ManufacturingAIIntegrationAutomationChina

Huadong Medicine Group ERP + AI Intelligent Data Entry Integration Project

Project Background

As a leading pharmaceutical manufacturer in China, Huadong Medicine Group handles annual procurement exceeding RMB 5 billion, generating a massive volume of daily purchase documents. However, all procurement data previously relied on manual transcription from PDFs and images into the SAP ERP system, followed by approval workflows initiated via DingTalk. The lack of data integration between the two systems caused chronic issues such as redundant data entry, information lag, and approval delays, significantly compromising procurement efficiency and supply chain responsiveness.

Core Pain Points

Low manual entry efficiency: Each purchase order took 2 hours to enter manually; even with 8 procurement specialists, the workload was unsustainable.

High data error rate: Manual entry errors reached 8%, leading to frequent returns and reconciliation disputes.

System data silos: No data connection between SAP ERP and the DingTalk approval system forced duplicate entry into both platforms.

Slow approval workflows: From data entry to final approval, the average turnaround was 3 business days, hampering supply chain response.

Solution

AI‑Powered Intelligent Recognition and Automatic Entry

We deployed a dual‑engine (OCR + LLM) purchase order recognition system supporting intelligent parsing of PDFs, images, and scanned documents. The system automatically extracts key fields—supplier information, material codes, quantities, prices—and maps them to SAP fields according to predefined rules. Entry time dropped from 2 hours to 5 minutes per order.

ERP–DingTalk Data Integration

A standard API integration middleware was developed to enable bidirectional syncing between SAP ERP and DingTalk. After data is entered in ERP, approval workflows are automatically triggered in DingTalk, and approval results are written back to ERP in real time. This eliminates duplicate entry and ensures full traceability of the approval process.

Intelligent Anomaly Alerts

A procurement anomaly detection model, built on historical data, automatically identifies and flags risks such as price deviations, quantity anomalies, and duplicate orders, empowering procurement managers to make rapid, informed decisions.

Results

Metric	Before	After	Improvement
Order entry time	2 hrs/order	5 mins/order	96%
Entry error rate	8%	0.5%	94%
Procurement team size	8 people	3 people	63%
Approval turnaround	3 business days	0.5 business days	83%