AI-Based Intelligent Invoice Processing System for Auditing and Compliance

The AI-based intelligent invoice processing system is an automatic auditing system. Its main purpose is to detect fraud and make invoice processing simpler using local computers. According to global financial data, manual auditing costs billions yearly due to human error and missed anomalies. Although large cloud-based models like GPT-4 can quickly process complex datasets, they still create huge data-control and compliance issues for banks' sensitive files. To avoid this critical risk, this proposed system runs entirely offline. The system combines a Generative Model (Llama 3.2 3B), Optical Character Recognition (OCR), and a thoroughly defined Heuristic Risk Scoring Engine to audit financial documents onpremises. Our technical innovation “Adaptive Semantic Compression”, allows the model to fit very large, multicompany datasets into its context window.