Invoice-to-ERP processing for a finance team

Client context

A distribution company receives several hundred supplier invoices a month by email, as PDFs and scans. A three-person accounts payable team enters them into SAP.

Problem

Each invoice is keyed in by hand and then checked against its purchase order and goods receipt. Mismatches are often found only at payment time. Near month-end the backlog grows, and the close is delayed while the team catches up.

Approach

1. Sample the documents. Take a few hundred recent invoices and group them by supplier and layout, to see how much variety there really is.
2. Measure extraction on real invoices. Test how reliably each field can be read before committing to a design or a price.
3. Keep the controls. Three-way matching and approval limits stay exactly as they are. Only the typing is removed.
4. Send uncertainty to people. Any field read with low confidence, and any mismatch, goes to a review queue.

Solution

• Invoices sent to a shared mailbox are saved to SharePoint automatically.
• Supplier, invoice number, dates, tax and line items are extracted.
• The invoice is checked for duplicates, then matched against the purchase order and goods receipt.
• Clean invoices are posted to SAP with the PDF attached.
• Exceptions appear in a review queue that shows the document next to the extracted fields and the reason it was held.
• A Power BI report shows volumes, exception reasons and processing time.

Measurable results

Every project starts by recording a baseline, and results are reported against it. For this kind of project we measure:

• Handling time per invoice
• Share of invoices posted without manual entry
• Exceptions by reason, to target supplier or process fixes
• Days to close accounts payable at month-end

Tools used

Power Automate, SharePoint, an AI document extraction service, SAP and Power BI.