Hands-Free Processing of Inbound Invoices

Most manufacturing companies need to procure materials and services from hundreds of external vendors. These invoices come in numerous layouts, and entering details from these to the ERP is often a nightmare for companies.
Some outsource the invoice data entry job to 3rd parties. Others utilize their own internal resources for the purpose. Either way, this is a time consuming and effort-intensive activity.

With the introduction of electronic invoices (e-Invoices), the Government of India standardized the format for Business to Business (B2B) Invoices and made this mandatory for companies with an annual turnover larger than 100 Crores (lowered to 50 Crores with effect from 01 Apr 2021). With this, the manual data entry for inbound invoices can be totally eliminated. But for this, they need a software that can seamlessly receive the signed e-Invoice acknowledgements from the vendors, and pull the validated line items for these from the GSTN portal.

A few solutions are available in the market that can semi-automate the processing of inbound invoices. These use technologies like Optical Character Recognition (OCR) and Machine Learning (ML) to extract the relevant fields from inbound invoices in PDF or image formats. However, these solutions are not widely implemented in the industry, probably due to unacceptable level of errors in the extraction process.

It is in the above context that IVL took on the challenge. We went on to design a solution that surmounts the technological hurdles of extracting field values from invoices and other external documents. The terms of reference used while planning the product were as under

• Not prohibitively costly to implement and run
• The same solution must handle both electronic and PDF invoices
• Near 100% extraction accuracy for e-Invoices and for PDF invoices (this latter needs to be “trained” for each vendor layout)
• Friendly User Interface for training new vendor invoices that can be used even by non-technical users
• 30% to 60% extraction accuracy for PDF Invoices of unfamiliar (“untrained”) layouts, thus reducing the manual effort by limiting this to train only the remaining fields with errors

What is more, in case the ERP used is SAP, IVL offers a SAP Certified Add-on product that PARK/ POST the extracted fields of inbound invoices at the ERP