The customer designs, develops, manufactures and markets a line of digital and mixed-signal integrated circuits. These circuits are used in the data communications, telecommunications, computers, and instrumentation systems industries. The company’s commitment is to provide customers with a high-quality, value-driven embedded solution, on-time, with world-class support.
While all invoices include similar fields, they’re sent to accounts payable operations in hundreds, sometimes thousands, of different formats from different suppliers. Considerable time is spent extracting data from each invoice, and the process of entering it into their Oracle ERP platform is manual, repetitive, and prone to errors. The variability of the formats makes it very difficult for their existing OCR platform to capture data due to the lack of consistency in invoice formats.
Cognitive Machine Reading (CMR), a combination of AI, cognitive technology, Machine Learning and pattern recognition, captures invoices from a shared folder. Its extracts key data elements and automatically categorises invoices in a way that emulates human data processing. CMR has built-in capabilities such as Natural Language Modelling, Data Contextualisation, and Data Enrichment providing curated data for downstream systems. CMR automatically flags missing or unreadable information as exceptions and routes tasks to the right people. Each manual intervention improves the accuracy and confidence of CMR’s Machine Learning models and multiple reports are available for further analysis.
After CMR’s implementation, the customer saw a 79% conversion accuracy rate, resulting in fewer hours spent handling exceptions and ensuring correct payments were made. CMR reduces overall invoice processing time by over 4,000 hours annually resulting in an increase in business processing efficiency