When suppliers fail: Constructor University researchers develop explainable AI for resilient supply chains

Supply chain disruptions can have swift and compounding consequences for manufacturers, leading to production delays, rising costs, and a scramble to secure alternative suppliers. Two researchers from Constructor University have developed a framework that leverages logic-driven artificial intelligence to provide more resilient and responsive supplier selection decisions. Published in Engineering Applications of Artificial Intelligence, the study demonstrates how AI can be leveraged to support scalable and explainable supplier solutions that account for operational continuity, cost, and social and environmental requirements.

The framework was developed by Professor of Industrial Engineering Dr. Omid Fatahi Valilai and PhD Researcher Omkar Vishwas Patil within the Emerging Technologies in Industrial Engineering research group. At its core lies Answer Set Programming (ASP), a form of declarative AI that acts as a decision-making layer, translating procurement rules, supplier capacities, disruption conditions, and sustainability requirements into clear instructions that the system can transparently evaluate. By integrating ASP into Robotic Process Automation (RPA) and Enterprise Resource Planning (ERP) systems, Prof. Dr. Fatahi Valilai and Patil created a closed-loop system that spans from decision to execution, allowing for decisions to be made and validated, repetitive tasks to be automated, and procurement transactions to be carried out.

“Our goal was to move beyond developing an AI model that simply recommends a supplier, and instead demonstrate how logic-driven AI can support the complete decision-to-execution process,” said Patil. “By building the decision layer on ASP with explicit rules for disruption and sustainability, we can make supplier-selection decisions that are responsive to changing supply chain conditions while remaining transparent and traceable.”

Transparent intelligence for effective decisions

The result is a system that can intelligently determine which suppliers remain available, whether they can meet demand, and whether backup suppliers are needed. It can then activate backup suppliers conditionally rather than automatically selecting every possible alternative. It can also enforce minimum social and environmental thresholds to ensure supplier selection is not solely based on lowest cost.

The framework is designed to make AI-supported decisions transparent and reviewable by human staff. Rather than simply recommending a new supplier, it can show the chain of reasoning behind the recommendation, such as the disruption that made a supplier unavailable, the determination that demand cannot be met through remaining suppliers, and the decision to activate backup suppliers. This audit trail allows human decision makers like procurement managers, auditors, and sustainability teams to understand, assess and, where necessary, challenge or adjust the system’s recommendations—supporting human decision-making rather than replacing it.

The researchers evaluated the model using a reproducible synthetic dataset containing networks of 100, 500, and 1,000 primary suppliers with corresponding backup suppliers. The experiments then simulated different scenarios including partial supplier losses, full loss of a high-capacity supplier, and multiple supplier outages across different demand levels.

Across 27 simulated scenarios involving supplier networks of up to 1,000 primary suppliers, the framework achieved a 100% fill rate while meeting the required social and environmental standards, and activated backup suppliers only when needed. A proof-of-concept demonstration also showed how the resulting decisions could be checked, translated into simulated purchase orders, and recorded in an audit log explaining the actions taken.

Designing tomorrow's supply chains

While promising, Prof. Dr. Fatahi Valilai and Patil stressed that this work used synthetic data and a mock ERP environment. A full deployment with real supplier information, live disruption signals and commercial enterprise systems remains a goal for future research.

“Supply chains are increasingly expected to be resilient, sustainable, transparent, and responsive at the same time,” said Prof. Dr. Fatahi Valilai. “For me, the important question is therefore no longer only how AI can predict what may happen, but how intelligent systems can make operational decisions that remain explainable, auditable, and aligned with sustainability policies.”

The paper, “A Logic-driven Integrated System for Resilient and Sustainable Supplier Selection in Supply Chains,” was published in Engineering Applications of Artificial Intelligence in August 2026. The dataset and supporting code are publicly available through Figshare and GitHub. Read the paper at https://doi.org/10.1016/j.engappai.2026.115955.