Study uses AI to automate and explain backup supplier selection

Logic-driven artificial intelligence can now automate and explain decisions on when to bring in backup suppliers during supply chain disruptions, according to a study published in Engineering Applications of Artificial Intelligence. Industrial engineering professor Dr. Omid Fatahi Valilai and doctoral researcher Omkar Vishwas Patil developed the framework to support resilient and responsive supplier selection without relying solely on lowest costs or automatic activation of every alternative.

The Tech TL;DR:

• Researchers at Constructor University built a logic-driven AI framework using answer set programming (ASP) to automate supplier selection during disruptions.
• The system evaluates operational continuity, cost, and social and environmental thresholds, activating backup suppliers conditionally rather than automatically.
• Tested across 27 synthetic scenarios with up to 1,000 primary suppliers, the model achieved a 100% fill rate while maintaining an audit trail for human staff.

Answer Set Programming Drives the Decision Layer

Supply chain disruptions trigger compounding delays and rising costs for manufacturers scrambling for alternative vendors. To address this, Fatahi Valilai and Patil integrated answer set programming—a form of declarative AI—into robotic process automation and enterprise resource planning systems. This combination forms a closed-loop system spanning initial decision-making to final execution.

The core ASP layer translates procurement rules, supplier capacities, disruption conditions, and sustainability requirements into explicit instructions. Rather than simply recommending a new vendor, the framework provides an audit trail showing the chain of reasoning. This trail includes the specific disruption that compromised a primary supplier, the determination that demand cannot be met by remaining vendors, and the conditional activation of backup options. Human decision-makers such as procurement managers and auditors can review, assess, and adjust these recommendations.

Evaluating Supplier Networks Through Synthetic Scenarios

The research team evaluated the model using a reproducible synthetic data set featuring networks of 100, 500, and 1,000 primary suppliers paired with corresponding backup suppliers. Simulations tested partial supplier losses, the full loss of high-capacity suppliers, and multiple simultaneous outages across varied demand levels.

Across 27 simulated scenarios, the framework achieved a 100% fill rate while enforcing required social and environmental standards. It triggered backup suppliers solely when primary capacity failed to meet demand.

The data set and supporting code are publicly available through Figshare and GitHub.

Future Research Targets Live Systems and Actual Supplier Data

While the simulation results demonstrate operational continuity and transparency, the current implementation relies entirely on synthetic data and a mock enterprise resource planning environment. A full deployment involving live enterprise systems, real-time disruption signals, and actual supplier data remains a target for future research, according to the authors.