Which Management Control System principles and aspects are relevant when deploying a learning machine?

How shall a business adapt its management control systems when learning machines enter the arena? Will the control system continue to focus on humans aspects and continue to consider a learning machine to be an automation tool as any other historically programmed computer? Learning machines introduces productivity capabilities that achieve very high levels of efficiency and quality. A learning machine can sort through large amounts of data and make conclusions difficult by a human mind. However, as learning machines become even more complex systems, they introduce an uncertainty not previously considered by automation tools. The algorithms can make their own associations, and the automation engineer will no longer know exactly how a learning machine produces its outcome. What is the motive for a learning machine’s decision? A learning machine in this context becomes more human-like compared to the older generation of automation computers. This thesis concludes that most contemporary Management Control System principles are relevant when deploying machine learning, but some are not. A Management Control System must in contradiction to a historically programmed computer, consider multiple human-like aspects while controlling a deployed learning machine. These conclusions are based on empirical data from web-articles, TED-talks, literature and questionnaires directed to contemporary companies using machine learning within their organizations