Towards a Pattern Recognition Approach for Transferring Knowledge in ACM

In Adaptive Case Management (ACM) systems, knowledge workers have the flexibility to deal with unpredictable situations. Compared with a classical BPM approach the extensive prescriptive process analysis and definitions are replaced by context-sensitive proposals, which is more suited for
knowledge-intensive work. Thus, it is vital that ACM systems support knowledge workers with knowledge captured from previous work which can be ambiguous for the system. This paper proposes an approach to support knowledge workers based on the knowledge previously applied by others in the form of a User Trained Agent that learns from ad hoc actions taken by knowledge workers to suggest best next actions for the current situation. The proposed best next actions are analyzed for coherence.