Conditions for Machine Learning Implementation : A qualitive study exploring Change Management in a test data analysis and management process

Sammanfattning

Background - Artificial Intelligence (AI), especially algorithms able to identify patterns in extensive datasets called Machine Learning (ML), facilitate challenges related to data management as decision support. This becomes particularly evident in management of growing data volumes and rapid technological advancements within the automotive industry. However, the success of an ML-implementation does not solely rely on technical aspects. The need for preparatory work related to human-machine collaboration should be prioritized by organizations. Management should pay close attention to less tangible factors during such implementations. Purpose - The study explores internal contextual conditions for ML-implementation, to understand how industrial organizations can facilitate change in test data analysis and management processes. The study aims to provide an understanding from a change management perspective by identifying employees' attitudes towards, and perceptions of, ML and translating them into internal contextual conditions. Method - A qualitative research design was applied in combination with an abductive research approach. This enabled the study to be guided by collected data flexibly. By using a critical realism perspective, the research dives deeper into the potential underlying mechanisms influencing the internal contextual conditions. Data collection consisted of 7 in-depth interviews and 1 focus group for data saturation. Results - Research question number 1 was answered using empirical data, and research question number 2 using a theoretical analysis of the empirical results. It concluded ten driving internal contextual conditions and nine restraining, leading to seven underlying mechanisms. Findings - Ensuring employees have sufficient AI and ML knowledge, and access to support functions, alongside implementing a learning cycle for testing and failure, promotes change. Continuous feedback and effective time management are crucial for transferring knowledge from past implementations. Fostering psychological safety enhances trust and encourages a supportive culture. Promoting autonomy increases motivation, while managing uncertainty by building trust in ML-technologies and providing clear definitions is vital. Clear communication and a well-defined vision from management are essential for commitment. Lastly, strategic commitment requires a communication strategy and clear organizational guidelines for ML-implementation to coordinate roles and reduce knowledge gaps. Additionally, balancing internal development and delivery of results, understanding the interest in how ML-tools impact workflows, and accessible strategic models are necessary to bridge knowledge gaps and drive change.