Artificial Intelligence, Lean Startup Method, and Product Innovations

Although artificial intelligence (AI) has the potential to drive significant business innovation, many firms struggle to realize its benefits. We investigate why some firms succeed in using AI for innovation, whereas others fail, focusing on the organizational support necessary for leveraging AI in both novel and incremental innovation. Specifically, we examine how the lean startup method (LSM) influences the impact of AI on product innovation in startups. Analyzing data from 1,800 Chinese startups between 2011 and 2020, alongside policy shifts by the Chinese government in encouraging AI adoption, we find that companies with strong AI capabilities produce more innovative products. Moreover, our study reveals that AI investments complement LSM in innovation, with effectiveness varying by the type of innovation and AI capability. We differentiate between discovery-oriented AI, which reduces uncertainty in novel areas of innovation, and optimization-oriented AI, which refines and optimizes existing processes. Within the framework of LSM, we further distinguish between prototyping—focused on developing minimum viable products—and controlled experimentation—focused on rigorous testing such as A/B testing. We find that LSM complements discovery-oriented AI by utilizing AI to expand the search for market opportunities and employing prototyping to validate these opportunities, thereby reducing uncertainties and facilitating the development of the first release of products. Conversely, LSM complements optimization-oriented AI by using A/B testing to experiment with the universe of input features and using AI to streamline iterative refinement processes, thereby accelerating the improvement of iterative releases of products. As a result, when firms use AI and LSM for product development, they are able to generate more high-quality products in less time. These findings, applicable to both software and hardware development, underscore the importance of treating AI as a heterogeneous construct because different AI capabilities require distinct organizational processes to achieve optimal outcomes. This paper was accepted by D. J. Wu, Special Issue on the Human-Algorithm Connection. Funding: Financial support from the Mack Institute for Innovation Management is gratefully acknowledged. Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2022.03905 .