METHODOLOGICAL APPROACH TO THE ASSESSMENT OF INNOVATION RISKS IN THE INTEGRATION OF AI SYSTEMS (ADAS/CMS) INTO THE OPERATIONS OF TRANSPORT ENTERPRISES

Abstract

The article substantiates a methodological approach to assessing innovation risks arising from the integration of Artificial Intelligence systems (ADAS/CMS) into the operations of domestic transport enterprises. The approach consists of four stages: 1) preparation and context establishment; 2) risk identification and classification; 3) risk assessment and analysis; 4) mitigation, monitoring, and strategy refinement. The study provides a systematization and categorization of key AI integration risks into six categories: technological (technical aspects of AI functioning, including component, algorithm, and integration reliability); operational (risks arising from AI-human interaction and enterprise processes); cybersecurity (threats of external interference affecting data integrity, control, and AI system confidentiality); legal and insurance (regulatory compliance, liability, and financial protection mechanisms); ethical and social (moral, societal, and psychological aspects, including bias and AI perception); and financial (economic aspects of AI implementation and operation). The dynamic risk assessment matrix is built as a 5x5 grid, taking into account the specifics of enterprises, such as: operational resilience, interaction with infrastructure and stakeholders. The approach was validated using SCRAM scenario modeling and expert methods. It was established that cybersecurity, financial, legal, and insurance risks pose the greatest threat, while ethical and social risks are moderate yet dynamic, requiring continuous monitoring. The developed risk mitigation strategies include technical, organizational, regulatory, and social measures that ensure the reduction of threat levels to manageable values. The proposed methodological approach is universal and can be adapted to various types of domestic transport enterprises with different levels of automation and stages of AI system implementation.