Inconsistent data
Claim amounts, receipt amounts, dates, vendor names, tax information, project codes, and employee details were often entered differently across documents.
Claims Reconciliation Case Study
Turning messy staff claims into payroll-ready reports.
Scellus helped a growing services company reduce claims processing time by 67% with document reconciliation, exception handling, and payroll-ready exports.
Regional services company · Malaysia · 8 weeks
67%Processing time reductionMonthly claims work dropped from around 160 hours to 52 hours.
2-3 daysClaims cycleThe cycle shortened from 10-14 days after implementation.
108hSaved monthlyHR recovered more than one hundred hours each month.
60%+Fewer follow-upsManual follow-ups fell as missing documents were flagged earlier.
0.5 dayPayroll preparationPayroll file preparation dropped from 2-3 working days.
Executive Summary
A faster, cleaner claims workflow for HR, Finance, and Payroll.
A mid-sized company with around 500 employees was struggling with a slow and messy staff claims process.
Employees submitted receipts, invoices, approval screenshots, payment proofs, travel records, and claim forms in different formats. HR had to manually check every file, reconcile amounts, chase missing documents, identify duplicates, and clean the final spreadsheet before payroll import.
Scellus implemented a claims reconciliation system that allows users to upload all expense documents into a shared folder. The system extracts key information, links related documents together, checks each claim against company rules, routes exceptions to a human reviewer, and generates a payroll-ready export.
Within the first three months, the client reduced monthly HR processing time from around 160 hours to 52 hours, shortened the claims cycle from 10-14 days to 2-3 days, and reduced manual follow-ups by more than 60%.
The Challenge
The staff claims process had grown beyond what spreadsheets and manual checking could handle.
Staff submitted claims through different channels. Some documents came through email. Some were uploaded to folders. Others were shared manually with HR or department admins.
The HR and Finance teams had to piece everything together before each payroll cycle. A single claim could involve a claim form, receipt, approval screenshot, payment proof, and corporate card line item, all with slightly different information.
Even after claims were approved, HR still had to clean, reformat, and prepare the final file for payroll import. This created a second bottleneck after the claim review process.
- Claim amounts, receipt amounts, dates, vendor names, tax information, project codes, and employee details were entered inconsistently.
- Required supporting documents were often missing until month-end checks.
- Finance struggled to trace each claim back to the employee, department, approval, payment reference, and payroll cycle.
- Duplicate submissions were difficult to detect when claims crossed different months or projects.
- Payroll import files still needed manual cleanup after review.
Key Issues
The operational risk was spread across intake, reconciliation, approval, and payroll preparation.
Incomplete documentation
Claims were frequently missing receipts, approval emails, payment proof, project codes, or tax invoices. HR often found these gaps only during month-end checks.
Untraceable transactions
Finance had difficulty linking each claim to the correct employee, department, approval, payment reference, and payroll cycle without digging through folders, emails, spreadsheets, and message threads.
Duplicate submissions
Some employees accidentally submitted the same receipt more than once, especially when claims crossed different months or projects.
Manual payroll preparation
After approval, HR still had to clean and reformat spreadsheets before payroll import, creating a second bottleneck.
The Goal
Turn messy expense documents into clean, structured, payroll-ready claim records.
The goal was not to remove HR from the process.
The goal was to remove repetitive checking, reduce avoidable errors, and let HR focus only on claims that required human judgment.
- Reduce repetitive document checking.
- Identify missing, duplicated, unclear, or out-of-policy claims earlier.
- Keep HR in control of exceptions and judgment calls.
- Generate an export that payroll could import directly.
The Solution
A folder-based claims reconciliation system with human exception review.
Scellus built a folder-based claims reconciliation system that processes expense documents automatically and flags exceptions for human review. Users upload claim-related documents into a designated folder, the system reads and links the files, validates each claim against policy and payroll rules, marks clean claims as payroll-ready, routes exceptions to HR, and exports approved claims in the required payroll format.
How It Works
The workflow moved from manual checking to structured reconciliation.
The system was designed around a simple operating model that preserved a familiar folder-based intake while adding extraction, linking, validation, review, and payroll export behind the scenes.
- 01
Folder-based intake
Employees and department admins continued uploading claim documents into a designated secure folder. The system monitored the folder and processed receipts, tax invoices, claim forms, approval screenshots, payment proofs, corporate card statements, travel itineraries, mileage claims, meal claims, vendor invoices, PDFs, images, and spreadsheets.
- 02
Document extraction
The system extracted employee name, employee ID, department, claim category, claim amount, receipt amount, transaction date, vendor name, tax invoice number, approval reference, payment reference, project code, cost centre, payroll cycle, and supporting document type.
- 03
Document linking and reconciliation
Related files were linked into a single claim package. Receipts were matched to claim forms, claim forms to employee records, payment proofs to transactions, corporate card lines to submitted receipts, and approval screenshots to the relevant claims.
- 04
Rule-based validation
Each claim was checked against company policy and payroll requirements, including amount matching, allowed claim period, required documents, duplicate receipts, employee record matching, approval thresholds, project codes, cost centres, payroll cycle, payroll code, and Finance review requirements.
- 05
Human-in-the-loop exception review
The system did not blindly approve claims. Missing receipts, amount mismatches, possible duplicates, unclear names, missing approvals, unreadable documents, missing project codes, missing payment proofs, and payroll mapping issues were routed to HR for review.
- 06
Payroll-ready export
After validation and review, the system generated a clean payroll file with employee ID, employee name, department, claim category, approved amount, payroll code, cost centre, payroll cycle, review status, supporting document status, reference number, and remarks.
Exception Handling
HR moved from checking every line to reviewing the claims that needed attention.
The system was designed to keep judgment with the HR and Finance teams while removing the repetitive checking work around complete, policy-compliant claims.
- 01
Approve
Reviewers could approve a flagged claim once the supporting documents and extracted data were verified.
- 02
Reject
Claims outside policy, unsupported by evidence, or confirmed as duplicates could be rejected before payroll export.
- 03
Correct
HR could correct extracted fields such as project code, cost centre, payroll code, category, amount, or employee mapping.
- 04
Request clarification
When a claim needed more context, HR could route it back for missing receipts, approval proof, payment evidence, or other supporting information.
03 — System
Implementation
The implementation connected folder monitoring, document extraction, reconciliation rules, employee records, exception review, audit history, and payroll export into one claims workflow.
Folder intake, extraction, and linking
Exception review and payroll export
The system focused on the parts of claims processing that created the most month-end drag: reading supporting files, matching related documents, detecting missing information, identifying possible duplicates, and preparing a clean payroll import file.
Clean claims moved through automatically. Claims with missing, unclear, duplicated, or out-of-policy information were routed to a human reviewer with the reason for review attached.
Technologies
- Secure folder intake
- Document extraction
- Claims reconciliation
- Rule-based validation
- Exception queue
- Audit trail
- Payroll export
Results
Within three months, the claims cycle dropped from 10-14 days to 2-3 days.
The client saw measurable improvements in processing speed, HR workload, payroll preparation, and claim traceability.
- 01
Faster month-end processing
HR and Finance no longer had to spend days manually checking receipts, claim forms, approvals, and payroll codes.
- 02
Better traceability
Each claim had a linked record showing supporting documents, extracted data, validation status, reviewer action, and payroll export reference.
- 03
Fewer payroll errors
Because the system exported claims in the required payroll format, HR spent less time cleaning spreadsheets and fixing import issues.
- 04
Lower duplicate-claim risk
Duplicate receipts and repeated submissions were flagged before payroll export, reducing the risk of accidental overpayment.
- 05
Easier staff adoption
The folder-based intake meant staff and department admins could continue uploading documents while the system handled reconciliation in the background.
- 06
Exception-based review
The team shifted from checking every claim manually to reviewing only the claims that needed human attention.
Measured improvement
| Metric | Before | After | Impact |
|---|---|---|---|
| Claims processing cycle | 10-14 days | 2-3 days | 67% to 80% faster |
| HR time spent on claims | Around 160 hours / month | Around 52 hours / month | 108 hours saved monthly |
| Manual follow-ups | 38% of claims | 14% of claims | 63% reduction |
| Payroll preparation time | 2-3 working days | Half a day | Up to 80% faster |
| Duplicate or questionable claims | Often found late | Flagged before payroll export | Better leakage control |
| Payroll import formatting issues | Around 12% of records needed cleanup | Under 1% needed cleanup | Cleaner payroll submission |
What Made This Work
This was not just OCR. The value came from reconciliation.
The system connected receipts, claim forms, approvals, payment proofs, employee records, cost centres, and payroll codes into a single traceable workflow.
That allowed HR and Finance to answer the questions that actually matter before payroll export.
- Is this claim complete?
- Does the receipt match the claimed amount?
- Has this receipt been submitted before?
- Is the claim approved?
- Is the employee record correct?
- Is the payroll code correct?
- Is this ready for payroll?
- If not, what exactly needs to be fixed?
The workflow changed from manual document checking into exception-based review.
In their words
What the team noticed.
-
Before this, claims processing took up a huge part of our month-end routine. We were checking receipts, chasing missing documents, and cleaning spreadsheets before payroll. The new system gives us a clear view of each claim and only highlights what needs attention. Our team now reviews exceptions instead of checking every line manually.
-
The payroll export made the biggest difference for us. Previously, we still had to reformat approved claims before importing them into payroll. Now the report is already structured properly, so we spend less time fixing files and more time reviewing the actual exceptions.
-
The traceability is what made the system useful for Finance. When someone asks why a claim was approved, we can see the receipt, approval, payment reference, and claim record together. We no longer need to dig through folders and emails.
Need to turn messy documents into payroll-ready records?
Scellus builds workflow automation systems that fit existing operations while removing repetitive manual checking.