AI Adoption in Reverse Logistics Condition Assessment : An Industrial Economics Perspective on Barriers, Enablers, and Circular Economy Requirements

Sammanfattning

As telecommunications networks evolve, large volumes of Radio Access Network (RAN) equipmentare returned during modernization and vendor-swap cycles and routed through reverse-logistics processes. Determining whether these units should be reused, repaired, or recycled is essential for operational efficiency, cost control, and compliance with circular-economy and sustainability requirements. Today, condition assessment relies heavily on manual inspection, technician judgement,and non-standardized routines, making the process time-consuming, inconsistent, and vulnerable to error. Although Artificial Intelligence (AI) has demonstrated strong potential in inspection and classification tasks, its adoption in telecom reverse logistics remains limited due to variable physical inputs, incomplete documentation, fragmented data sources, and increasing sustainability-driven reporting demands. This thesis investigates how data quality, workflow characteristics, and sustainability requirements influence the feasibility and organizational readiness for AI-supported condition assessment of returned RAN equipment. The study follows a qualitative single-case design and draws on semi-structured interviews with practitioners across logistics, repair, sustainability, field operations,and AI/ML functions, complemented by internal documents, workflow descriptions, and sustainability reports. This multi-actor perspective provides a 360-degree view of the operational, organizational,and regulatory factors shaping AI adoption in reverse-logistics environments. Four analytical perspectives guide the analysis. Sociotechnical Systems theory explains the interaction between human routines, tacit knowledge, and technical systems. Data Quality Framework highlights accuracy, completeness, and consistency as prerequisites for reliable AI performance. AI Readinessand Maturity Models assess organizational capabilities, governance, and cross-functional alignment. Circular Economy principles clarify sustainability objectives such as reuse maximization, waste reduction, and regulatory compliance. Together, these frameworks provide an integrated Industrial Economics perspective on AI adoption, linking technology integration to resource efficiency, coststructures, and value recovery. The findings show that inconsistent defect data, incomplete metadata, and fragmented information systems constrain AI feasibility. Manual inspection routines depend heavily on tacit knowledge, limiting standardization and complicating algorithmic integration. Sustainability objectives create both incentives and constraints, as organizations must meet circular-economy targets while also generating auditable data for regulations such as the EU Corporate Sustainability Reporting Directive (CSRD). Organizational readiness is shaped by coordination practices, knowledge distribution, and governance structures, with significant gaps between strategic ambitions and operational realities. The study contributes to literature by examining AI adoption in a domain where tacit knowledge, workflow variability, and sustainability pressures intersect, an area largely overlooked in prior research. It offers practical insights for decision-makers seeking to improve reverse-logistics performance, strengthen data governance, and advance circular-economy objectives.