新加坡国家监管 利用人工智能打击汽车索赔欺诈.pdf

1、Confidential.Fermion 2024.All rights reserved.Combating Motor Claims Fraud with AI&a Relentless Focus on DATASebastian Tan,Singapore Country Director9,000+Franchise Dealers&Repairers1,000+Loss Adjusters,Lawyers24,000+Agents,BrokersAbout MerimenRegional PartnersCustomer BaseOver 20 Years of Expertise

2、 across 10 countriesAsias leading Claims Technology provider150+InsurersDigitize&Automate 2 Million claims a yearProblem:Fraudulent Motor Claims costing losses&livesSource:General Insurance Association of Singapore https:/gia.org.sg/images/resources/For-Consumers-PDF-file/CASE-2018-motorfraud.pdfAlm

3、ost 20%of motor claims in SG are fraudulentSeverely affects loss ratios,contributes to rising premiumsStaged accidents harm lives-drivers/passengers/pedestriansCrash Course on Motor Claim FraudAccident is fakeAccident is real,but Damage is fake Accident is real,damage is real,but Passengers are fake

4、 Accident,damage,passengers are real,but Claimant is fake Solution:How Tech&AI can help combat FraudAs-IsTo-Be1.Automated Checks2.Assisted by AI3.Machine-readable Data databases,doc metadata,Internet,IoT,etc1.Manual Checks2.Staff Expertise3.Human-readable Data documents,interviews,Internet searchSus

5、picious ClaimsSend to SIUAll Incoming CasesFirst Challenge:Your Process&Data looks like thisInsurersRepairersCustomersLoss AdjustersPRIVACYPROFITSREPLACED BY AIALL OF THE ABOVESecond Challenge:No One will share Data(and neither will you)LawyersCase Study:Failed Pilots for AI Fraud in SE Asia2017 202

6、4:Over 10 failed insurer pilots by leading AI providers from US/EuropeKey Reasons reported by users:1.AI only ran basic checks on limited data,no Industry Data checks or Image analysis etc2.Overly high False Positive rate due to inaccurate/un-localized AI model3.Standalone portal not integrated to c