Agentic AI in Supply Chains: Balancing Trust and Autonomy in Disruption Scenarios
Abstract
Artificial Intelligence (AI) is no longer just a tool that supports human decisions. It is becoming an active participant in the economy, one that negotiates, transacts, and coordinates autonomously with other AI agents at a speed and scale that no human organisation can match. This shift, from AI as assistant to AI as autonomous economic actor, represents one of the most significant innovations in how organisations operate, compete, and create value. Yet this innovation is outpacing governance. As AI agents increasingly interact with each other rather than with people, organisations are deploying autonomous systems without fully understanding when those systems should act independently, when they should defer, and when they should escalate to human oversight. Without answers to these questions, the promise of AI-driven innovation carries substantial and poorly managed risk. Supply chains are where this tension is most directly felt. They are already among the most complex environments organisations manage, and they are precisely where the cost of getting autonomous AI wrong is highest. Disruptions propagate fast, decisions must be made under uncertainty, and the consequences of miscalibrated autonomy can cascade across entire networks of operators, suppliers, and partners. This research proposes an innovative governance framework that addresses this challenge directly. By developing Trust-based thresholds that dynamically regulate how much autonomy an AI agent exercises depending on the reliability of its counterparts, the research introduces a new model for governing Agent-to-Agent interactions in high-stakes supply chain environments. The framework is grounded in empirical data from the European Gas Transmission Network and validated against the 2022 European Gas Supply Crisis, one of the most significant supply disruptions in recent European history. The expected outcome is both a theoretical contribution, an innovative framework for Trust-governed Autonomy in multi-agent systems, and a practical one: empirical evidence that supply chains governed by this model can recover from disruption autonomously, without continuous human intervention, opening new possibilities for resilient, AI-enabled supply chain innovation.