Predictably Satisfied: Contributions of Artificial Intelligence to Intra-Organizational Communication

Abstract

Artificially intelligent (AI) applications make data-driven predictions that enable personalization on a large scale. As such, recent advances in AI's predictive power might have the potential to create more productive work environments. Using a principal-agent model and understanding AI as signal production technology, I show that, within an organization, higher accuracy of an AI's predictions reduces information asymmetries and fosters truthful communication. Detailed information about employees allows for individually tailored management which ultimately raises production and profits. Using observational data, I test the main implications of the theoretical model concerning optimal behavior and heterogeneity in employee satisfaction. I exploit a unique individual-level panel dataset from Attuned, a Japanese startup that developed an AI tool measuring employees' intrinsic motivation. Empirical results show that when communicating via the AI tool, employee satisfaction increases over time. The effect is particularly pronounced in the long run, for initially dissatisfied individuals, in small teams, and for those whose motivational profile resembles that of their teammates. This heterogeneity suggests personalized work experiences due to managers' better targeting.