In this study we provide the first documented application of the CBEAT Audit framework as an empirical means of evaluating the potential of using voice technologies to assess applicants for employment in the Ghanaian context. Our research assumes that applicant scoring, vocal characteristic recording, and voice transcription represent a set of culturally specific values about emotional expression and communication competence that disadvantage those who speak Fante from Ghana. Our method for analyzing this issue is through an evaluative qualitative single-case study design using only secondary document data as sources. The process consists of five stages involving reviewing documents selected by their level of relevance to the purpose of this study, extracting pertinent details from these documents through an evidence extraction and mapping process, utilizing the evidence gathered through the extraction and mapping stages to create a composite profile of all evidence collected for each dimension of the CBEAT Model (i.e., social and environmental), conducting validation of the findings against three external regulatory standards (i.e., E.U. AI Regulation and the United Nations Convention on the Rights of Persons with Disabilities), and finally mapping all findings onto the institutional context of Ghana. Our results highlight that for each dimension of the CBEAT structure, there exists a combination of aspects that when combined, create a cumulative risk profile that would be impossible to detect if using only one metric per dimension. For example, the word error rates of automatic speech recognition error systems compared to standard test weights for Ghanaian Akan (Fante) speech provide an 80.3% error rate on commercial systems as opposed to approximately a 5% rate on standard test weights. The issue of regulatory restriction on inference of emotionality from speech scoring systems was also explored. For instance, in the case of the E.U. AI Act, the inference of emotionality for workplace purposes is prohibited under Article 5(1)(f). However, the Ghana Data Protection Act of 2012 does not define biometric information and therefore does not prohibit using emotional inference to score an applicant's suitability for a position in the workforce. The overall contribution of this study is three-fold: 1) it will enable a new and practical method of replicating the implementation of a previously conceptualized audit framework (CBEAT); 2) it provides a unique cumulative risk assessment based upon multiple sources of relevant data; and 3) it will provide the Ghanaian government with sound and effective policy recommendations regarding how to govern emotional data in Ghana. This research is timely as Ghana moves towards establishing a national regulatory framework for artificial intelligence through the drafting of a new Data Protection Bill and establishment of a Responsible AI Office; thus, this audit will serve as a significant input for policy development by regulators who will be responsible for crafting legally enforceable standards to govern the use of AI within Ghana.