The Development of Al in Multinational Enterprises - Effects upon Technological Trajectories and Innovation Performance

Abstract

This paper investigates how the development of AI-related inventions by Multinational Enterprises (MNEs) affects their technological trajectories and innovative performance. I combine a matched-pair analysis with an extension of the Difference-in-Difference method to analyse these effects over a novel panel dataset of MNEs. This dataset links over 30 thousand MNEs to more than 10 million patents that these companies owned directly or indirectly (i.e., through their subsidiaries) in the period from 2011 to 2019. The results indicate that MNEs introducing AI-related inventions increase the relatedness of subsequent inventions by about 10 per cent compared to a control group. These results are robust when accounting for a self-selection bias. AI is thus being used to reinforce the existing technological trajectories, rather than to disrupt them. The results also suggest that the number of subsequent inventions is about 40 per cent higher for MNEs that introduce AI during the observation period compared to the control group, without significant effects on the intensity of R&D expenditures per invention. It is argued that this increase in innovative performance is linked not only to knowledge dynamics created by learning about AI but also by AI’s technical potential to be used for learning.