Supply Chain Dive published an analysis of why supply chain AI deployments fail, drawing on interviews with West Monroe partner Brian Pacula, Bain and Co. partner Dheera Anand, and University of Tennessee professor of practice Alan Amling. It cites Starbucks' decision, reported by Reuters in May 2026, to scrap its computer-vision inventory counting tool across North America after roughly nine months because it miscounted or mislabeled items. A Gartner survey released April 29, 2026 found 56% of chief supply chain officers see integrating AI with legacy systems as a large hurdle and 50% cite limited internal expertise. The experts advise targeting frequent, time-pressured decisions with available data, validating ROI and data availability before pilots, codifying planners' tribal knowledge, and applying a "left-on test" — whether the system could run unsupervised over a weekend.