Invoice Document Processing
Client: Multi National Manufacturing Enterprise
Business Problem: Client operates in many Eastern European and erstwhile Soviet Union countries. They use large number of local vendors for raw material, spare parts and other consumables purchases. These vendors do not have proper IT systems and they send their Accounts payable Invoices in PDF files, Scanned files, by Fax and physical mail. Every vendor had their own different format for the invoice. The Accounts payable department has to employ large staff to manually look at these documents and verify, enter these invoices in SAP system and clear the payment.
Data Challenges: Many scanned, Faxed and documents sent in mail were smudged, not printed properly and need lots of guessing.
AI Solution:
- We built a joint team of customers Finance team, our domain experts and AI tech team to list key data fields required to populate SAP Invoice transactions. For example : Vendor number, Vendor Name, PO#, Item#, Item quantity, shipped date, received date, other and the invoice amount and invoice currency etc.
- We batched similar looking invoice formats in to batches and trained the AI Engine to recognize these key fields as one training was not workable with such varied formats from each vendor. Lots of iterative process was followed to keep improving the accuracy of recognition and capture of the key fields.
- Wherever possible we used the Core SAP systems to fill in smudged letters or faint letters to fill up the blanks.
- The AI engine was integrated with SAP and the Invoice screens were pre-filled with maximum number of fields automatically recognized by the AI Engine. The Billing agent has to fill up only few fields and complete the SAP Transaction.
Solution Benefits:
- 30% staff reduction in the Back office head count.
Note: These case studies are experiences and accomplishments by our Think-Tank members