Systematic Process Selection for LLM-Based AI Implementation : An Operations Management Perspective on Efficiency in Administrative and Knowledge Work

Sammanfattning

Organizations are adopting LLM-based AI rapidly, yet most struggle to translate adoption into real value. This thesis examines how operations management frameworks can explain where and under what conditions LLM-based AI improves administrative and knowledge work processes. The study is based on twelve semi-structured interviews across twelve organizations in the Swedish energy and utilities sector, in collaboration with Intric AB. The theory that is used combines Lean waste analysis, Task-Technology Fit theory and organizational condition assessment. LLM-based AI primarily reduces the lean wastes skills underutilization and motion by freeing qualified staff from routine execution and making information directly accessible, addressing inefficiency within individual process steps rather than between processes. The four process-fit dimensions examined in this study each influence a different aspect of implementation, and the four organizational conditions help explain varying results between organizations. These findings are synthesized into a practical selection framework where organizational readiness functions as a boundary condition. Beyond efficiency, capability expansion emerged as a distinct outcome where AI enables previously infeasible work rather than effectivizing existing tasks. The main theoretical contribution is the extension of Task-Technology Fit logic from the individual to the process level. The main practical contribution is a selection guidance framework offering organizations a structured alternative to unstructured experimentation.