AI Camera as a Control Technology: Evidence from the Field

Abstract

Advances in artificial intelligence (AI) have created new opportunities to improve organizational productivity, yet little is known about the conditions under which AI improves productivity. We study a seaport stevedoring firm that implemented an AI-assisted camera system generating different types of alerts through internal and external cameras. Drawing on the enabling-coercive control framework, we classify these alerts according to the form of control they embody and examine their effects on employee productivity. We find that the adoption of the AI-assisted system is associated with lower productivity. However, further archival and interview evidence reveals substantial heterogeneity across alert types. Alerts associated with coercive control are linked to lower productivity, whereas alerts associated with enabling control are linked to higher productivity. Our findings suggest that productivity gains from AI are more likely to arise when AI augments, rather than constrains, employee decision-making.

Keywords: Artificial Intelligence, Management Control, Process Control, Monitoring Technology, Decision Facilitation

JEL Classification: D24, D82, M41, M54, O33