The Dominant Do Not Share: Limits to Data Network Effects in Oilfield AI Technology

• In Event: Middle-Range Theorizing and Its Empirical Tests

Abstract

We identify limits to the “data network effects” often attributed to artificial intelligence (AI) technology. Consulting firms (“service providers”) in the digital oilfield industry rely heavily on AI-based predictive models, which seem poised to help generalist providers increase their market concentration by leveraging the disproportionate amount of data they receive from their clients (“operators”). Yet the market continues to feature a strong presence of specialist providers. Our key argument—inductively derived from interviews with consultants, engineers, data scientists, and managers—centers on market-dominant clients’ restrictions on the use of their data. Such clients view their data as core competitive assets and therefore work only with service providers who use the data for project-specific purposes, prohibiting its incorporation into generalized tools beyond the immediate engagement. By contrast, specialist providers working with less dominant, more niche clients often negotiate rights to retain and reuse client data, on the grounds that these clients face less competition and perceive mutual benefit in co-developing analytics capabilities. Over time, these arrangements enable specialist providers to build richer, more differentiated datasets and to train models that, in some domains, surpass those of their generalist rivals.

Authors

• Isha Dhruva, Rice University

• Minjae Kim, Yale University