Robotic Process Automation
Automation Anywhere was the RPA tool used to stitch together the Rules Engine, the Imaging and Upload system and the ML Processing and Payments Processing systems for the client.
Two separate open source OCR engines were deployed and the output automatically compared for each invoice scanned; in cases where the engines differed in their output, the invoice and the mismatch was sent to an exception queue for manual arbitration and resolution.
Machine learning models were deployed to learn the invoice templates of multiple suppliers leading to faster OCR of key data points from the invoices. Data points extracted in this step were augmented with metadata from the client’s database to automatically tag and categorize the invoices belonging to different product categories, vendors, amount buckets and time periods enabling better and more intuitive analysis by the client’s management.
Anomaly detection and Error detection algorithms were deployed on top of the OCR solution to derive maximum benefit from automation of the process; potential fraudulent invoices, duplicate invoices and fat-finger errors / typos were reduced, leading to increased confidence in the automated payout solution.