DSPy a system for AI to Write Prompts and Do Fine Tuning
RPA case study for claims processing
1. Background
The client is a global multiline insurance companywhich has core service offerings inGeneral Insurance, Life andFarmers. It employs more than50,000 people worldwide serving customers inmore than170 countries and territoriesaround
the globe.
One of the keyaspects of customer satisfaction is timelyandaccurate settlementof the claim payments.
The client’s claim payments team forthe US geographyis a 300 memberteamwhichhandles the payment processing forall claim settlements across businesslines.
This particular activityis handledbyanoffshore vendor and the client is lookingto drive processefficiencies throughvarious digital levers.
We have beenentrusteduponbythe client to understandthe process, advise themaboutthe improvementopportunitiesanddraw the actionplanfor execution.
Following are the key asks of the client for the Claim Payments team:
Increaseefficiencyof the process byoptimizingthe total number ofresources
100% accuracyof payments since the error rate of payments has beenaround15% inthe lastone year whichhas ledto lot of customer dissatisfactionandescalations
Improve touch-less processingas muchas possible (ina 24*7 environment)
Complywithindustryregulations
Reduce exceptions
Improve overallscalabilityto other geographies while looking at the growingmarket size
Process Details:
The processstarts withthe upstream claim handling team taking over the First notificationoflossrecordandassigning the evaluator.
Once the Claimhandler gets the required inputs fromevaluator andis able toadjudicate, she arrives at the amount to be paidandher team works onfilling out the detailsona payment request form (PRF) whichis thensent over to the
offshore payments teamfor processing.
The offshore agent goesthrough the PRF andafterdoingthe requireddue diligence, she processesthe request furtherforpayment.
There are 3 paymentruns ina given dayandbased onthe processedrequests the payments goout to the claimants.
The mode inwhichthe agents receive the PRF varies for the type of business line (General liability/ AutoClaim/Workers Compensationetc)andthe infrastructure set upin the respective Claimhandler’s office inthe onshore locations.
Useful Data:
Average Handling time forprocessing one PRF:6.5 mins per transaction
Volumes handled(Monthly)
Workers Compensation Payments:220,000
General LiabilityPayments:115,000
Auto Payments:80,000
SpecialityClaim Payments:38,000
Other Claims:65,000
Percentage of Exceptions:18%
ErrorPercentage:15%
Compliance toRegulatoryPractices:95% complied(Opportunityto improve upto100% whichis the desiredlevel)