Manufacturing

Smart Invoice Process

Business Challenge

Low Satisfaction with RPA-Based System

The customer was not satisfied with the success and performance of the document digitization solution that was currently dependent on RPA robots. There were problems due to unused pages not being carried over to the following years and digitized documents not being correctly matched with the data.

  • Language Support –
    In particular, invoices sent by suppliers and issued in different languages such as Chinese, German and Polish could not be digitized successfully.
  • Robot Dependent Work –
    The processes of interpreting the data in the digitized documents and running the business rules were based on the decisions of the robots. This structure was inflexible, prone to errors and caused loss of time.
  • Supplier Invoice Variations –
    Since each supplier sent invoices in different formats, it was quite difficult to determine and interpret a common structure.
  • High Operational Load
    These manual verifications created a significant workload for insurance personnel, increasing the potential for human error and causing frequent delays in accident insurance processing.
  • Employee Satisfaction
    Long turnaround times for approval or denial negatively impact employee morale and confidence in the reimbursement process, while also increasing compliance risks for the organization.

Manufacturing / Smart Invoice Process

Agentic AI - Smart Invoice Process

API Integration

Requests are sent to GDX Services

Customer's existing RPA processes trigger AI-powered GDX services via API. End-to-end automated solution. Fast response time (2-3 seconds).

Document Submission and Operations

| GDX

Upload Document, GetStatus, GetResult

UploadDocument: Submits documents to the AI system for processing. GetStatus: Returns the current status of the submitted document. GetResult: Retrieve extraction results after document is processed or manually validated.

Document Understanding Process

| GDX

Extracts Required Fields from attachments Like Amount, Date

SendHumanValidation: Sends the document to a human validator based on business rules; validation messages can be attached. DeleteDocument: Deletes all data related to the submitted document.

MultiPage & MultiInvoice

Supports various multi-page, multi-invoice documents and extracts data accurately

MultiPage & MultiInvoice Support: Handles various multi-page, multi-invoice documents and extracts data accurately from complex document structures.

Multilingual & Handwriting

Recognizes various languages like Chinese, Japanese, German, English and handwritten text

Multilingual & Handwriting Recognition: Supports various languages like Chinese, Japanese, German, English and handwritten text. Fully AI-driven classification and data extraction process.

Data Logging and Monitoring

Extracted data is stored in a database and key fields

Extracted data is stored in a database and key fields. This data can be retrieved for reporting when needed. Monitored daily via ServiceNow dashboards or custom reporting tools.

????
Application Processing

How Agentic AI Helped

  • Custom Invoice Model Integration
    Unite GDX’s invoice model was customized to customer needs and integrated with Agentic AI support.
  • API-Based Access
    The customized model was made available to the customer via API service for seamless integration.
  • Robot-Free Architecture
    This new model, which is not dependent on RPA robots, provided high success rates in very short periods of time.
  • Automated Document Processing
    96% of the documents were successfully processed without any human intervention or quality control.
  • AI-Driven Satisfaction Recovery
    With Unite GDX and Agentic AI-supported technology, customer dissatisfaction was largely eliminated.

Results

  • Multi-Supplier Format Digitization
    Successfully digitized invoice formats from multiple suppliers with Agentic AI support.
  • Multilingual Invoice Interpretation
    Supplier names, PO numbers, and shipment info from invoices in Chinese, German, and Polish were accurately digitized.
  • High Volume Throughput
    Many more documents were successfully processed in a very short time.
  • Minimal Human Intervention
    Human control and intervention were significantly minimized.
  • Partial & Multi-Invoice Handling
    Incomplete invoices and multi-invoice documents were successfully detected and processed with near-perfect accuracy.
  • Email Body Invoice Recognition
    Invoice data written in the body of emails was correctly identified with 96% accuracy.
  • Reduced Processing Time
    Average processing time dropped from 4 minutes to 10 seconds.
  • 12,000+ Hours Automated
    Over 12,000 hours of automation achieved, increasing overall workforce efficiency.