The successes of using artificial intelligence (AI) in analysing large-scale data at a low cost make it an attractive tool for analysing student data to discover models that can inform decision makers in education. This article looks at the case of decision making from models of student satisfaction, using research on ten years (2008-17) of National Student Survey (NSS) results in UK higher education institutions. It reviews the issues involved in measuring student satisfaction, shows that useful patterns exist in the data and presents issues involved in the value within the data when they are examined without deeper understanding, contrasting the outputs of analysing the data manually, and with AI. The article discusses risks of using AI and shows why, when applied in areas of education that are not clear, understood and widely agreed, AI not only carries risks to a point that can eliminate cost savings but, irrespective of legal requirement, it cannot provide algorithmic accountability.