accelerating output. Beyond these gains, AI has been shown to enhance productivity (Dell’Acqua et al., 2026) and support core functions such as summarization, translation, and document drafting. Nevertheless, cyberloafing – an employee’s non-work-related internet use during work time (Jiang et al., 2024), such as browsing news, personal emailing, social media, shopping, or gaming – remains a significant workplace concern. The extensive integration of AI into workplace workflows gives rise to a distinct phenomenon – AI-enabled cyberloafing. While the broader phenomenon of AI affecting employee non-work behavior has begun to receive scholarly attention, there are three gaps remaining for further study. First, (1) no existing study empirically isolates perceived slack as the mechanism through which AI-freed time surplus translates into distinct forms of cyberloafing behavior. Second, (2) prior cyberloafing typologies do not differentiate the distinct motivational pathways such as avoidance, curiosity, and socialization that AI tools may uniquely enable. Third, (3) no study has experimentally tested organizational monitoring as a moderator of how employees allocate post-task discretionary room toward cyberloafing behavior.

In this study we consider three primary motivations through which employees engage in AI-enabled cyberloafing. First, work avoidance, whereby employees use AI tools as a means of escaping demanding or monotonous tasks. Second, curiosity, where the novelty and conversational nature of AI systems naturally draw employees to explore their capabilities beyond work-related purposes. Third, socialization and entertainment, as conversational AI offers an accessible and low-effort substitute for social interaction and leisure during working hours. To better understand this emerging phenomenon, this study pursues the following research questions: How does perceived efficiency of AI tools relate to employees’ perceived slack at work? How does perceived slack relate to employees’ engagement in AI-enabled cyberloafing? How does organizational monitoring moderate the relationship between perceived slack and AI-enabled cyberloafing? In S-O-R terms, these three motivations are distinct forms of the cyberloafing response, not mediators or manipulated factors, so monitoring and slack (stimuli) may shape each form differently through perceived slack (organism). This study is expected to make three contributions to the emerging literature on generative AI in the workplace. First, it extends cyberloafing research by introducing and empirically examining AI-enabled cyberloafing as a distinct workplace phenomenon. Whereas prior research has largely focused on conventional internet-based non-work behavior, we argue that AI changes the form of cyberloafing because the same tool can be used seamlessly for both productive and non-work purposes within a single workflow. By distinguishing among avoidance/escape, exploration/curiosity, and social/entertainment uses, the study offers a more granular conceptualization of the heterogeneous motives underlying non-work AI use during work time.

Second, the study contributes to theory by explaining how AI-enabled work conditions translate into nonwork behavior. Rather than treating AI only as a productivity tool or only as a source of workplace risk, this paper develops a more balanced account of the “efficiency paradox” of workplace AI. Specifically, it uses the S-O-R framework to argue that AI-supported work conditions can increase perceived efficiency and create discretionary room in the workday, which may then spill over into AI-enabled cyberloafing. In doing so, our study shifts the conversation from whether AI improves work to what employees may do with the time and cognitive room AI helps create. We define perceived slack as a situational appraisal of discretionary time and low pressure immediately after task completion, distinguishing it from organizational slack (a firm-level resource surplus) (Nohria & Gulati, 1996), time affluence (Kasser & Sheldon, 2009), and psychological detachment (disengaging from work during non-work time) (Sonnentag & Fritz, 2007). We further distinguish it from self-licensing (“I was productive, so I have earned this”) (Merritt et al., 2010), a motivational entitlement rather than a time-resource appraisal, by manipulating available time directly while measuring perceived slack and efficiency separately.

Third, this paper contributes to practice and governance by introducing organizational monitoring as a boundary condition on the efficiency-cyberloafing relationship. We emphasize that monitoring is not simply a control mechanism, but it is discussed as a contextual factor that may shape whether AI-enabled efficiency gains are redirected toward work or toward non-work behavior. By testing this mechanism in a 2 × 2 vignette experiment that varies monitoring and AI-freed time slack, the study provides an initial governance perspective on how organizations might preserve the productivity benefits of AI while limiting unintended misuse. Participants are employed U.S. adults recruited via Prolific who currently use AI tools at work. All vignettes present a common workplace scenario in which a knowledge worker uses an AI assistant to complete a client update task. Beyond deterrence, we advance a redirection hypothesis: monitoring may not reduce total non-work use but shift it from overt channels toward the harder-to-detect in-workflow AI channel (a monitoring × channel interaction rather than a uniform main effect). To attribute effects to the manipulation rather than sample composition, we control for trait cyberloafing and task self-efficacy (Bandura, 1997), check randomization balance across the four cells, including attention, comprehension, and realism checks.

This paper develops and tests a behavioral account of AI-enabled cyberloafing: perceived efficiency increases perceived slack after task completion, which employees may redirect toward non-work behavior depending on the monitoring context, across three conceptually distinct cyberloafing forms. In doing so, it moves beyond productivity-focused accounts of workplace AI and offers a practical view of how employees allocate post-task discretionary room and how organizations might govern it.

References

Dell'Acqua, F., McFowland, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science. https://doi.org/10.1287/orsc.2025.21838

Jiang, H., Siponen, M., Jiang, Z. J., & Tsohou, A. (2024). The impacts of internet monitoring on employees' cyberloafing and organizational citizenship behavior: A longitudinal field quasi-experiment. Information Systems Research, 35(3), 1175–1194. https://doi.org/10.1287/isre.2020.0216

Download

Abstract Only

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.