The goal of this master thesis is to investigate the characteristics of cognitive automation and its limitations when applied to business processes or tasks. It is motivated by the increasing relevance of advanced process automation through the use of cognitive computing capabilities, including artificial intelligence and especially machine learning. Companies that reduce their operating costs and increase their effectiveness have a competitive advantage. This can be done by using robotic process automation to automate repetitive and mundane everyday tasks that are inherently prone to human error. Cognitive automation improves upon this by further extending the usefulness of traditional automation through robotic process automation by automating (partially) unstructured and cognitively challenging processes/tasks. It can be used both to support human knowledge workers by using natural language processing/generation and as a replacement by directly communicating with customers. In this work, a systematic literature review is used to define the characteristics of robotic process automation, cognitive automation and business processes typically automated by them. It shows that the event-driven approach of cognitive automation allows flexible deployment and widespread use, but is still limited by the capabilities of artificial intelligence algorithms, by problems due to scarce and incorrect data, and by the continuing need for human exception management.