Summary
The client processes a high volume of insurance claims arriving by email. Claim validation and payment processing were manual, producing long processing times and a higher risk of error.
We automated intake, eligibility verification and payment on UiPath. The robot retrieves claim forms from Lotus Notes, extracts data from the PDF documents, validates eligibility in a mainframe application and processes payment for eligible claims. Processing time fell from 55 minutes to 15 minutes per transaction, with accuracy at 97%.
The client
The client operates in the insurance sector across multiple products with strict compliance and accuracy requirements. Claims processing is a critical business operation owned by the claims and operations teams. Customers submit claims by email in PDF form, and each requires detailed validation against policy and eligibility rules held in a mainframe system.
The problem
The claims process ran end to end on manual effort.
- 01Monitor Lotus Notes for incoming claim forms
- 02Download and review PDF claim documents
- 03Extract claim and policy information by hand
- 04Log into the mainframe to verify eligibility
- 05Process payment for eligible claims, or route the rest for review
What it was costing
- High manual effort and 55 minutes of handling time per claim
- Risk of human error in extraction and eligibility checks
- Inconsistent decision-making leading to rework
- Delays in settlement affecting customer satisfaction
Objectives
Goals
- Reduce end-to-end claim processing time
- Improve accuracy and consistency in claim validation
- Automate eligibility checks in the mainframe application
- Ensure timely payment processing for eligible claims
- Handle business exceptions efficiently rather than manually
Success criteria
- Processing time per claim
- Accuracy percentage on eligibility decisions
- Reduction in manual effort
What we built
A single end-to-end process with a hard split between what the robot decides and what a person decides. Eligible claims complete automatically; anything failing a rule is flagged with a clear status for business review.
Lotus Notes monitoring
The robot watches the claims inbox and downloads PDF attachments as they arrive.
Claim data extraction
Required claim and policy details are extracted from the PDF documents through document processing.
Mainframe eligibility check
Eligibility is validated in the mainframe against predefined policy rules — no interpretation, no judgement calls.
Automated payment release
Payment is processed automatically for claims that pass every rule.
Exception flagging
Ineligible claims are flagged with clear status updates and routed to the business team for handling.
Delivery
Phased delivery — assessment through stabilisation
- Process assessment and requirement gathering
- Document analysis and extraction logic design
- Mainframe interaction and eligibility rules
- Bot development and unit testing
- UAT with the claims team
- Production deployment and stabilisation
Challenges and how we handled them
Stakeholders
Results
| Measure | Before | After |
|---|---|---|
| Processing time per claim | 55 minutes | 15 minutes |
| Accuracy | Variable, error-prone | 97% |
| Eligibility check | Manual mainframe lookup | Automated against rules |
| Ineligible claims | Identified late | Flagged immediately with status |
| Audit logging | Partial | Detailed logs per claim |
Quantitative
- End-to-end processing time reduced from 55 minutes to 15 minutes per claim
- Accuracy improved to 97% on eligibility validation and payment processing
- Eligible claims processed without manual touch
- Ineligible claims automatically flagged for business review
Qualitative
- Teams focus on exceptions rather than routine claims
- Consistent decision-making across claim types
- Improved customer experience through faster settlement
Business value delivered
Faster settlement
A critical insurance process modernised end to end, with settlement times cut by nearly three quarters.
Compliance confidence
Eligibility rules applied identically every time, with the logs to prove it.
A foundation to extend
The pattern now supports further automation across the wider claims lifecycle.