The numb efficiency paradox: AI work pressure, affective numbing, and professional judgment in organizationally digitally mediated work.

Artificial intelligence is reshaping accounting work by accelerating task cycles, expanding exception monitoring, and increasing reliance on automated recommendations. In accounting and auditing, these changes are consequential because work quality depends not only on efficient task completion, but also on ethical vigilance, professional skepticism, and careful evaluation of AI assisted outputs. This study introduces the numb efficiency paradox, a process in which accounting professionals may continue to meet work demands while becoming less affectively responsive to pressure, warning signs, and professional consequences. Using a single three wave longitudinal panel of accounting and auditing professionals recruited through occupation based prescreening and study specific eligibility screeners (T1, n = 512; T2, n = 417; T3, n = 361), the study examines AI related work pressure, occupational resilience, affective numbing, sustained work efficiency, ethical vigilance, and professional skepticism. The declining wave-specific sample sizes reflect panel attrition and quality screening rather than separate studies or newly recruited respondents. The results show that AI related work pressure was prospectively associated with both occupational resilience and affective numbing. Occupational resilience was prospectively associated with later sustained work efficiency. Affective numbing was not significantly associated with sustained work efficiency, but it was negatively associated with ethical vigilance and professional skepticism. AI autonomy and supervisor support showed beneficial direct associations with higher resilience and lower affective numbing, whereas the hypothesized interaction effects were not supported. These findings distinguish resilient efficiency from detached persistence and suggest that AI implementation should be evaluated not only by whether accounting work remains efficient, but also by whether professionals remain ethically vigilant and professionally skeptical when using AI supported systems.