Logistics

Background: The rapid diffusion of large language models (LLMs) such as Claude, ChatGPT, Gemini, LLaMA, and Mistral is reshaping logistics and supply chain management by embedding generative intelligence into planning, coordination, and governance processes. While prior studies emphasize algorithmic capability, far less is known about how differences in diffusion pathways shape productivity outcomes, managerial cognition, and institutional control. Methods: This study develops and applies an integrative analytical framework—the AI Diffusion Triad—comprising Productivity, Perspective, and Power. Using comparative qualitative analysis of five leading LLM ecosystems, the study examines how technical architecture, access models, and governance structures influence adoption patterns and operational integration in logistics contexts. Results: The analysis shows that diffusion outcomes depend not only on model performance but on socio-technical alignment between AI systems, human workflows, and institutional governance. Proprietary platforms accelerate productivity through centralized integration but create dependency risks, whereas open-weight ecosystems support localized innovation and broader participation. Differences in interpretability and access significantly shape managerial trust, learning, and decision autonomy across supply chain tiers. Conclusions: Sustainable and inclusive AI adoption in logistics requires balancing operational efficiency with interpretability and equitable governance. The study offers design and policy principles for aligning technological deployment with workforce adaptation and ecosystem resilience and proposes a research agenda focused on diffusion governance rather than algorithmic advancement alone.

Background: The administration and management of humanitarian logistical operations are considered critical factors. The extreme and unexpected nature of events such as natural disasters poses logistical challenges to humanitarian support agencies providing aid to the affected area. These logistical challenges are characterized by fluctuations in demand that generate uncertainty in the required capacity for aid, all of which are emphasized at the last link in the humanitarian supply chain: the point of distribution for humanitarian aid (the so-called last-mile problem). The objective of this research work is to support the decision-making process regarding capacity adjustments and the closure of the distribution points established in the disaster area. Methods: In response, a Markovian Decision Model for Capacity Adjustment was developed, focused not only on reducing traditional logistics costs but also on minimizing human suffering by incorporating so-called deprivation costs. Results: The model establishes adjustment policies for capacity for each aid period, and the existence of a monotonous policy that establishes an optimal threshold for closure decisions was demonstrated. Conclusions: It is possible to efficiently adjust the capacity at the distribution points and minimize the costs (both logistical and deprivation) associated with each decision period.

Background: This study evaluates an additive manufacturing (AM) network designed to balance economic performance, lead time, and environmental impact within the healthcare logistics and supply chain. Methods: An integrated framework is proposed that identifies optimal AM facility locations using spatial K-means clustering and optimizes delivery routes through a multi-objective vehicle routing problem with time windows (MOVRPTW). This framework was applied to a case study in Phra Nakhon Si Ayutthaya, Thailand, utilizing hospital geocoordinates, demand profiles, and CO₂ emission factors to evaluate centralized versus decentralized network configurations. Results: Findings demonstrate that hub structures derived from K-means clustering achieve the highest economic efficiency, reducing the AM part cost per unit to 698.51 Baht. In contrast, a fully centralized network resulted in a significantly higher unit cost of 4759.79 Baht, while clustering based on hospital types yielded a unit cost of 959.34 Baht. Quantitative results indicate that the multi-objective approach provides a superior trade-off, achieving lead time requirements while maintaining operational costs and emissions. Conclusions: The results indicate that the proposed framework, particularly through spatial clustering, offers a practical decision-support tool for designing AM networks that achieve a balance between operational efficiency and sustainability objectives in healthcare logistics.