This paper surveys the existing academic literature on artificial intelligence (AI) technologies for anti-money laundering (AML). We review the state-of-the-art AI methods for AML and extend the discussion by proposing a framework that utilizes advanced natural language processing and deep-learning techniques to facilitate next-generation AML technologies. Our framework utilizes unstructured external information to assist domain experts, aiming to decrease the workload for the human investigator. We bridge the gap between the current AML methods and state-of-the art AI, highlighting new trends and directions in AI that can be used to develop the AML pipeline into a robust, scalable solution with a reduced false positive rate and high adaptability.