Immersive VR simulations provide realistic experience in skills that are expensive or dangerous to learn principally through real world practice, for instance conning large ships. Although the cost of simulators can be amortized over many training sessions, the ongoing cost for human instructors is a serious impediment to optimal utilization of simulations. An artificially intelligent tutoring system (ITS) capable of performing some of an experienced human instructor's role may, without diminishing student proficiency, enable a reduction in dollar and manpower costs by helping students overcome common and predictable problems during simulation-based training. ONR supported an R&D effort to develop and test such an ITS for the Conning Officer Virtual Environment (COVE) shiphandling simulator used to train Naval Surface Warfare Officers (SWO). COVE ITS detects both process (observation of visual cues) and performance (correct orders at the correct time) as conning officers conduct a briefed evolution in COVE. It measures the student's proficiency and detects problems that may arise. Modeled on instructor behavior, the ITS can give pointers before problems become unmanageable, but not before the student has a chance to see the effects of an error. This system is being evaluated at the Surface Warfare Officers School (SWOS). A study to assess proficiency gains of students taught by COVE ITS with oversight by an instructor, compared to proficiency gains of a control group who received the usual one-on-one instructor training used two runs, one instructed and a second uninstructed test run. Data analyzed to date show no differences in performance between the two groups during the test run, either on instructor scoring or on behavioral measures. Additional evaluation will test the effectiveness of the ITS when one instructor supervises two or three COVE stations. Plans for further development include expansion to different levels of students and to a practice-only mode.