Research on Intelligent Audit System for Employee Expense Reimbursement Based on Agent Technology

Abstract

Expense reimbursement auditing is a high-frequency task in corporate finance departments, requiring auditors to verify invoice authenticity, expense compliance, and approval process integrity across multiple dimensions. This work is repetitive and prone to errors due to auditor fatigue. This paper proposes an intelligent expense reimbursement audit assistant based on agent technology to automate preliminary auditing tasks. The intelligent agent integrates three core functional modules: invoice verification, standard compliance checking, and quota verification. It automatically identifies invoice information from reimbursement documents and validates authenticity through tax authority interfaces, cross-checks expenses such as meals and travel against corporate standards to detect overages, and verifies whether cumulative reimbursement amounts exceed departmental budgets. The system employs a rule engine architecture rather than complex machine learning algorithms, enabling finance personnel to maintain audit rules independently without continuous IT support. In practical application at a consulting firm, the system reduced document auditing time from an average of 15 minutes to 5 minutes per claim, improved anomaly detection rate by 40%, and significantly reduced manual audit costs and compliance risks. This paper elaborates on key implementation aspects including audit rule base construction methods, common anomaly type identification logic, and human-machine collaborative audit process design, providing enterprises with a low-barrier, high-efficiency solution for financial digital transformation.