The modelling of deception is an active field in artificial intelligence. Episodic formulae were introduced, applied and refined in several previous articles, some of them concerned with identity, and some, with tasks of guessing agents' minds. The AURANGZEB project is specifically concerned with combining this kind of representation with argument structure data (in particular, Toulmin's). Two previous articles were concerned with the MURAD subproject, whereas this paper introduces the AJIT model. In MURAD, the kind of deception consisted of having a victim misunderstand a perpetrator's real intentions. In AJIT, instead, a powerful agent is outsmarted by means of exchanged identities.