Tag Archives: exploitation

How Does AI Change Drug Development? Evidence from Clinical Trial Phases and Drug Types

Kwon, Angela Eunyoung, Jaecheok Park, Gene Moo Lee. “How Does AI Change Drug Development? Evidence from Clinical Trial Phases and Drug Types,” Working Paper.

  • Presentations: KrAIS (2025), CIST (2025), INFORMS (2025), UBC (2025), WISE (2025)

We examine how pharmaceutical firms’ AI capabilities influence drug development outcomes, focusing on clinical trials. Clinical trials progress through three phases that differ in regulatory scrutiny and evidentiary requirements. We measure firm-level AI capabilities using job postings and clinical trial outcomes using the number of trials initiated across phases. We find no significant overall effect of AI capabilities on clinical trial activity. However, this average relationship masks meaningful heterogeneity. AI capabilities are associated with increases in incremental innovation (refinement trials) but not radical innovation (new trials). These effects are stronger for biologics, where market incentives are high, than for small-molecule drugs, where learning hurdles are relatively low. AI capabilities also matter more in early-phase trials, where regulatory barriers are lower, and have no detectable influence in Phase III. This study contributes to the healthcare IS literature by identifying the nuanced and context-dependent business value of AI in drug development. It also offers practical guidance for pharmaceutical firms and policymakers on where AI investments are most likely to enhance R&D productivity.

Balance by Machine Redirection? The Role of Machine Learning Investments in Organizations’ Innovation Search and Long-Term Survival

Lee, Myunghwan, Timo Sturm, Gene Moo Lee, “Balance by Machine Redirection? The Role of Machine Learning Investments in Organizations’ Innovation Search and Long-Term Survival”, 2nd round R&R at MIS Quarterly.

  • Presentations: JUSWIS 2024, KrAIS Summer 2024
  • Best Short Paper Award at KrAIS Summer Workshop 2024.

Organizations’ long-term survival depends on their ability to balance innovation search between explorative and exploitative innovation outputs. Machine learning (ML) investments can reshape this balance by enabling organizations to develop capabilities that generate innovation opportunities from data beyond those readily accessible through human-led innovation. Yet, it remains theoretically unclear whether ML helps organizations move toward greater balance or reinforces their innovation imbalances. We theorize that ML investments operate as a context-dependent directing mechanism: how they affect innovation search depends on organizations’ prior innovation tendencies and the types of ML capabilities they develop. We test this theory using a novel organization-year-level measure of ML investments in a longitudinal sample of 2,916 organizations. Combining panel regressions, long-difference analyses, survival models, and interviews with ML practitioners, we find that ML investments are associated with greater innovation balance by strengthening organizations’ underrepresented innovation tendencies: they redirect exploitation-oriented organizations toward explorative innovations and exploration-oriented ones toward exploitative innovations. The shift toward explorative innovations is strongest for organizations with unsupervised capabilities, and our survival and mediation analyses provide suggestive evidence that greater innovation balance may be linked to lower organizational failure risk. Our study extends research on IT-enabled innovation by showing that ML can become an additional source of innovation whose direction depends on organizations’ prior innovation search. In doing so, it offers a context-dependent view of ML-enabled innovation and explains how ML investments can help organizations counteract path-dependent innovation imbalances.