From experimentation to scaling: what shapes the funnel of AI adoption?

Firms tend to manage the diffusion of complex and uncertain technologies through a 'funnel', whereby firms first experiment (test and learn), then decide to exploit and scale technology. We study how internal and external factors typically affecting Artificial Intelligence (AI) adoption influence the successive steps of the funnel for a sample of large firms worldwide. We demonstrate that competition influences AI adoption more at the exploitation than at the experimentation phase. Conversely, technology complements are more relevant at the start of the funnel, in contrast to organizational complements which are most relevant to ensure embedding of AI technologies in business practices.