It is a big challenge in the information technology field to develop techniques to help users in their tasks. For that purpose, we need to develop assistants able to help people without disturbing them in their main task. This paper discusses building an assistant based on traces and making use of a trace-based system. Traces can be used as a knowledge source to discover other useful knowledge but also to reuse experience. The main idea of the approach is to find useful episodes (previous experiences) in interaction traces and to reuse them to provide users with contextualized help. The assistant can then adapt the retrieved episodes to provide assistance to the user by reusing previous experiences. This work has been done in partnership with SAP-BO. We have implemented our proposal in the SAP explorer project which aims to ensure that all business users have easy access to all the information they need to make confident decisions based on up-to-date, reliable information, so they need to help business users while doing their tasks. The main contribution described in this paper is the algorithm enabling us to retrieve past episodes corresponding to a “task signature” (a description of the main characteristics of the episode). For that purpose, we have implemented a solution based on Finite State Machines.