Work

Capability atrophy: When AI automation erodes the dynamic capabilities it was meant to build

Year: 2026
Type: article
Abstract: Firms adopt artificial intelligence (AI) automation to strengthen their dynamic capabilities, yet automation can also remove managers from the routines through which those capabilities are renewed. Drawing on the microfoundations of dynamic capabilities, organizational learning theory, and the automation–augmentation paradox, this conceptual paper develops the concept of capability atrophy and specifies when AI automation weakens, rather than builds, firm-level dynamic capabilities. Theory synthesis across research on skill erosion, organizational learning, and human–AI collaboration yields nine propositions, together with a multilevel, multi-wave mixed-methods design that uses behavioral and system-discontinuance measures to test them. The propositions hold that automation intensity relates to dynamic capability change in an inverted-U pattern, turning negative beyond a substitution threshold. Erosion travels through reduced experiential engagement, weaker skill retention, and poorer reliance calibration, and it becomes organizational when tacit knowledge is concentrated and apprenticeship paths close. Retained decision roles, frequent exceptions, and task variety are proposed to raise the threshold, whereas analytics capability alone is not. The paper places capability atrophy within existing skill-erosion research and adds a firm-level, conditional account with explicit falsification criteria. For practice, it implies that managers should track override rates, unaided diagnosis time, and other observable indicators of capability retention alongside the returns on AI investment.
Language: English
Open Access status: hybrid
OpenAlex ID: W7237349141
FWCI:
Cites: 61
Cited by:
Related to: