Prerequisites for an AI implementation in a safety-critical supply chain

Sammanfattning

The increasing use of artificial intelligence (AI) within supply chain management creates new opportunities for organizations to improve decision-making, efficiency, and resilience. However, implementing AI in security-critical industries such as the defense sector involves significant organizational, technical, and security-related challenges. This study explores which organizationaland digital conditions may enable successful AI implementation within supply chain managementin a security-sensitive and complex industrial environment. The study also examines howorganizational structures, support, and ways of working influence the implementation process andthe organization’s ability to utilize AI in practice. The research was conducted as a qualitative case study at BAE Systems Bofors. Empirical data wascollected through pilot interviews, thirteen semi-structured interviews with employees fromdifferent supply chain functions, and a complementary system walkthrough. The collected data was analyzed using thematic analysis. The findings identify four central themes influencing AI implementation: Running on Empty, Mindthe Gap, The Missing Blueprint, and Built for Yesterday. The results show that successful AI implementation depends not only on technical capabilities, but also on organizational readiness, structured processes, high-quality data, competence development, and clear governance. Keybarriers include resource constraints, lack of standardized work processes, insufficient systemknowledge, fragmented data management, and limited trust in digital tools and AI-based solutions. The study concludes that AI implementation in security-critical supply chains requires a holisticand structured approach where technology, processes, people, and security considerations areintegrated. Based on the findings, the study proposes a framework intended to support future AI implementation initiatives within the case company and similar security-sensitive industrialcontexts.