AI-Powered Invoice Processing and PO Automation

CUSTOMER

A multi-location pharmacy retail organization operating approximately 50 stores was processing a high volume of supplier invoices every day.

The organization worked with hundreds of suppliers, each using different invoice formats. Invoices arrived in multiple formats, including paper copies, PDFs, and carbon-copy documents.

The existing process required significant manual effort from store receiving teams and the centralized PO department.

CHALLENGE

The organization was processing approximately 500+ invoices per day from around 700 suppliers.

Key challenges included manual invoice reviews, scanning received products, manual reverse PO creation in Microsoft Dynamics 365 Business Central after merchandise arrived, comparing the PO against supplier invoices to identify discrepancies, pricing/tax adjustments, and supplier claims processing.

The organization had approximately eight employees working on PO processing, yet invoices frequently accumulated in a backlog. Large supplier invoices of 10 to 14 pages made manual line-by-line reconciliation extremely time-consuming. Store receiving teams also annotated exceptions (“Not Scanned” items) by hand on paper invoices, which needed manual routing and processing.

COGNIC’S SOLUTION

Cognic proposed an AI-powered invoice processing platform combining Azure OCR, OpenAI API, intelligent document processing, business rules, and ERP integration.

The system automates repetitive invoice processing while keeping human review at critical steps for operational verification.

Step 1: Invoice Receipt & Upload

The store receives the physical supplier invoice together with the merchandise. For the initial POC, the paper invoice is scanned and uploaded into the Cognic platform.

Step 2: Azure OCR Processing

Azure OCR processes the invoice to extract vendor info, invoice number, dates, item codes, product descriptions, quantities, unit prices, line amounts, tax details, totals, and handwritten annotations where detectable.

Step 3: AI Validation

OpenAI processes the extracted information to normalize invoice fields, interpret layouts, validate data, identify potential discrepancies, interpret handwritten exception notes, and match lines with item catalogs.

4. Side-by-Side Invoice Review

The user receives a side-by-side review interface showing the original invoice PDF/image on the left and structured OCR data on the right. Users can verify and edit extracted info before final approval.

5. Exception Detection & Claim Processing

The platform automatically flags mismatch exceptions (items not scanned, missing items, quantity mismatches, price mismatches). Reviewers can select items for supplier claims, and the system automatically calculates the claim values and resulting payable amounts.

6. Contract Price Comparison

The platform compares invoice pricing with contracted vendor pricing stored in the ERP, automatically selecting the lower price according to the organization’s business rules and flagging exceptions.

7. PO Creation & Business Central Integration

After validation, the system prepares the PO using the validated invoice data. The production architecture supports integration with Microsoft Dynamics 365 Business Central, handling PO header/line creation, vendor details, final pricing, and status synchronization.

POC APPROACH

Cognic proposed starting with a controlled POC focusing on selected high-volume suppliers first to validate OCR accuracy, invoice-to-PO matching, pricing comparison, handwritten note detection, and user workflows before a broader deployment.

TECHNOLOGY ARCHITECTURE

• React for the side-by-side invoice review interface
• Azure OCR for document text extraction
• OpenAI API for AI-based validation and handwritten note interpretation
• Backend application for workflow orchestration and business rules validation
• Database for invoice, vendor, pricing, exception, and audit information
• Microsoft Dynamics 365 Business Central for ERP integration

BUSINESS BENEFITS

• Reduced manual invoice processing and faster PO turnaround.
• Reduced invoice backlog and faster identification of missing items.
• Faster supplier claim generation to recover short-shipment credits.
• Reduced manual price reconciliation by matching contracted pricing.
• Better visibility into receiving exceptions and audit tracking.
• Scalable process to onboard hundreds of suppliers.

FUTURE ROADMAP

• Phase 1: Scan paper invoices, OCR extraction, AI validation, and manual review.
• Phase 2: Automated PO creation in Microsoft Dynamics 365 Business Central.
• Phase 3: Digital invoice processing directly from email, EDI, and supplier feeds.
• Phase 4: Fully paperless store workflow using mobile/tablet receiving apps.