Once defined the development environment, the implementation models were developed according to the modular structure of the conceptual models, as follows: (i) UFO.owl; (ii) BusinessProcess.owl; (iii) ITService.owl; (iv) ITComponents.owl and finally (v) ConfigurationItem.owl. Due to the expressivity restrictions inherent in the implementation languages, the main issue concerning the mapping from conceptual models into implementation models is related to the treatment of the reduction in semantic precision. In order to maintain this reduction at an acceptable level, the most relevant losses that were found were related to the transformation of all ontologically well-founded concepts and relations into OWL classes and properties, respectively. Regardless of the application scenario, this mapping must consider the information contained in the notation used for the development of the conceptual models, such as cardinality, transitivity, domain and range. With respect to cardinality and transitivity, in OWL it is not possible to represent them simultaneously (Bechhofer et al., 2004). As a result, this work considers that the representation of cardinality restrictions is more relevant to the implementation models developed in this section. In addition, to represent the cardinality restrictions in both directions inverse relations were used. For instance, the relation “requests” is represented by the pair of relations “requests” and “is requested by”. However, according to Rector and Welty (2001), the use of inverse relations significantly increases the complexity of automated reasoning. Thus, they should be used only when necessary. With respect to domain and range, an issue that should be considered is how to organize and represent many generic relations. For example, if a generic relation “describes” is created, it is not possible to restrict the domain and the range. In this case, the design choice was to use specific relations like “describes_Software_ComputerProcess”, which is represented as a sub-relation of a generic relation “describes”. Finally, with respect to SWRL restrictions, this language has neither negation operators nor existential quantifiers (Horrocks et al., 2003). In addition, the SWRL language might lead to undecidable implementation models. Nevertheless, this issue may be worked around by restricting the use of rules and manipulating only those that are DL- safe (Motik et al., 2005). As an attempt to make this tangible, consider an implementation of the axiom A7, which concerns the competence question QC2, discussed in Section 4. This implementation is represented by the rule R7a.