Human-in-the-Loop AI for resilient supply chain operations: a framework for augmented decision-making and ethical deployment

Purpose This research examines how Human-in-the-Loop (HITL) artificial intelligence (AI) can support resilient decision-making, operational performance, transparency, trust and ethical accountability in operations and supply chain management (OSCM). Design/methodology/approach The research adopts a design science research approach combined with comparative case analysis. A HITL AI framework, termed the adaptive oversight architecture (AOA), was developed and evaluated in two contrasting OSCM contexts. System logs, operational observations, interviews and performance indicators were analysed to examine decision outcomes, human oversight, trust and accountability. Findings HITL integration was associated with improved exception-handling accuracy, reduced human override frequency, shorter task-completion times and higher user satisfaction across the two cases. Qualitative findings further indicate that trust calibration, training, interface design, workflow integration and organizational governance influence how effectively human oversight operates. The findings also identify risks of automation dependence, accountability diffusion and inappropriate reliance on algorithmic recommendations. Research limitations/implications The research is limited to two case firms, which constrains the generalizability of the findings. While the dual-case approach provides in-depth insights, broader studies involving larger datasets or cross-industry comparisons are needed to validate the proposed framework. Future research could extend the empirical base by employing longitudinal analysis, simulations or hybrid models that test HITL effectiveness under different operational conditions. These directions will allow for a more comprehensive understanding of the scalability, adaptability and long-term organizational impacts of HITL-enabled supply chains. Practical implications For practitioners, the research highlights how HITL can be implemented to strengthen supply chain decision-making while ensuring compliance with ethical and regulatory standards. It suggests that organizations should prioritize training programs, design user-friendly interfaces and establish override mechanisms to maximize HITL effectiveness. The findings emphasize the importance of aligning AI investments with human oversight structures, rather than pursuing automation in isolation. Managers can use the framework to evaluate readiness, map decision tasks suitable for HITL integration and design workflows that balance speed, transparency and accountability. Social implications By positioning humans as active collaborators, HITL contributes to responsible AI adoption that safeguards jobs, reinforces accountability and reduces risks of algorithmic bias in supply chain operations. The approach ensures that technology complements rather than replaces human expertise, fostering inclusive digital transformation. From a societal perspective, embedding ethical safeguards into AI decision-making promotes trust in enterprise systems and enhances their legitimacy. The findings encourage the development of AI solutions that not only drive operational efficiency but also align with broader social values of fairness, responsibility and human dignity. Originality/value This research conceptualizes HITL as a dynamic socio-technical arrangement rather than a static control mechanism. The proposed AOA explains how human authority, algorithmic decision-making and organizational governance interact through oversight, trust calibration and distributed accountability, providing a mechanism-based conceptualization of HITL for OSCM.