Artificial Intelligence (AI) in Command and Control (C2) raises questions about the interaction between operators on different levels on the one hand and AI-supported information systems on the other. For some purposes, the aggregation and analysis of large “big data”-sets creates potential for prediction and calculations of probabilities, while for other purposes human heuristics may be as promising depending on the situation. What has been more rarely discussed, is the very quality of data underpinning such calculations, and thus also operators’ awareness of the validity of predictions in relation to any specific situation. This is an urgent debate, considering the fact that full transparency may be impossible and underpinning data may be based on exercises, simulations, real-time data, or a mix thereof. Based on a fundamental classification of different kinds of uncertainties, this article discusses how data of different origins and quality can be managed and communicated to allow for operators to assess on what basis predictions are made. The article further suggests that looking at other fields of research may be useful for exploring unconventional ways of highlighting the existence and quality of different kinds of data of different origins in order to assess its predictive power. Finally, the article discusses how AI may change the role of analysts with regard to such issues.