The Client
Customer: Global developer and manufacturer of advanced materials for coatings, printing & packaging, and industrial manufacturing markets
Size: 300+ employees
Transaction volume: 200/month
The Problem
With each supplier having a different invoice format, invoice processing was tedious, and it was all too easy to make mistakes.
Finance executives had to manually identify Purchase Order (PO) numbers on the invoices, then pore over the quantities and amounts among a jumble of numbers on each invoice, trying to match them with the POs in the ERP system.
Existing technologies like Optical Character Recognition (OCR) rely on templates and does not scale well.
The Solution
CFB Bots designed and implemented a solution that integrated several best-in-class technologies to support straight-through processing of invoices.
The RPA bot handled the entire end-to-end invoice processing, with Machine Learning technology classifying and extracting data. To ensure accuracy, in the rare cases where the extraction by the bot is below a pre-defined confidence level, the case is escalated to human workers for validation (exception handling).
These technologies worked together seamlessly to make this possible:
UiPath Document Understanding
Intelligent invoice extraction: Using Machine Learning models, invoices are extracted quickly and accurately
UiPath Action Center
Human-in-the-loop validation: human workers work alongside the robot, handling exceptions—the rare cases where the confidence level for the bots' extraction is not sufficiently high
UiPath Attended Robots
Once invoices have been extracted and matched with PO information, an attended bot posts them to the ERP system
UiPath AI Center
To get the Machine Learning models to extract invoices as accurately as possible, the models are trained in the UiPath AI Center with documents relevant to the client
