Recently, rehabilitation researchers have identified a number of problems with current Dictionary of Occupational Titles (DOT) based computerized job-matching systems. These DOT based systems tend to utilize a straight-line search and employ rigid cut off scores in job selection for people with disabilities. This method provides profiles without considering the adaptability of individual clients and the possibilities of job modification and accomodation. Consequently, these systems do not take advantage of the wealth of clinical knowledge developed by experienced rehabilitation professionals over years of clinical practice. Most of these systems are overly concerned with the "person-job requirements fit" aspects and basically overlook the "person-job environment fit" needs. As an alternative, this paper presents a conceptual framework for developing a second generation computerized job-matching system. Utilizing some common elements from the current literature on artificial intelligence, the' authors conceptualized the development of a knowledge-based (expert) job-matching system capable of reasoning, making decisions as would a vocational expert, learning from cumulative job matching experience, and handling ambiguous data commonly found in real life job-matching situations. The conceptualized system uses "fuzzy set" mechanisms to maximize the amount of information available for a successful job search.