From AI Adoption to Exploitation: the Role of Complements and Competition

This paper studies the diffusion of artificial intelligence (AI) within firms, from exploration to local adoption to full-scale exploitation. The optimal timing of technology adoption represents a balance between preempting the risk of competition and time needed to acquire necessary complements, to ensure a successful return on investment. We formulate and test the idea that this balance changes along the adoption curve from experimentation to exploitation. We first model the decision of a firm facing Cournot competition to explore then exploit AI and assess the role of a variety of internal complements (technological and organizational) as well as competitive rivalry in these processes. Based on this theoretical model, a reduced form model of internal diffusion of AI is then estimated. Three results emerge: (1) rivalry triggers a competitive technology race that prevails in the exploitation more than in the exploration phase; (2) direct AI complements (such as machine learning) favor both adoption and exploitation, while indirect complements (such as cloud and big data) matter more for the experimentation than for the exploitation phase; (3) organizational complements are important for exploiting AI at scale, while technological ones drive exploration and adoption more than exploitation.