AI Based Detection of Ghost Workers in the Public Sector Payroll System of the Democratic Republic of Congo

Public payroll digitization can improve payment administration while leaving uncertainty about whether recorded recipients occupy authorized positions. This article examines how artificial intelligence could support detection of potential ghost workers in the Democratic Republic of Congo. It critically synthesizes published payroll-governance research, institutional reform accounts, and scholarship on entity resolution, anomaly detection, and accountable machine learning. The analysis identifies a gap between contextual descriptions of payroll fragmentation and locally validated detection using independently established outcomes. It proposes an identity, screening, and verification framework that connects employee records to periodspecific employment entitlement and payment evidence. Rules and record linkage provide the baseline; learned scores prioritize investigations where they add measurable value. Four propositions concern linkage quality, investigative yield, explanation usefulness, and institutional conditions. A prospective design-science pilot would compare methods under common review budgets, use temporal validation, and investigate selected non-alerts to address verification bias. Precision at the review budget, false alerts, review time, and justified case resolution are distinguished from national savings. The article is a structured conceptual review; it reports no new payroll dataset, trained model, or empirical validation. It concludes that a useful AI contribution depends on reliable identity evidence, defensible evaluation, and an accountable correction process that protects legitimate employees.