AWS Case Study: Bahrain Public Sector Organization & Gen AI Document Handling

Project background & customer issues

The customer’s process involves reimbursing employers only after confirming that employees have been paid. This validation was undertaken manually, requiring officers to cross-check employee names, CPR numbers (personal identification number for Project background & customer issuesboth citizens and residents), payment amounts, and pay periods between uploaded payslips and corresponding bank statements.

This manual process presents several critical business challenges,

• Operational inefficiency: each request involves time-consuming, repetitive data extraction and comparison across multiple documents for many employees, leading to processing delays and high administrative overhead.
• Human error risk: manual verification increases the likelihood of inconsistent validation and data entry errors, which can result in incorrect reimbursements or delayed employer payments.
• Limited scalability: as the number of participating employers and employees grows, the manual validation process cannot scale effectively, causing potential backlogs and bottlenecks.
• Compliance and auditability risks: errors or delays in verification can expose the customer to financial mismanagement risks, compliance gaps, and audit challenges, undermining trust in the reimbursement program.

Risks of Not Addressing the Challenge

If not addressed, these issues would have measurable business impacts,

• Longer processing cycles, reducing the organization’s ability to meet service-level agreements with employers.
• Increased operational costs, as more staff would be required to handle growing workloads manually.
• Higher risk of inaccurate disbursements, leading to financial losses.

Project objectives

The customer’s primary objective in the project is to leverage AWS AI and automation capabilities to enhance the accuracy, efficiency, and scalability of their reimbursement validation process. The project aims to transform a manual, document-intensive workflow into an intelligent, automated system that ensures data integrity and operational transparency.

This change in processes will,

• Validate that the uploaded documents are the intended documents (Pay slips & bank statement) and output the confidence level for each document.
• Fields to be extracted and matched (Names, CPR, Amounts -base, month & years).
• Showing the data loaded vs the previous months to help detecting any anomalies.

Category Alignment: Generative AI Consulting Services

This case study aligns with the AWS’s Generative AI Consulting Services specialization.

The customer manages over 200,000 business transactions annually with employers, each submitting around 6 documents monthly—amounting to nearly 200,000 pages processed through OCR every month.

The implemented solution leverages Amazon Textract to automatically extract key data fields such as names, CPRs, payment amounts, and pay periods. Extracted data is securely stored in Amazon DynamoDB, with lifecycle policies to transition documents to S3 Glacier storage every 3 months for cost optimization.

For advanced insights, Amazon Bedrock will allow officers to retrieve and summarize information instantly. This end-to-end AI automation reduced manual processing time by over 80%, improved accuracy, and provided scalable infrastructure for future AI-driven compliance and reporting use cases.

AWS Services

This project used the following AWS services,

• Networking
  • AWS Site-to-Site VPN
• Storage
  • Amazon S3
  • Amazon ECR
• Compute
  • AWS Lambda
  • Amazon ECS
  • AWS Fargate
• AI & ML
  • Amazon Bedrock (Flows, Agents, Models)
• Databases
  • Amazon Aurora (PostgreSQL)
• Integration
  • Amazon API Gateway