AI Decision Support Systems (AI DSS) have attracted interest in the domain of crisis management. Crises are usually complex, the outcome is often uncertain and they usually take place under high time pressure, leading decision-makers to rely heavily on tacit knowledge: experience-based, personal, contextual knowledge. This thesis researches the epistemic consequences of the use of AI DSS on the decision-making process in crisis management, with a specific focus on tacit knowledge. The findings indicate that the introduction of AI DSS can change the knowledge mix in crisis decision-making by increasing the availability and prominence of explicit knowledge. Due to automation bias, decision-makers may prioritize AI-generated outputs over their own tacit knowledge. Furthermore, this thesis argues that the use of AI DSS can enable cognitive offloading of epistemic tasks, which may lead to skill erosion and increased dependence on such systems. This potential erosion poses risks for crisis management in cases where there may be issues with the performance and reliability of AI DSS model outputs, but also as tacit knowledge remains crucial for assessing AI outputs and adapting to novel situations.