A small elite group of companies have scaled AI in supply chain operations, delivering triple-digit productivity gains, faster response times and lower error rates, according to new research from GEP and the University of Virginia's Darden School of Business.
The GEP–UVA Darden study finds that while AI experimentation is nearly universal, only about 5% of supply chain AI initiatives have successfully moved beyond experimentation to enterprise-wide scale.
"Most companies aren't failing at AI because of the technology," says Michael DuVall, global head of strategy at GEP and co-author of the study. "They're failing because they're automating broken processes. The companies seeing outsized gains redesign how work gets done, put real governance in place, and hold AI to clear business outcomes before scaling."
Key takeaways:
· While 5% of initiatives have successfully industrialized, the vast majority remain stalled. Specifically, 22% are caught in the pilot phase, and 74% are either stuck in planning or have no formal roadmap for execution, leaving a significant divide between AI potential and operational reality.
· The barrier is not budget, AI and agentic capabilities. It is management discipline.
· Companies that scaled AI are significantly more likely to operate with a dedicated AI steering committee that ties funding directly to enterprise value delivery.
· Rather than approving isolated experiments, successful companies manage AI initiatives as a structured portfolio — progressing use cases deliberately from evaluation to pilot to scale.
· AI scalers document system logic through digital audit trails at materially higher rates than their peers, reinforcing trust, compliance and accuracy.
· Organizations that have scaled AI are 2-3 times more likely to modernize elements of their talent strategy, redefining roles and aligning incentives to AI-enabled operating models.
