Developing industrial AI capabilities: An organisational learning perspective

Incumbent industrial firms are putting in a lot of effort in developing capabilities for machine learning (ML) systems that help them better predict and perform a variety of industrial and business processes and decisions. Given the data-, process-, and organizational structure-related requirements for effective implementation of such systems, these organizations encounter a major challenge in developing capabilities in this context. However, the existing literature has yet to unravel the organizational processes and practices associated with artificial intelligence (AI) capability development and deployment in industrial incumbent firms. The present study frames AI adoption in established industrial firms as a process of history-embedded, situated organizational learning involving explorative and exploitative learning. Based on a qualitative study of seven firms utilizing ML algorithms in their industrial and business processes, we develop a grounded model that explains AI capability building as both enabled and constrained by perceptual and functional triggers and barriers, leveraged via communicative and structural practices, and resulting in ongoing and interdependent processes of exploration and exploitation. The study contributes to the literature by showing how the convergence of organizational learning and AI technology's unique features promotes a distinct dynamic of AI capability building and deployment. • Industrial firms increasingly invest in machine learning to improve productivity. • Such firms often have a heavy asset base and legacy resources. • Developing and deploying AI capabilities is necessary for these firms. • We demonstrate organizational learning processes for a situated AI capability. • We distinguish between explorative and exploitative learning processes.