Auditors increasingly have access to complete popu-lations of payroll records, card transactions, and financial state-ments, but black-box anomaly scores are difficult to defend in audit documentation and can displace, rather than inform, pro-fessional judgment. This study develops and evaluates XAT, an explainable audit toolkit that combines transparent audit rules with peer-group statistical baselines, so that every flagged item comes with a reason an auditor can verify. XAT is applied to three real U.S. public datasets: 544,338 New York City payroll records (FY2024), 109,011 Oklahoma state purchase-card trans-actions (April–June 2024), and 35,166 public-company firm-years labeled with SEC enforcement actions. In payroll and expense data, where no fraud labels exist, we seed 2,098 realistic anoma-lies (padded overtime, salary overpayment, ghost employees, un-supported payments, split purchases, duplicates, personal-use pur-chases, and inflated amounts) into the real records. Within a review budget of 1% of records, XAT surfaces 31.1% of seeded payroll anomalies and 26.0% of seeded card anomalies, versus 8.0% and 9.7% for an isolation forest and 13.8% and 13.6% for rules alone. For revenue recognition, a sparse, glass-box logis-tic model of revenue red flags achieves an out-of-sample AUC of 0.647 against SEC-identified fraud, close to a black-box gradient-boosting model (0.671) and well above a Beneish-style revenue in-dex (0.553). The results show that explainable tools can match or exceed opaque detectors while producing evidence that supports, rather than replaces, auditor judgment.