You are opening our English language website. You can keep reading or switch to other languages.

Automating Invoice Processing with AWS Bedrock for Equipment Rental Company​

Company Name

Location

Finland

Client

Renta Group is a full-service machine and equipment rental services company. It receives maintenance, service, and spare parts invoices from various vendors.

Challenge

Employees manually input data from the invoices into the service and maintenance management system, decreasing productivity while increasing error-prone. Reducing manual work also boosts employee satisfaction.

Solution

DataArt’s team built a solution that leverages GenAI capabilities, particularly AWS Bedrock foundation models, to extract data from PDFs containing invoices. This approach automates invoice processing and reduces the time spent on manual effort, transforming unstructured data into a structured format suitable for database storage.

The primary challenges were managing PDFs with invoices from various vendors in different formats and the lack of initial testing data. Proper preprocessing of the PDF data, specifically correct PDF-to-text conversion and table extraction, helped to resolve these issues and achieve high extraction accuracy. While Textract enables data and information extraction from invoices, Bedrock foundation models allow for structuring and reformatting of the data without the need for model retraining.

Additionally, the development team created testing data from scratch. They also chose an LLM (Claude) that performed well in Finnish, the language of all documentation.  

Technologies

Bedrock (Anthropic Claude)
Textract
ECS Fargate
Lambda
DynamoDB
S3

Outcomes

  • 50% reduction in the time spent by employees on manual invoice processing
  • 78% data extraction accuracy across all vendors
  • Enhanced productivity and work satisfaction, along with reduced stress levels
  • Contribution to cost-efficiency within the business
Contact Us
Please provide your contact details, and we will get back to you promptly.