Optimizing Supplier Delivery Reliability in Project-Driven Supply Chains through AI-Supported Value Stream Mapping: A Data-Driven Framework for Enhancing Coordination and Predictive Control

This thesis explores how artificial intelligence (AI) can be integrated into project-driven supply chains (PDSCs) to improve supplier delivery reliability. The research focuses on a real Engineer-to-Order (ETO) project in the energy sector, where delivery delays and coordination gaps exposed structural weaknesses in procurement and logistics. A simplified Value Stream Mapping (VSM) approach was used to analyze the delivery flow and identify where reliability breaks down. Three AI models were introduced to support earlier decision-making. LightGBM was positioned to flag delay risks after supplier confirmation. A fuzzy inference system (FIS) was used to assess shipment readiness from incomplete signals. Clustering was applied for post-delivery supplier segmentation. One model—FIS—was simulated using real shipment data and confirmed to match high-frequency coordination failures. The others were validated through interviews and mapped to case data. Applied studies showed that these models improve forecasting reliability and streamline logistics decision-making in environments similar to the case context. The redesign does not propose automation. It introduces predictive checkpoints where human judgment is weakest. Procurement and logistics staff confirmed the realism of the approach and expressed interest in piloting readiness scoring and segmentation tools. A phased roll-out strategy was proposed, starting with Excel and PowerBI overlays before deeper integration. The findings show that interpretable AI can improve operational clarity and strengthen Lean coordination—without disrupting existing workflows. This creates a foundation for smarter, more reliable project execution.