Adanced analtics and innoation in inancial rime omliance...helped to lessen the impact of COVID-19...

Advanced analytics and innovation in Financial Crime Compliance The future is now

Advanced analytics and innovation in Financial Crime Compliance | The future is now

03


02

Glossary of terms Foreword

AI ArtificialIntelligence

AML Anti-MoneyLaundering

BAU Business as Usual

CFT CounteringtheFinancingofTerrorism

DMODataManagementOffice

FCC Financial Crime Compliance

FEAT Fairness, Ethics, Accountability and Transparency

FI Financial Institution

GC Group Compliance

MAS MonetaryAuthorityofSingapore

ML MachineLearning

NLP NaturalLanguageProcessing

NS NameScreening

POC Proof of Concept

RPA Robotics Process Automation

TM TransactionMonitoring This third white paper, co-published by Deloitte and UOB, examines the use of innovation and advanced analytics inaworlddominatedbydigitaltechnologyanddisruption.Wewilltouchonpotentialrisksthatstemfrombusinessdisruptionsinunprecedentedtimes,includinghowtheglobalcoronaviruspandemichasresultedinariseinfinancialcrime.WedescribehowtechnologyandinnovationarenecessaryinweatheringunforeseencircumstancesandinachievingbetteroutcomesforFinancialCrimeCompliance(FCC).

Thefinancialservicessectorisnowfacinggreaterchallengesfromsophisticatedcriminalswhohavefoundwaystoprofitfromanincreasinglydigitalisedeconomy,acceleratedpartly duetotheCOVID-19pandemic.Effortstoenhancedetectionbyaugmentinginvestmentsmadeinartificialintelligence(AI)andmachinelearning(ML),analyticsandroboticprocessautomation(RPA)havepaidoff.However,moreworkstillneedstobedonetoensurethatthesectorisabletoadequatelyrespondandcurbvariousrisksincludingfinancialcrime,andmaintainthetrustithasestablishedwithitsrelevantstakeholders.

OurwhitepaperexaminestheongoingjourneyofUOB’sAIanti-moneylaunderingsolution,fromproofofconcept(POC)toproductionstage,explaininghowitgraduallycalibratedmodelsforintegrationintocurrentbankingoperations.ItoutlinesthejustificationfortheBank’sinvestmentinadvancedanalytics,AI/MLandrobotics–notinghowthesehavebeeninstrumentalinmitigatingmajordisruptions.

DeloitteandUOBpreviouslypublishedtwowhitepapersin2018and2019.Thefirstwhitepapertitled,“Thecaseforartificialintelligenceincombatingmoneylaunderingandterroristfinancing”1explainshowfinancialinstitutions(FI)canleverageinnovationtomanageFCCeffectively.ItsharedUOB’scasestudyinsuccessfullypilotingmachinelearningtoidentifysuspiciousaccountsandtransactionswithgreateraccuracy.Thesecondwhitepapertitled,“Thefutureoffinancialcrimecompliance”2, depicted the future-state of FCC that incorporates AI, ML, RPAandnaturallanguageprocessing(NLP)tomanageevolvingfinancialcrimerisks.ItdetailswhatisinvolvedtooperationaliseMLforFCC,takingreferencefromUOB’ssuccessfullyimplementedMLmodel.

SharingUOB’stransformationstory–onitsuseofinnovativetechnologiestocombatfinancialcrimeprovidesinsightintotheimplementationprocessandchallengesexperienced.Itshedslightonthegovernanceofthetechnology,theengagementrequiredwithstakeholderstobuildtrustinthesolutions,andhowtointegratetheseintothebusinessasusualoperatingenvironment.WehopetheinsightssharedinthiswhitepaperwillencourageFIstofocusonapplyingFCCtechnologies,reapingitsbenefits,whilehelpingtoinnovateinandenhanceFCCeffortsacrosstheindustry.


0504

Introduction HowcanFIsembracethisnewreality of innovation?

Theneedtoinnovateandtoadopttechnologyhasneverbeenmorepronounced.Technologyanddigitisationarenolongera“goodtohave”forbusinesses.Businessesneedtostayconnected,and overcome constraints of physical mobility withthehelpoftechnology.Agilityishighlypriced.ThishasadirecteffectonFCCwhereembracinginnovationwithuseofAIandMLandcutting-edgetechnologywillenhancecapability,effectivenessandefficiencyincombatingfinancialcrimes.

Anna Celner Deloitte Global Banking & Capital Markets Practice Leader

Theglobalpandemic,aswellasgeo-politicaltensionsandloomingtradewarsdominatedtheheadlinesin2019and2020representinganewglobalrealitymarkedbydisruptiveevents.COVID-19haspromptedgovernmentsfromallcountriestotakedrasticmeasures3fromlockdownstoenforcedbusinessclosures.Traditionalbusinesseshavebeenhithardbythesemeasures,especiallywhereoperationsremainbrick-and-mortar-based.

Inresponse,businessesandFinancialInstitutions(FIs)haveacceleratedinvestmentsintransformingtheirbusinessmodels,andembracingdigitisationandenhancingremoteworkingcapability.WhilethismovetodigitisationhashelpedtolessentheimpactofCOVID-19disruption,accordingtoaFinancialActionTaskForce(FATF)publicationinMay20204,ithasalsobroughtnewchallengesandheightenedconcernsindealingwithnewandvariedformsoffinancialcrimes.

Wideningsophisticationincrimessuchasfraud,cybercrime,humantrafficking,slavery,crimesagainsttheenvironment,onlinechildexploitationandorganisedpropertycrimenecessitatesevengreatereffortstocombatfinancialcrimes.Thereisthereforeanurgentneedfortheindustrytoexploreandtoapplyinnovativetechnologicalsolutionsthatcanaddressthesecomplexitiesandrisks.Wehopethispaperinspirestheindustrytoembarkonthisjourneyandtobuildamorerobustfinancialcrimeriskmanagementecosystem.

In2020,worldwiderevenuesforAI/MLcompaniesareexpectedtoexceedUSD150billion,representinga12.9%increasefrom20195.ThebankingindustryinvestedatotalofUSD5.6billioninAI-enabledsolutionsin2019.Accordingtoastudy,companiesseeAIandMLasimportantcomponentsintheirstrategywheresignificantinvestmentshavebeenandwillbemade.RiskmanagementhasalsobeenhighlightedasthetopdomainforAI/MLimplementation.6

TheincreaseinAI/MLinvestmentunderpinstheincreasingdependencebybusinessesontechnologytomanageenterprise-widerisk.ThischapterexaminesthevariousinvestmentsmadeintotechnologiessuchasAI/MLanddataanalytics,andhowthishasbeenagamechangerforFIsinmanagingfinancialcrimerisks.

Effectiveness and efficiencies of advanced analyticsAsmorepeoplegoonline,dataisbecomingplentifulandpervasive.FIsandorganisationshavebeenanalysingdatatounderstandtransactionbehavioursandspendingpatterns.Theyarealsodesigningnewproductsandservicestomeetchangingcustomerneeds.Forexample,Singapore-headquarteredbankUOBhasusedinsightsfromtransactiondatatopersonalisethebankingexperienceforconsumerandbusinesscustomersacrossitsnetworkinAsia.

IntheFCCspace,datahasalsobeenusedextensivelyinidentifyingbadactorswhotrytouseFIsasconduitstolaunderillicitfunds.Typically,suchsurveillanceincludesidentifyingcomplexmoneylaunderingtypologies,anomaloustransactionsandsuspiciousfundflownetworks.

ThepositiveimpactofdataanalyticsonFCChasbeenimmense.Forinstance,itwasreportedthatananalyticssolutionappliedbyaFIuplifteditscapabilitiestodetectandtodeterfraudulentattempts.Thisresultedina26percentincreaseinsuspiciouscasesinvestigatedanda40percentincreaseinsubmissionofprovenfraudcasesforcriminalprosecution.Collectively,thistranslatedtoasubstantialrecoveryofmoneylostfromfraudfortheFI.7


07


06

Swift detection of suspicious human behaviourAI/MLhasalsobeenatopicofinterestasFIscommitheftybudgetstomanagingrisksmoreefficientlyandeffectively.OurfirstwhitepaperdiscussedtheapplicationofMLalgorithmswithself-learningcapabilitiesthatenableFIstoploughthroughlargevolumesofdataforpotentiallysuspiciouscustomertransactionbehaviours.Implementation of such platforms enables FIs to direct resourcestotacklefraudalertsthatarelikelytobetrue,reducingtimeandeffortonfalsepositives.ThisplacesFIsinagoodpositiontoaddressmorefraudincidentswithoutasignificantincreaseinmanpower.Furthermore,accurateandswifteridentificationoffraudfacilitatesafasterrecoveryoffundslost.

Forexample,abankinIndonesiausesMLtodetectnewsuspectedfraudpatterns.Theimplementationofthisplatformhasreapedsignificantresultswitha30percentreduction in the number of fraud incidents due to more accuratedetection.8

Adaptability in changing circumstancesAsAI/MLmodelscanadapttochangingFCCpatternsovertime,theyoffersignificantbenefitsinthecurrentdisruptiveenvironment.9TheadaptivelearningcapabilitiesofAI/MLaresometimesoverlookedandundervaluedwhenbenefitsofthisattributearenotapparentintheinitialstageofinvestmentsmadeintothesetechnologies.Someorganisationsmayseetheseasnewtechnologies,andquestioniftheycanbedependableanddefensibleunderintensescrutiny.

Yet,traditionalsystemsarenotthebestwhenitcomestoagility.Despitetheeffectivenessofrule-basedsystemsindetectingtransactionanomalies,ever-changingcustomerbehavioursandtransactionpatternsmeanthesesystemshavetobeconstantlyre-calibrated.Thisisahighlymanualexercise.

Thatiswhymanyorganisationsaremovingtowardsmodelsthatadapttothechangingenvironmentandself-learntoprovideinsightsthatcanbeactedonbycomplianceofficers.Shiftingfromthelimitationsoftraditionalrule-basedsystemstolearning-basedMLmodels,canhelpFIsvastlyimproveaccuracyindetectinganddeterringpotentialfinancialcrimes.

Automating repetitive jobs Inourpreviouswhitepaper,wedelvedintoRoboticProcessAutomation(RPA).Wehighlightedthekeybenefitsofautomationandhowitisnowa“must-have”forFIstoachievescaleandvaluemoreefficiently.Theautomationofrepetitiveandlow-valueactivitiesensuresthathumanresourcesaredeployedefficientlyandhighervalueactivitiesreceivemoreattention.Thisway,humanexpertisecanbemaximisedtocombatfinancialcrime.

UOB,withDeloitte’sassistance,successfullyimplementedRPAintransactionmonitoring.Withrobotstakingon manual and repetitive processes, this has led to a decrease in human error and an improvement in the standardisationoftransactionmonitoringprocessesandauditingofactivities.TheBankwashenceabletoachieveareductionof30percentinman-hoursspentonthesemanualprocesses.Typically,thesetaskswouldhavebeencumbersometoperforminremoteworkingcircumstancesduringthepandemic.TheuseofRPAhasenabledtheefficientperformanceofthesetaskswithoutdisruption.

Progress will result in more benefits COVID-19hasnecessitatedtheurgentadoptionoftechnologyanddigitisationtocontinuebusiness-as-usual(BAU)operations,withremoteworkingnowtheglobalnorm.OurfindingsalsodemonstratethatinvestingininnovationandtechnologyhelpskeepFIsaheadinthesevolatiletimes.

Withthecompetitionfromfinancialtechnology(FinTech)firms,establishedFIscannotaffordtorestontheirlaurels.FIshavetoinnovatecontinuouslytoavoidtheerosionoftheirbusinessadvantage.Theyalsoneedtodevisemarket-friendlycoststructures,facilitatetransactionswithminimalfrictionandsafeguardrevenues.Innovationisnotonlyrelevanttobusiness(front-line)butalsoincomplianceandmoreparticularly,inFCC.AsFIsinnovateandcompetefrombusinessperspective,compliancegenerallyandFCCneedtokeeppacetocontinuetobeeffective.Forinstance,asfundsmovefasteracrossborders,tradetransactionsbecomemorecomplex,customerbehaviourchangerapidlyandcriminalsconjure-upnewapproachestolaundermoneythroughFIs,thecapabilityforsurveillanceanddetectionoffinancialcrimemustalsobecomeequallyrobust.ThiscanbeachievedwiththeuseofAI,MLandRPA.

Investmentsintoinnovationandtechnologyalsocannotbeaone-offoccurrence.Constantrefinementstokeeptechnologycurrentareessentialinmanagingeverchangingfinancialcrimerisksandregulatoryexpectations.ThiscallsforthedevelopmentandimplementationofmoresustainableandadaptivetechnologiessuchasML.Theseareself-learningandcanautomaticallycalibrateasthepatternsoffinancialcrimesadvance.

Ashighlightedinthepreviouswhitepapers,employeesalsoneedtobetrainedtobeproficientusersoftheoutputofdataanalytics,AI/MLandRPA.ThiswillensuretheyarecapableofsupervisingandoperatingFCCtechnologies. The trajectory to achieve the end-goalContinualinvestmentintoAI/MLtocombatFCCisrequiredtoaddresstheincreaseddimensionoffinancialcrimerisksdevisedbyincreasinglysophisticatedcriminals.ItisalsocrucialforFIstoensurethattheyquicklydeveloptheseinnovationstostrengthentheirriskmanagementcapabilitiesandtostayaheadofthecriminals.

Aspreviouslymentioned,AI/MLmodelsusedforFCCenableFIstostrengthensurveillanceagainstfinancialcrime.ThesetoolsenhancetheFIs’abilitiestoidentifyanomalies,soastomitigatemoneylaunderingandterroristfinancingrisks.

Current landscapeTheuseofadvancedanalyticsandinnovativetechnologiesforFCCisstillinitsinfancy.Whatisclearisthatmanagementbuy-inisrequiredbeforeanyFCCapproachcanbetransformedwithnewtechnologies.GivennewtechnologiesrequireinitialfinancialinvestmentsbeforeefficienciesandeffectivenessforFCCcanbedemonstratedandrealised,faithisneededthatthesenewtechnologieswillwork.Convincingstakeholderscanbeachallenge,andinvestmentstosupportdevelopmentaresometimesmadeintranchesasthetechnology’ssuccessisrealisedstepbystep.

ForAI/MLsolutionstobedefensible,developmenttimelinesmayalsobeextended.Thisistoavoidrisksandregulatoryimplications,shouldtheseAI/MLmodelsfail.

InvestmentsintotechnologiesforFCCwillbecriticalforFIstokeepabreastofevolvingfinancialcrimethreats.FIsthathavebeendigitalisingtheirserviceswouldhaveseensomereturnsoninvestments amid the tumultuous times, as they were able to avoid acompletestandstillofoperations.Beyondthis,FIsalsoneednewapproachesandadvanceddataandtechnologycapabilitiestocontinueeffortstobecomemorerobustandeffectiveinmanagingfinancialcrimerisks.

Ho Kok Yong Deloitte Southeast Asia Financial Services Industry Leader


09


08

The end-goalForAI/MLmodelstodelivertheirmaximumpotentialforFCC,allparties(FIs,employees,serviceprovidersandregulators)willneedtohavetrustinthem.Thisleveloftrustinmodelshastobetheend-goalfortheindustry,oritwillimpededevelopment.

Providingasecurebankingenvironmentdeepensconsumertrustandconfidenceinthefinancialsystem.Ultimately,FIsrelyontrustfromtheircustomerstobuildasustainablebusiness.Inthatlight,theymustpreservestakeholdervalueandsupportfromgovernmentsandinstitutionalinvestors,amongothers.

Bridging the trust gapAnybreakthroughsintheuseofAIandMLforFCCwouldbeundervaluedwithouttrustinthetechnologysolutionfromstakeholders.Inaddition,regulatorsalsoneedtotrustthedecisionprocesstoembraceinnovationandsolutionsbeingassuredthatthesemodelsareexplainable,defensibleandcanaddressFCCriskseffectively.

Appropriate perimeters and rubrics need to be created toprovetheeffectivenessofAI/MLmodelsasatrustedallyforhumansintacklingFCCissues.ThispointwasreiteratedattheG20DigitalEconomyMinistersMeetingbySingapore’sMinisterforCommunicationsandInformation,SIswaran.HehighlightedtheparamountimportanceofupholdingtrustandsecurityinthedeploymentofAIanddataflowinanincreasinglydigitalisedworld10.

To address this requirement, many have either establishedorsuggestedtangibleframeworksandguidanceforsuchAImodels.Forinstance,theEuropean Commission11 published a white paper on AIthatemphasisesfocusingontrustworthinessintheusageofAI/MLasitsetsoutpolicyframeworkstoensureagreateruptakeofAI/ML.ThisringstrueintheUnitedStatesaswell,wheretheWhiteHouseOfficeofScienceandTechnologyPolicy12providedgovernmentagencieswithguidelinesandprinciplesfor“consideringregulationsorpoliciesrelatedtoAIapplications”.Publictrustanddisclosure,andtransparencyarelistedaskeyprinciples.

TheInstituteofInternationalFinance(IIF)andDeloittehavealso,inOctober2019,releasedawhitepapercallingforacombinationofregulatoryreform,culturalchangeandthedeploymentofnewtechnologiestoenhancehowFIscounteranti-moneylaundering(AML)threats.13Engagingkeystakeholdersinvariousstagesofthedevelopmentprocessofinnovativesolutionsisanecessarystep–itbridgesthetrustgapandbuildsconfidenceintheuseofsuchtechnologyforcombatingfinancialcrime.

TheMonetaryAuthorityofSingapore(MAS)hasalsopublishedasetofprinciplestopromotefairness,ethics,accountabilityandtransparency(FEAT).Theseareintendedas“anindustrybenchmarkandguidewhenthinkingabouthowtouseAIanddataanalytics”14.TheFEATprinciplescanalsohelpstrengtheninternalgovernanceofAIapplicationsaswellasbuildgreatertrustandconfidenceinAI/MLsolutions.

InOctober2019,theInstituteofInternationalFinance(IIF)andDeloittepublishedawhitepapercallingforacombinationofregulatoryreform,culturalchange,andthedeploymentofnewtechnologiestobettercounterthreatsposedbyillicitmoneyflowsthroughtheinternationalfinancialsystem.Innovation,withtheuseofAI/MLandRPA,isanecessarysteptowardsbridgingthegapandbecomingmoreeffectiveincombatingfinancialcrimeandbuildingtrust.Inaddition,asdiscussedintheDeloitte/IIFwhitepaper,recommendations such as public-private partnerships,improvinginformationsharing,andreformingsuspiciousactivityreportingareallnecessitiesinsharpeningcapabilitiestocombatemergingfinancialcrimethreats.

Michael Shepard Deloitte Global Financial Crime Practice Leader

GiventhatsuchtechnologiesarequicklybecomingembeddedwithinFCCprogrammes,theindustryshoulddeepencollaborationandacceleratethebuildingofthesecapabilitiestobolstertrustandbuildanecosystem.Thenextstep,istocreateindustry-levelmonitoringutilitiesincorporatingAIandML,amongstothers.WiththeuseofAI/MLandotherinnovation,asFIsbecomemoreeffectiveatmanagingfinancialcrimerisks,theindustrycouldtogetherembarktowardsagreaterwin-winphenomenontocombatfinancialcrimemoreeffectively.

Radish Singh Deloitte Southeast Asia Financial Crime Compliance Leader

ToprovideasetofguidelinesagainstwhichFIscanvalidatetheirsuccess,theMASbroughttogetheraconsortium15 consistingofFIsandFinTechs,ofwhichUOBisafoundingmember.Itsaimistocreateaframeworkknownas“Veritas”toprovideFIswithaverifiablewaytoincorporatetheFEATprinciplesintotheirartificialintelligenceanddataanalytics(AIDA)solutions.Whilestillintheearlystagesofdevelopment,thisframeworkseeksto“promotethe responsible adoption of AIDA"16.Inasimilarvein,Deloittehasenvisagedthat“TrustandConfidence”shouldformthefoundationonwhichallAI/MLmodelsarebuilt.ThecompanyhasbeenabigadvocateinbuildingtrustbetweenmanandmachinetoworktowardsacommonsetofgoalssincethefirstwhitepaperpublishedwithUOBonAI/MLinFCCin2018.

Industry players such as Microsoft17 and the Gartner Group18havealsoproposedtheuseofframeworkswithafocusonusingmaturitymodelstobolsterconfidenceandtocatalysegreateradoptionofAI/ML.Maturitymodelsareframeworksthathelpindustryplayersmeasuretheirreadinessandpotential(i.e.theircurrentandfuturestate)toimplementAI/ML.

Specifically,partiesinvolvedshouldbe:1)providedwiththemeanstomeasurethematurityofFCCAI/MLmodels;and2)abletoidentifyandimplementadequategovernanceandriskmanagementaroundspecificmodels.

ThereisthereforeaneedtoharmoniseAML/CFTrequirementsandtheprinciplesgoverningAI/MLtobuildanadequateframeworkforFCCoperations.IntheFCCspace,Deloitteviewscollaborationintheformofpublic-privatepartnerships(PPP)ascentraltoimprovingthe“legalandregulatoryframeworkandriskmanagementtoolkittoenhanceeffectiveness”19.


11


10

Measuring maturityWetaketheviewthatthedevelopmentofamaturitymodelwillprovideanindustry-wideyardstickforuseofAI/MLmodelsinFCC.Currently,noindustrybenchmarksexisttomeasureandtestthematurityoftheframeworkfordeploymentofadvancedanalyticsandinnovation.Therearealsootherobstacles:

1) Protracted development timeframe: VariousFIshavecollaboratedwithregulatorytechnology(RegTech)companies,forexample,UOBandTookitaki.Custom-builtmodelscanbeco-createdinsuchcollaborationstofitanFI’sspecificrequirementsandarchitecture.Intheabsenceofanindustrymaturitybenchmark,FIsandtheirvendorshavetodefineanidealstateandtocharttheirowncourseintermsofaddressinggapsinormeasuringtheefficacyandrobustnessoftheirmodelsandthegovernanceframework.Unsurprisingly,thisexerciselengthenstheproductiontimelineandissubjecttomuchchallengeduetothelackofabenchmarkframeworkforcomparison.

2) Inadequate user reliance: FIswhohavebeenidentifyinggapsduringthedevelopmentprocess,maybereluctanttotrustthemodel,especiallyintheabsenceofanobjectivebenchmarktoevaluateitsmaturity.

3) Duplicate operations: ThelackofregulatoryguidancehasresultedinFIsbeingreluctanttorelyentirelyonAI/MLmodels.FIsusebothAI/MLmodelsandtraditionalFCCmethodstotacklethesamealerts,resultinginduplicatework.

4) Regulatory scrutiny: RegulatorsmayincreasescrutinyonaFItoensuretheAIandMLmodel’sefficacyasitisadaptedforuseinFCC.Themodelneedstobeexplainableandrobustinmanagingfinancialcrimerisks.ThereiszerotoleranceforfailuregiventhatthestakesaretoohighforanyfinancialcrimetopassthroughanFI.However,literatureprovidingclearregulatoryguidelinesspecifictoFCC-relatedAIandMLsystemsiscurrentlyunavailable.

5) Inadequate internal governance principles and guidelines for assurance framework: As with any modelsdeployedandprocessesputinplacetomanagerisks,thereisaneedforagovernanceand

assurance/testingframework.DeloitteandUOBhavebroughttogethervariousprinciplesandstandardsbasedonourexperiencewhileworkingtogetheronthisjourney.Theseincludebestpracticesandinternationalregulatoryprincipleswhichcouldbeappliedbyanalogy,giventhattherearenoexistingdirectguidanceforreference.Wecreateddocumentationongovernanceandmodelriskmanagementprinciplesaswellasprocessestoaddresslowervaluealertswithdueconsideration.

Closing the maturity measurement gapApotentialmaturitymodelforuseofAI/MLinFCC,asdiscussedabove,canhelptheindustrybetterassurestakeholdersthattheAI/MLsolutionsarerobustforuse.Specifically,thefollowingarerequired:1) Providing a standard measure of maturity: Theindustryshouldbeabletogaugethemodels’capabilities

and maturity in a way that enables them to discern competent models from those that require further improvement.Thisinturnallowsthemtooperatefit-for-purposesystemswithassurance(inthecaseofFIsandemployees)andwithstandanyheightenedscrutinyontheiroperations(inthecaseofregulators).Amaturitymodelwillalsoaidinsettingstandardstomanageandmitigatemodelsubjectivityandbias.Italsofacilitatestheinteroperabilityofchampionandchallengermodelstocontinuallyensurefitnessofpurpose.Inaddition,thereshouldalsobeguidelinestodefinetheacceptableindustryapproachtogovernanceandongoingassurance.

2) Shortening development timelines: Usingthesaidyardstick,FIsandtheirpartnerswouldhaveareferencepointfortheirdevelopmentandimplementationroadmapandcanmorequicklyidentifyandaddressgapswithintheirAI/MLmodelsforFCC.

3) Facilitating strategic decision-making: Inthelongerterm,FIswillbeabletoproperlypositiontheircurrentsituationintermsoforganisationalmaturityaswellasmakestrategicdecisionswithvisibilityonfutureAI/MLmodelsaccordingtoadevelopmentroadmap.Wearehopefulthatinthenearfuture,therewillalsobefurtherguidanceontheuseofalgorithmsinmanagingfinancialcrimerisks.

4) Better training and awareness: Withanindustryyardstick,thisreferencepointwillalsohelptoguidestakeholders’understandingofsuchmodels.

Characteristics of a maturity modelBasedonthejourneyofDeloitteandUOBaswellasworkandresearchundertakeninthisspace,maturitymodelshavetwokeycomponents:1) Staging Mechanism: Roadmapsettingoutthestagesofanorganisation’sAImaturity–rangingfrom

aspirationaltoadvancedimplementations.

2) Guiding Principles: Asetofprinciplesunderpinningdevelopmentandoperationalisation.Theseprinciplescanbesummarisedintofourlargecategories:i)Culture;ii)GovernanceandTraining;iii)Data;andiv)ModelArchitecture.

AmaturitymodelfortheuseofAI/MlmustbetailoredtoaddressspecificneedsinFCC.EventhoughthecurrentgeneralmaturitymodelsintheirpresentformsareinadequateforFCCpurposes,theycanprovideabaselinetostartwithwhendesigningabespokemodeltoaccountforthepeculiaritiesoffinancialcrime-relatedrisksandissues.Addedconsiderationsinclude:

AscoringmatrixcanbeusedtohighlightwhereFIsareperformingwellandwherethereisaneedforimprovement.AstagingframeworkshouldalsobeconstructedtoprovidethedirectionforfutureFCC-relatedAI/MLmodels.

AppendedisasuggestedmaturitymodelframeworkbasedonDeloitte’sexperiencethusfar.WebelievethisformsthestartingpointfordevelopingamaturitymodelthatwecanberefinedandenhancedalongsidedevelopmentsinAI/MLforFCC.

Compliancewithregulatoryrequirements Theexplainabilityofmodelsandalgorithms

Establishingcultureprinciplessuchas“TonefromtheTop”

Designationofrolesandresponsibilities(acrossThreeLinesofDefence)

Undertakingarisk-basedapproach Maintainingdocumentationandanaudittrail

Puttinginplaceclearpoliciesandproceduresforgovernance,riskmanagementandescalation

Adequatetrainingandawarenessforstaff


13


12

Deloitte's suggested FCC maturity modelFigure1outlinesoursuggestedguidingprinciplesincorporatingkeyFCCrequirements.

Figure 1: Guiding Principles

Culture Data Governance & Model Risk Management Model Architecture Assurance Training

Tone From the TopEncouragingresponsibleuseofAItoenhance compliance capabilities and cultivate an innovative culture

Standardisation of Data Data should be uniform across the FI without overlap

Model Risk ManagementMonitoringofmodeldesignandconceptualsoundnessshouldbeongoing

Integration into BAU OperationsModelshouldbedesignedforsmoothandadequateintegrationintoBAUoperations

External Validation Modelshouldbesufficientlyvalidatedbyindependent third parties such as Deloitte

Human Resource & TrainingOngoingtrainingtoensureproficiencyinoperatingmodel/expertiseaswellasrecruitmentofrelevantSubjectMatterExperts in both Data Analytics & FCC

Data-Based Decision MakingStrategicdecisionsmadebytheFIaredrivenby data analysis in reliance on the model

Adequacy of Data Pools / Lakes Data used should be adequate and sufficientforthemodel’spurposes

Model GovernanceAdequateandsufficientmonitoringofgovernance(controls)shouldbeestablished

Efficiency Modelgeneratesalertswithgreateraccuracy,significantlyincreasingtruehitsandreducingfalsepositives

Internal ValidationInternal validation conducted within the FI should aspire towards an automated self-validation module conducted solely by the model

Risk Based Approach / Effective Cascade of Risk AppetiteAdoptingRBAasencouragedbyregulators,calibratingmodeltoriskappetiteofFIsetbyseniormanagementwhererelevant(e.g.Howalertsaretriaged)

Customer IdentificationEntity resolution abilities of models must be adequate for purposes of accurately identifyingindividualcustomers

Risk Profiling & Management Modelshouldbeabletoconductriskprofilinginvariousareasforbetterunderstandingandmanagementofriskexposure

ExplainabilityStakeholdersmustbeabletounderstanddecision paths, model should be able to outputcleardecisionpaths(Noblackbox)

Model EffectivenessModeleffectivenessshouldbeconstantlymonitoredviaparallelruns,challengermodels and below-line models

Quality Management Healthchecksondataqualityshouldbeconducted periodically and consistently

Adequate OversightSeniorManagementshouldbeawareofthekeyrisksaswellasmakedecisionsaroundthem

Ongoing MonitoringA periodic review of model performance metrics should be conducted to monitor performance and model health - conditions triggeringre-developmentsandre-validationsshouldbepre-defined

PrivacyCustomers’privacyshouldkeptinlinewithFIs’internalprivacyrequirementsandregulatoryrequirements

Policies & ProceduresImplementation of clear processes approvedbyseniormanagementfortheescalation of alerts and suspicious activity

Data Aggregation From Business FunctionsData owned by various business functions intheFIshouldbeaggregatedintoacentralrepository/pooltofacilitateoversight

Roles & Responsibilities Riskownershipandsegregationofdutiestoappropriatepeople,assigningofresponsibilityforthemodel(E.g.Postmortemreviews:continuousgapcheckingtoensureiftherearegaps)

Documentation & Audit TrailFIs’customerrecordsshouldberetainedonfileperFCCregulatoryrequirementsofatleast5years


15


14

Figure2illustratesDeloitte’ssuggestedFCCmaturitymodel which builds on the above principles and incorporatesastagingmechanism.Deloitteisofthe view that maturity can be adequately measured by assessingamodelagainstascoringmatrixalignedwiththeprinciplesabove.

Implementation

Maturity Stages

Internal Assurance

Stage 0: Rules-Based Models

Utilising supervised rules-based models

• Supervisedmodels:Operationalrules-basedmodelsrequiringhumansupervisioninmodeltuningandoptimisationisusedascomplementary tool for BAU purposes

• Compliancetestinganddocumentationarerequired

Internal Audit by Current Teams

• InternalassuranceconductedbyFIs’ownauditteamspercurrentpractices

• Silo-edviewofriskexposurebasedonscopeofauditconducted

Stage 0: Traditional FCC 3rd Line Assurance

Stage 1: Hybridised Rules-Based Models

Implementing AI/ML aspects (unsupervised / self-learning)

• Existingsupervisedmodelsaresupplementedwithadditionalself-learningmoduleswhichprovideinsightstohumansformodeltuningofrules

• FinetuningMLmodelthroughdeploymentofchallengermodelsaspossible alternatives

• Compliancetestinganddocumentationarerequired

• SystemsarepartiallyintegratedtoprovidetheFIwithabetterunderstandingofriskexposure

Transitioning to a supervised self-validation model

• Self-validationmoduleisdevelopedrunningparallelwithtraditionalinternalassurance practices

• Humaninterventionisrequiredtoverifyoutputandresultsfromthesystem

Stage 1: Hybridised Assurance

Stage 2: Intelligence-Led Models

Transitioning to intelligence-based models

• Intelligence-basedFCCoperationsbyanalysingcustomerbehaviouralpatterns,increasingnumberoftruepositivesforinvestigationalerts,drawingfromalldatasources

• Model’sautomatedMLsystemcanindependentlygeneratenewFCCinsightssuchasemergingtypologiesandpatterns

Moving towards automated self-validation

• Self-validationmoduleisdevelopedintandemwiththeintelligence-ledmodel

• The model is able to conduct internal assurance independently without human intervention

Stage 2: Intelligence-Led Assurance

Stage 3: Holistic Surveillance

Establishing an end-to-end view

• ComprehensiveoversightofFCCaspectswithintheFIacrossallthree lines of defence

• CompleteintegrationofFCCmodelssuchasTM,NS,KYC/CDD,Testing,Assurance,SanctionsamongstotherstoprovideaholisticvantagepointoftheFI’sriskexposure

• Conductingrealtimeandbatchsurveillancealertingdependinguponrisk,frequencyandseverityofevents

• Predictivemodellingwhichenablescomprehensivecollusionacrosschannels, products and behaviours

Establishing an end-to-end view

• End-to-endviewofFIs’currentFCCassuranceprogrammeusingdashboardsandvisualisationsofriskstohighlighttheareasrequiringattentionbytheFIs

• Anintegratedviewrisksallowsseniormanagementtoachievebetteroversightleadingtomoreinformeddecisionmaking

Stage 3: Holistic Assurance

Figure 2: Suggested FCC maturity model


17


16

Governance and risk managementGoing down traditional routesAsidefromadvocatinganindustry-widematuritymodel,anotherapproachtobuildingtrustinstakeholdersistoensurethatrobustgovernanceandriskmanagementstructuresareembeddedwithintheframeworkwhendeployingAI/MLmodels.

RegulatorshavecontinuedtostressFIstoensurethatcompliance-relatedissuesarelayeredwithstronggovernanceandriskmanagement.ThishasbeenaprominentfacetoffinancialcrimecomplianceandwillundoubtedlycarrythroughtoAI/MLmodels.

What to look out for?Atthecoreoftheissue,allmodelsneedtobeexplainable–sothathumans,especiallyend-users,understandtheunderlyinglogicthatdrivesthedecision-makingprocess.AnAI/MLmodelalsoneedstoincludeadequateoversightandriskmanagement,clearpoliciesandproceduresforescalation,designationofrolesandresponsibilitiesandmodelexplainability.

Thematuritymodelproposedabove,alignedwiththesaidprinciples,canpartiallyaddressconcernssurroundinggovernanceandexplainabilityofthemodel.ButwiththelackofuniformityinapproachofAI/MLuptakeinFCC,theindustrycanonlyprovideasetofgenericguidelines.FIswillneedtoadaptframeworksaccordingtotheirwidergovernancestructure,technologyarchitectureandspecificneeds.

DeloitteandUOBrecognisedtheseconsiderationswhenimplementingtheBank’salerttriage.Wedevelopedthelowpriority(L1)alertmanagementguidelines.TomakeAI/MLmodelsreliableandrobust,“confirmed”falsepositivesaresegregatedaslowpriority(otherwiseknownasthe‘L1bucket’).BothUOBandDeloittehavedevelopedguidelinesonhowtomanagesuchL1alerts.Wehavealsoco-createdgovernanceprinciples,regulatoryexpectationsandcompliancerequirements.

Use case: Low priority (L1) alert management guidelinesL1alertsholdahigherprobabilityofbeingfalsepositives.UOBandDeloittehavedevelopastreamlinedapproachtoworkingwithL1alerts.First,thetransactionalertsmonitoringteamanalysestheL1alertstoruleoutanyprobabilityoftruepositivesbeingerroneouslyembeddedintheL1bucket.ThesealertsarethenfilteredbymappingthemagainstriskindicatorssetoutinUOB’sinternalpoliciesandFCCriskgovernanceprinciples.TheguidelinesalsoestablishprudentoperatingproceduresfortheteamintheeventanyL1alertisidentifiedtohavepotentialrisksorpreviouslinkagetoSTRs.

Model risk management guidelines for the use of AI/ML in FCC Oneofthemanyprinciplesprovidedinthisdocumentincludetheneedforongoingcalibrationtoensurethatthemodelcontinuesoperatingasintended.

DeloitteandUOBdevelopedandimplementedguidelinestoensurevisibilityofthemodelbyapplyingtraditionalAML/CFTrequirementsaswellasMAS’suggestedFEATprinciples.Theguidelinescoverthefollowing,amongstothers:a) Policiesandproceduresb) Oversightfromseniormanagementc) Explainabilityofthemodel’sdecisionpathsd) Managingbiase) Applyingmodelgovernanceprinciplesbasedoninternationalpracticef) Assuranceguidingfundamentalconsiderations

Embeddingtheseprinciplesintoamodelwithtangibleandconcretestepsensurescomplianceaswellaseffectivenessofthemodel.Withtheseinplace,trustcanbestrengthenedasallpartiesinvolvedareabletounderstandtheBank’sapproachtomanagingriskevenwhenemployingnon-traditionaltools.

Building trust and confidenceWhilesomeFIshavebeeneagertoimplementAI/MLsolutionsintotheirFCCoperations,otherstakeholdershavebeenslowertodoso.ThesestakeholdersarenotopposedtodeployingtheuseofAI/MLinFCCbutarewaryoftheconsequences,shouldthesemodelsfailtomeettheirobjectives.TheimplicationsareheightenedwhenanAI/MLmodelfailsinidentifyinginstancesofmalfeasancebyaFI,itsemployeesorcustomers.

UOB and Deloitte have published this series of white papers with the aspiration that other industryplayerscantakereferencefromcitedreliableusecasesandembarkonsimilarjourneys.

Oursuggestedapproachtoconstructingtangibleframeworksandbenchmarksforthemeasurementofmaturityaffords:

• FIs better visibility in terms of next steps

• Otherstakeholderstheabilitytoratemodelsanddecidehowmuchtrusttoplaceinthem

Establishinggoodgovernanceandriskmanagement,anddemonstratingthattheyhavebeencarefullyconsideredandimplemented,willgofarinbolsteringregulators’confidenceinaspecificAI/MLmodel.Shouldtheseareasbeachieved,theindustrywillbesignificantlyclosertothedesiredend-stateofhavingallstakeholders(e.g.FIs,employees,regulators)placetheirtrustinthesetechnologiesforthepurposesofFCC.ThiswillacceleratetheuseofAI/MLintheindustry.

ChallengesTherearealsootherfactorsforconsiderationwhenimplementingsuchaframework:1) Harmonising regulatory compliance and internal controls–Importingregulationsbuiltfortraditional

FCCoperationsintoacompletelynewterritoryofuseforAI/MLinFCCrequiressignificantworkinharmonisingtherequirementsofcomplianceandcontrols.Thisisnecessarytomanageriskalongsidegoodgovernanceandaccountability.WesoughttohighlightthesecoreprinciplesthroughouttheseriesofwhitepaperscreatedbyUOBwithDeloitte.Thisjourneyisacontinuousoneasmodelsbecomeincreasinglyadvancedandsophisticated,andprinciplesneedtoevolve.Itisnotaone-offinvestment.

2) Multiple stakeholders–Formulatingabestpracticeframeworkrequiresinputfromanentireindustryandpresentssignificantlogisticalchallenges.Thepreferencesofdifferentplayersaddtothecomplexityandlackofhomogeneity.Whileitisunlikelythattherewillbegreatdisagreementintermsofthebroaderprinciplesandcomponents,therecouldbesomedifferencesasdetailsareworkedoutacrosstheindustry.Thebroadprinciplescanserveastheuniversallyapplicablebaseline.EachFIcouldthenworkintherequisitedetailsbasedontheiruniquearchitectureandneeds.

UOB and Deloitte have published this series of white papers with the aspiration that other industry players can takereferencefromcited reliable use cases andembarkonsimilarjourneys.


19


18

UOB’sapproachtodevelopingaRegTechecosystemUOB’s journey OurpreviouswhitepaperstouchedontheBank’sjourneywithDeloitteandRegTechfirm,Tookitaki,fromPOCtothetechnical-livestageofthemodel.UOBandDeloittehavesincemadesignificantprogressbyconductingvalidationexercisesoftheproductionmodelstogobusiness-liveinthesecondhalfof2020.Thisincludesanindependenttechnical-livevalidationconductedbyDeloitte,aninternaljointbusinessvalidationandperiodicinternalperformancemonitoring.

Figure 3: UOB AI Journey - A prudent approach to developing a RegTech ecosystem

2017– 2018

Independent Rules Based

Model Validation

Feb2018

Kick-off2-in-1 ML

Proof of Concept

Jul2018

ConductedInternal Validation

By DMO

Aug2018

Oct2018

ConductedExternal

Independent POC Validation

Nov2018

Jan2019

Joint White Paper

Vol 1

Project Approved& Implementation

Kick-off

May2019

Nov2019

RegTech AI Governance

Framework

Nov2019

Joint White Paper

Vol 2

Mar2020

Independent Post Tech Live

Validation

Jun2020

2020

Joint Business Validation

GC Performance Validation

Joint White Paper

Vol 3

Year 1: Our POC Journey

Year 3: Our Productionalisation Journey

Year 2: Our Implementation Journey

POC SuccessfullyCompleted

Tech LiveIn Production

TM & NS

Vol. 1: Entitled “The case for artificial intelligence in combating money laundering and terrorist financing”

Vol. 2: Entitled “The Future of Financial Crime Compliance”

Monthly

Business Live

Oct2020

Dec2020

Journeycontinues…

Governancemanagement

approach

Conducting independent model validations - DeloitteAsdiscussedinourfirstwhitepaper,thePOCAI/MLmodelunderwentatwo-foldvalidation–firstbyUOB’sdatascientistswithintheBank’sDataManagementOfficeandnext,byDeloitte.ItprovesthatthePOCmodelisconceptuallysoundandcapableofdeliveringgoodmodelperformance.

RecognisingthatthisAI/MLRegTechsolutioncouldplayastrategicroleinenhancingtheBank’seffectivenessinAMLriskmanagement,UOBandDeloitteinitiatedadditionalindependentassessmentandvalidationofthesemodelspriortogoingbusiness-live.UOBalsoworkedwithDeloittetodevelopaRegTech-specificAI/MLmodelmanagementframeworktoguidekeyaspectsoftheAIgovernanceandmodelarchitecture.Thisinturnensuredthemodel’sveracityandstability.

Governance AI model management frameworkInVolume2,welaidoutamodelgovernanceframeworktoguidetheimplementationofMLmodelsinthefollowingareas:a) Modelriskmanagementb) Managingbiasesc) Explainabilityofmodelsd) Applicationofdataprivacy

e) FEATprinciplesf) Datamanagementg) Assuranceandtestingofmodelsh) Incidentresolution

TheobjectivewasfortheBanktomitigateandtomanagepotentialrisksfromtheuseofmodelsthatmightaffectitsregulatorycomplianceobligation,customers,shareholdervalueandreputation.

Inpreparationforbusiness-live,UOBintegratedgoverningprinciplesintotheBank’sbusinessoperationsandcontinuestolayoutbuildingblocksforeffectiveandsustainableAI/MLgovernancepostbusiness-live.Thisconstructalsoformsthebasisofourvalidationregimeofanymodel’sgovernancestructure.

ThekeypillarsarelaidoutinFigure4.

Figure 4: Governance RegTech AI models in UOB

FEAT Principles

Principles to Promote Fairness, Ethics, Accountability and Transparency (FEAT) in the Use of AI and Data Analytics

Governance Pillars of AI Model Risk Management Framework

GC worked with Deloitte to develop a RegTech AI Management Model framework which includes the governance pillars.

UOB Model Risk Governance Framework

Risk committee forms the second line of defense performing risk and control oversight functions

The new AI/ML model management will be subsumed under the Bank’s Model Risk Governance Framework to provide an overarching structure to effectively manage model risk.

Independent Validation

Assess risk management and controls

Internal Review

• External Audit• Internal Audit

Board Risk Management Committee

Senior Management Committee

Management Working Group & Committee

Supporting Technologies and Processes

Data Management

Model Risk Management

Model Lifecycle Management

Organisation and Governance


21


20

Complianceisaboutdoingthingsright.Astrongriskmanagementandcomplianceculturedemandsfinancialinstitutionsexaminehowtheirsystemsmeasureupagainstcurrentthreatsandthenewwaysinwhichcriminalsseektoinfiltratethefinancialsystem.Investingintechnologiessuchasartificialintelligenceandanalyticsareimportantastheygivefinancialinstitutionsthefirepowertheyneedtofightbackandtokeepthesystemsecure.WhenUOBbeganourtransformationjourney,wedidsonottocreatenewtechnologies,buttoensurewewerestrengtheningourdefences.Welookedaheadtoseewhatneededtobedonetoservecustomerswell,tokeeptheirtrustinusasaresponsiblebankandtoexceedtheirexpectations.Wehaveandwillcontinuetobeguidedbytheseobjectives.

Victor NgoHead of Group Compliance, UOB

RegTech AI model architecture validation approachInadditiontorobustgovernance,themodelneedstoenabletheBanktofulfildesiredbusinessobjectivesandexpectations.Assuch,thesecondaspectofDeloitte’smodelmanagementframeworkprovidesacomprehensivesetofguidelinesanddimensionsthatcanbeusedtoapproachanymodelvalidationexercise.Thescopeofeachvalidation exercise is dependent on the extent to which models can be tested, the availability of techniques, as well asspecificmodelrisklevels.OurvalidationapproachaimstoassessthekeydimensionssetoutinFigure5.

Figure 5: Model architecture validation framework

Conceptual soundness

Governance &

Operations

Data

M

anag

emen

t

Ongoing

Monitoring

Outc

ome

Anal

ysis

Vendor Selection

Independent model validations by DeloitteInpreparationforthenextstepofitsbusiness-liveprocess,UOBengagedDeloittetoconductanindependentvalidationofthetechnical-livemodel.Thisservedtoevaluatethesoundnessofthemodelgovernance(usingtheGovernanceAIModelManagementFramework)andsolutionarchitecture(viatheModelArchitectureValidationFramework).

Deloitte’svalidationrevealedpositiveresultsfortheproductionmodel’sperformance,leadingUOBtoconcludethatitisconceptuallysoundandrobust.

Figure6illustrateshowthemodelworksconceptuallyonalertsgeneratedbyTransactionMonitoring(TM)andNameScreening(NS)systems.

Figure 6: How the AI/ML model sorts alerts from TM and NS systems

The set above visually represent a bulk of alerts. Only a smaller amount is confirmed to be True Positive. However, the entire set must be reviewed in the same manner.

L3 - AI predicts high priority for x% of the alerts

L2 - AI is on the fence for y% of the alerts

L1 - AI predicts low priority for z% of the alerts

Alerts

Before After (Example)AI-ML solution looks at multi

AML risk dimensions concurrently and recognize

relevant risk-patterns

L3

L2

L1

AI


23


22

End user validations in UOBInordertooperationalisethemodel,UOBsoughttodetermineifthemodelwassusceptibletoanysystematicmisclassificationofalerts.TheBankconductedaninternalreviewbycomparingsystem-generatedresultsagainstthoseperformedbytheBank’sanalysts.Anymismatcheswereevaluatedbythevalidationteamtoascertainiftheinconsistenciesobservedstemmedfrommachineorhumanerror.Theassessmentdidnotrevealmajorgapsbetweenthemodel’spredictionsandoutputbybusinessusers.TheconclusionthatthishadbeenasuccessfulexercisegaveUOBtheassurancetoimplementthemodel.

Periodic performance monitoring UOBhasinstitutedaperiodicperformancemonitoringprocess.ThisassessmentprocessrequirestheBanktoexaminefourkeyaspectsofthemodel:

Wehaveobservedthatthemodel’sperformanceisoperatinginanoptimalrange.Thisisdespiteanincreaseintransactionvolumewhenbankingtransactionpatternsshift,orduringseasonalfluctuationssuchasfestiveseasons.

Model's prediction outcomeDuringthevalidationprocess,weobservedthatthepredictionoutcomefromthemodelremainedconsistentwhencomparingtheresultsgeneratedduringPOCandfromtheparallelrunintheactualoperatingenvironment.

Name screening modelsUOB has observed that the NS models for Individual and Corporate customers performed within the prediction boundariesestablishedduringPOCandTechnical-Livestages(setoutinoursecondwhitepaper)achievedabove96 per cent truepositivealertsconcentrationintheHighpredictionbucket(L3).

Sustainability – maintain low misclassification ratios; prioritisation ratios; true positive alerts concentration; and high accuracy rates observed during model training

Flexibility – machine is able to self-learn continuously from changes in behavioural patterns and automatically recommend to humans how to further enhance the model

Resilience – ability to adapt to larger volumes and larger values of transactions, as well as more cross border-transactions due to seasonality (e.g. Christmas, Lunar New Year) or events affecting long-term trends such as disruptions discussed in this paper

Applicability – of the model to different business segments (Corporate; Private Banking) and rules configured in the rule-based TM and NS systems

Transaction monitoringThe TM model presented positive outcomes with 96 per cent true positive prediction accuracy in theHighpredictionbucket(L3)whichflagsalertsdeemedastruepositivealertsorhighlysuspicious.ThiswasachievedduetoUOB’sTMmodelrelyingonthousandsofclues(features)whenanalysingtransactionbehavioursandpredictingthelikelihoodoftruepositivealerts.Giventheseparameters,themodelencountersasignificantlyhighernumberofinstanceswherethelinebetweenatrueorfalsepositiveislessevidentascomparedwiththeNSmodels.

Figure 8: Results showing the effectiveness of AI/ML models

Engaging stakeholdersEngagementofbothinternalandexternalstakeholdershasbeenkeytoourjourneyofimplementingnewtechnologiestocombatfinancialcrime.Tohelpstakeholderstrustthattheseinnovationscanworkreliablyandresponsibly,wehadtoensuretheywereclearontheoperatingmodelofthesesolutionsaswellastheoutcomesproduced.

UOB’sinitiativeisbuiltonthebackoftheMonetaryAuthorityofSingapore’sstrongencouragementforfinancialinstitutionstoleveragetechnologytocombatmoneylaunderingandterroristfinancingrisks.Forinstance,theuseofdataanalyticscanhelpimprovethedetectionanddisruptcriminalbehaviour,leadingtobettersupportoflegitimatebusinesses.Asmorefinancialinstitutions implement enhanced detection capabilities, coupled withclosepublic-privatecollaborationintargetingkeyrisks,thefinancialsystemwillcontinuetoenhanceitsresiliencetofinancialcrime.

96.3%96.0% 97.9%

Total L3 High Priority Bucket Population deemed True

Positive / Highly Suspicious

Total True Positive Population can be found in L3 High Priority Bucket

TransactionMonitoring Individual Corporate

Name Screening

RegTech Eco-system

Stakeholders

IoTProviders

AI PlatformProviders

ExternalAdvisors Customers

FinancialInstitutions

SolutionPartners

Regulators


25


24

Benefitsinatimeofdisruption

UOB: Benefits in a time of disruptionAsatthetimeofpublication,theCOVID-19pandemiccontinuestodisrupteconomies,industriesandbusinessesacrosstheglobe.

UOB,likemanyothercompanies,transitionedquicklytoremoteworkingwithoutcompromisingthespeed,safetyandsecurityofitspolicies,programmesandprocesses.ThiswasduetotheBank’songoingtechnologyinvestments.

IntheareaofFCC,UOB’searlierinvestmentsandintegrationofautomationandAIintoitsoperatingenvironmentmeantthattheBankavoidedmuchofthedisruptionarisingfromCOVID-19.ItstechnologyinvestmentstoenhanceitssystemsalsoenabledtheBanktocombatfinancialcrimeseffectively,evenasillicitactivitiessurgedduringthepandemic.

1) COVID-19 pandemic: Withgreaterimpetusforacashlesseconomyamidthepandemicandascashlesstransactionscontinuetogrow,existingmodelsusedinsurveillancesystemsforFCCneedtoberecalibratedtoreflectthecurrentsituation.Typically,suchchangescanbealengthyandcostlyexercise.ButthisisgreatlymitigatednowwithUOB’sMLcapabilities.Thetechnologyquicklymakessenseofdatatoidentifynewpatternsandinsights.

TheBank’suseofRoboticsProcessAutomation(RPA)hasalleviatedmanpowerconstraintsforFCCinSingapore,whereUOBisheadquartered.Robotsperformrepetitiveandcomputationallychallengingworkwhichfreesuptimeforhumananalyststomakedecisionsandjudgementsbasedonaccurateinformation.Complianceanalystsnolongerneedtogeneratetime-consumingreportsmanuallyandonsite.Thisproveduseful,particularlyduringthecircuitbreakerinSingaporewhentheBank’sComplianceteamwaslargelyworkingfromhome.UOBisintheprocessofimplementingRPAinitstransactionmonitoringprocessacrossitsglobalnetwork.

Foreveryalert,therobotextractscustomerprofileinformationandtransactiondatafromvarioussystemstoformasinglereport.ThisisthenbeefedupusingAdvancedAnalyticsandNaturalLanguageProcessingtoprovidegreaterdatapointsandavisualrepresentationofthecustomer’sflowoffunds.Itenablesanalyststofocustheirattentiononsuspiciousalertsamidahighalertvolume.

Figure 9: Overview of UOB’s RPA system

TM Alert

Examples of RPA DA Output

RPA

+ =

Data Analytics

AIAlert Prioritization

Integration

InsightData Analytics

Prod

uctiv

ity

RPA

PredictionsAI/ML• Analyst Notebook

• Profile Write Up• Flow Through of Funds• L1 Alert Management• SONAR Form prefill• CAD Transaction Report

Visual AnalyticsNatural Language reportNetwork Link AnalyticsAI Output ComputationData TransformationData Transformation

2) ASEAN network connectivity: OnboardingnewcustomersrequiresperformanceofduediligencetoidentifycustomerswithhigherAML/CFTriskprofiles.Coupledwithstrongregionalnetworkconnectivity,oursolutionprovidedamechanismtoenableamoreeffectiveidentificationofextendedlinkagesofcustomerthatmaynotbeapparentatthepointofonboarding.

Theuseoftechnologyintheformofnetworklinkanalytics(NLA)hasproveninvaluableinprovidingabigpictureviewintheareaofTM.NLAexaminesdirectandindirectrelationshipsbetweencustomersandtheirtransactioncounterpartiesforthefollowinginsights:

• Customeridentification–identifyingcustomerswithshellcompanycharacteristics

• Counterparties’analysis–Understandingcustomers’counterpartiesandtheirtransactionswithUOBcustomers.

• Flowoffunds–visualisingcustomers’flowoffundsandidentifyingnewhigh-risktransactionpatternsandbehavioursmoreeffectively


27


26

AtUOB,werecognisedearlyonthevalueofinsightsfromcomplexdatasetsinenablingustodriveinnovationforourcustomers.Inanticipationoftheever-increasingvolumeandvelocityofdatathatarebeinggeneratedeachday,andintegraltoourstandardisedregionaltechnologyplatform,wedesignedandbuiltarobustandsecuredataarchitecture.Onthatfoundation,wecreatedadatalakeatanenterpriselevel. Incombatingfinancialcrime,thisuniquedataarchitecture enables us to have a holistic view of qualitydataacrossalllinesofbusinesses.ThismeansthatweareabletotestmorerigorouslyandaccuratelyAI/MLsolutionswithinourAMLriskmanagementsystemstoenableswifterandmoreeffectivedetectionofcriminalsevenastheybecomemoresophisticatedintheirtechniques.This is crucial as we continue to invest in technologytoenableasafeandsecurebankingexperienceforourcustomersforthelongterm.Susan HweeHead, Group Technology and Operations, UOB

UOB FCC: The way forwardTodrivefurtherinnovationintheFCCspace,UOBhasmappedoutfiveareasoffocusforitsAML/CFToperations.Itsaimistoleveragetechnologytodrivedata-drivendecisionmakingbycomplianceofficers.

• Robust Enterprise Applications: AML/CFTmonitoringcapabilitieshavebeenbuiltintoenterpriseapplications.InsteadofjustusingtraditionalAML/CFTapplications,theBankcannowharnessdataanalyticsandmachinelearningtodeepenitsunderstandingoftheriskprofileandtransactionbehaviourofcustomers.

• Big Data: AML/CFTdatapointsresideindozensofenterprisesystemsacrosstheBank.AcentralbigdataplatformaggregatesthesedatafortheuseofAI/MLintandemwithAML/CFTanalytics.WorkingwithtechnologypartnersthatprovideBigDataas-a-service(seefigure10)buildsontheBank’sdatainfrastructuretoprovideitwiththeflexibilityandscalabilitytodeployanAI-optimisedinfrastructureplatforminashortertimeframe.Inthisregard,UOBpartneredHewlettPackardEnterprise(HPE)toenabletherapiddesignanddeploymentofAIsolutionssuchasitsAnti-MoneyLaunderingSuitewhichUOBandTookitakico-created.HPEalsodeliveredapubliccloudexperiencewhichgaverisetobettercosteffectiveness,controlandagilityfortheBank.

• Data Analytics: Withfinancialsystemsbecomingincreasinglyglobalised,extractingknowledgeandinsightsfromAML/CFTdatacontinuestobecrucialandcannolongerbetheskillsetsofjustafewprofessionals.TheBankhaslaunchedseveraltrainingprogrammes,includingitsflagshiplearninganddevelopmentprogrammeforallemployees,totrainitspeopletobedataconversant.Datachampionsacrossallfunctionsandbusinessunitsareabletotapdatadashboardsandnetworkanalyticstoolstoanalyseandtovisualisedatatopowertheirdecision-makingprocess.WithintheBank’scompliancefunction,effortsarealsounderwaytointegrateAML/CFTadvancedanalyticsintoothercomplianceprocesses.

• Artificial Intelligence / ML: AI/MLhavebeensuccessfullyimplementedforTMandNS.TheBankislookingtoextendtheimplementationofAI/MLintoadditionalareassuchasSanctionPaymentScreeningandKnowYourCustomer(KYC)riskprofiling.

• Automation and processes uplift:Automation,dataanalyticsandAIcanmakeefficientdailycomplianceoperations.RPAcanbridgethegapforuserslookingtousedataanalyticsandAIineverydaydecisionmaking.

Figure 11: Five Pillars of continuous innovation

ThefivepillarsofcontinuousinnovationintheFCCspacecannotexistinsilos.UOBhasdemonstratedthroughtheexamplesabovethattheyworkbestinharmony.

Optimised infrastructure for

data analytics

Support Services and Operations

Design and Implement

Figure 10: Big Data as-a-service

Automation & Processes

UpliftAI/ML

Big Data

Data Analytics

Enterprise Applications

Financial Crime Compliance RegTech Ecosystem


29


28

Thenextlap–IntegratingAI/MLintofinancialcrimecompliance

UOB’s point of viewThebenefitsofadvancedanalyticsandtechnologicalinnovationplayanincreasinglycrucialrolewhenitcomestofightingfinancialcrime.Thisextendsbeyondimprovingefficienciesandinsulatingorganisationsagainstunexpectedmacroeconomicevents.Thereisnotimemorepressingandrelevantthannowforsuchtechnologiestobecomepartofdailyoperations.UOB’sjourneyhasdemonstratedthatthisisbothfeasibleandpractical.

TheinterconnectednessofthefinancialsystemmakesitevenmorepertinentforFIstoembracedigitaltransformation.Oneweaklinkinthefinancialsystemcanresultinaglobalwebofsuspicioustransactionsandpayments.AmultipliereffectinflaggingsuspiciousactivitiesandcombatingfinancialcrimecanbeachievedoncemoreFIsadoptnewFCCtechnologies.StakeholderssuchasFIs,regulatorsandindependentvalidatorssolutionarchitectsneedtoworktogetherasanecosystemtoexpandtheuseofadvancedanalytics,AI/MLandRPAinareassuchas:i) monitoringAMLcustomerriskbyaggregatingcustomerdatafromvarioussourceswiththehelpofa

centraliseddatarepository;ii) monitoringtrade-basedmoneylaundering(TBML)risksandredflags;andiii) effectivesanctionspaymentscreening

FCCstandards,suchasthegovernance,riskmanagementandmaturityassessmentstandardsforuseofAI/ML,alsoneedtobestrengthenedcontinuallywiththeuseoftechnologytoaddressnewthreats.Suchinitiativesshouldinvolveaclosepartnershipbetweenthepublicandtheprivatesectors.

ManagingriskisintegraltohowUOBensuresthesustainabilityofourbusinessandcreateslong-termvalueforourcustomersandstakeholders.Enablingthisisourstrongriskmanagementframework,policiesandprocessesaswellasinvestmentintechnologyandinnovation.WithincreaseddigitalisationcomesnewdimensionsofrisksintheareaoffinancialcrimeandassuchtechnologybecomesevenmorepertinentforFIstosafeguardcustomersandthefinancialsystem.Theriskmanagementguidelineswhichweco-developedwithDeloitteprovideFIswithastartingpointtoensurethatrobustpoliciesandprocessesareinplaceastheytapAI/MLtomanagenewthreats.

Chan Kok SeongGroup Chief Risk Officer, UOB

Holistic surveillanceForFIstomakemoreinformedstrategicdecisions,thereisaneedtoshiftthecomplianceregimefromasilosapproachtoonethatismorecomprehensiveandrobustinmanagingmaterialrisks.Suchanapproachuses“datafromallrelevantsourceswithinthefinancialinstitutiontotransformthevisualisationoffinancialcrimerisks.”2

Our envisioned solution architecture Deloitte’senvisionedholisticsurveillancearchitecture–fromthefirststepofsynthesisingvariousdatastreamstothelaststepofgeneratingariskexposurereportforend-users–issetoutinFigure12.

FIsneedtobecomemoreagileindetectingandpreventingfinancialcrimeandvisualisingtheirriskexposurewithacustomisabledashboardmayhelpthisprocess.Thevisualisationwillprovideinsightsontheconnectionsbetweendataoncommunication,transactionsandbehaviour.Bothinternal(conduct)andexternalthreatscanalsobeexaminedandflaggedforfinancialcrime,inadditiontoexistingmonitoringandscreeningefforts.

Figure 12: Holistic surveillance architecture

Connect

Data types

Ingest

Text extraction

Store

Storage types

Analyse

Analysis types

Case Management

Case types

Report

Report types

Structureddata

Multipleentities

Opportunity

POTENTIALLYSUSPICIOUS

Near-future Key Value Database Graph Analysis

File Storage

Graph Storage

Unified Searching

Potentialcollaboration

Unstructureddata

MIS Reporting

“Let’s act before we lose out”

Followup Task

DataDocument

StatusCollaboration

PersonA

PersonB

PersonC

Regularcontacts

Buysfrom

Wife of

Samecountry

as

Oppositegender to

Friendof


31


30

Conclusion

Thecontinuedeffectivedetectionandpreventionoffinancialcrimerequiresongoingeffortandinvestmentinoperationalisingtechnologiessuchasautomation,advancedanalyticsandAIinthemainstreamFCCframework.TheCOVID-19pandemichasalsounderscoredtheneedtoadoptthesetoolstoimproveadaptabilityandagilitydemandedbyanincreasinglyconnectedworlddefinedbyconstantchange,disruptionandglobalevents.FIsthathaveincorporatedtechnologiesforFCCwouldfindtheyaremoreadeptduringthesetryingtimestomitigaterisks.ThishasbeenthecaseforUOB.

EnsuringarobustFCCprogrammeisanongoingeffortgiventhatcriminalbehaviourscontinuestomorphandbecomemorecomplexasbadactorstakeadvantageofthechanging,moredisruptedandmoreconnectedworld.Thisinturndemandscompliancefunctionstobeasagiletoputinpreventativemeasurestoensurethatfinancialsystemdoesnotbecomeaconduitforillicitactivities.RecenteventshavedemonstratedthatemployingtheuseofAI/MLandRPAhasenabledUOBtoridethroughsuchchallengeswithgreatereaseandemergeonbetterfooting.

BeyondmeetingBAUneeds,investmentintheseinnovationshascarriedgreaterbenefitsinunprecedentedcircumstancessuchastheCOVID-19pandemic,asseenfromUOB’sjourneythusfar.SetwithinthecontextofheightenedregulatoryfocusandFCCrequirementscoupledwithlimitedresourcesinFCCfunctions,FIshavebeentaskedtodomorewithlessinthefightagainstfinancialcrime.Withthatinmind,theapplicationofinnovationsuchastheuseofAI/MLmodelsforNSandTMrepresentsthedawnofmoreeffectivecomplianceregimesandusherstheriseofwideranddeeperapplicationoftechnologiesasmootedabove.Movingintoapost-pandemicworld,theindustrymaywishtotakethesamestepsasorganisationssuchasUOBandothertechnology-orientedFIstostayrelevantandreadytocombatnewwavesoffinancialcrimesregardlessofpeace-timeordisruption.

Thepotentialoftheseinnovationscanonlybefullyrealisedwhenrobustandadequategovernance,aswellasriskmanagement,areembeddedwithintheinnovationframework.Thisisafundamentalandvitalsteptowardswidespreadoperationalisationanditsimportancecannotbeemphasisedenough.

TappinginnovativetechnologiesenableFIstotakeastepforward,towardsthevisionofholisticsurveillance.OncetheFIhasestablishedrobustgovernanceframeworksacrossallmodels,technologysolutionscancreatealayeroverexistingsystemsintheFItobringtogetherawiderangeofdataandtoprovideseniormanagementwitha360-degreeviewofrisksacrosstheorganisation.Thiswillnotonlyprovidegreatertransparencyontheinherentandresidualrisksinthebusiness,butalsoensurethatFIstapintoallavailabledatawhilemakingriskdecisions.

Inourview,theuseofAI/MLandRPAenhancestheriskmanagementcapabilityofanFCCprogramme.ThiswillbringabouttheresultanteffectofgreatertrustintheFIbyitscustomers,regulatorsandotherstakeholders.

WhilenewdisruptionsundoubtedlyposeseriousthreatstoFIs,theyalsopresentFIswiththeopportunitytoacceleratethedevelopmentofnewFCCcapabilitiesandtools.

Asevidencedbythosethathaveworkedtostayaheadofthecurve,whatisneededareindustry-wideeffortsandclosecollaborationofstakeholderstoconcretisethepathwaytothrivingFCCfunctionsinthisnewworld.

Asexploredinourseriesofwhitepapers,thefutureofFCCisnotadistantyonder–itisherenowforadoption,creatingasystematicallyinterwovencommunitythatcombatsfinancialcrimewithsharpenedcapabilityanddeeptrustinthesystem.

Wesummarisekeyareasasbeingthefollowing:1) EncouraginganFCCmaturitymodel–creatinganindustry-wideagreedstandardforbenchmarking

ofanFI’sprogressandreachingaconsensusonthegeneraldirectionofdevelopmentwillprovideanimplementationroadmapforreference.

2) Ensuringarobustmodelgovernance–governanceframeworkswithhighlevelsofgranularitytailoredforuniquemodelsaswellasindividualFIs’widergovernancestructuresshouldbedeveloped,basedonFCCregulatoryexpectations,controlsandrobustriskmanagementstandards.


33


32

End notes

1) RadishSingh,NickLim,EricAng,‘TheCaseforArtificialIntelligenceincombatingmoneylaunderingandterroristfinancing’,Volume1,November2018,DeloitteandUOB,https://www2.deloitte.com/sg/en/pages/financial-advisory/articles/the-case-for-artificial-intelligence-in-combating-money-laundering-and-terrorist-financing.html

2) RadishSingh,MinLiu,NickLim,EricAng,‘TheFutureofFinancialCrimeComplianceACompellingUseofInnovationinaConvergingDigitalandPhysicalWorld’,Volume2,November2019,DeloitteandUOB,https://www2.deloitte.com/sg/en/pages/financial-advisory/articles/financial-crime-compliance.html

3) YuanYang,EdwardWhite,RobinHarding,KiranStacey,CliveCookson,NajmehBozorgmehr,MilesJohnson,SteveBernard,JackFrancklin,‘Howcountriesaroundtheworldarebattlingcoronavirus’,FinancialTimes,March10,2020,https://www.ft.com/content/151fa92c-5ed3-11ea-8033-fa40a0d65a98

4) FATF,‘COVID-19-relatedMoneyLaunderingandTerroristFinancingRiskandPolicyResponses’,May2020,https://www.fatf-gafi.org/media/fatf/documents/COVID-19-AML-CFT.pdf

5) IDC,‘IDCForecastsStrong12.3%GrowthforAIMarketin2020AmidstChallengingCircumstances’,August4,2020,https://www.idc.com/getdoc.jsp?containerId=prUS46757920

6) FintechnewsSingapore,‘3in4BanksinAsiaWillInvestinMachineLearningThisYear’,April23,2019,https://fintechnews.sg/30005/ai/refinitiv-ai-and-machine-learning-to-transform-financial-services/

7) SAS,‘Wherehumancapabilitiesfail’,https://www.sas.com/en_us/customers/allianz-fraud-management.html

8) PriyankarBhunia,‘Enhancingcustomerjourneysandimprovingfrauddetectionthroughmachinelearning’,April13,2018,https://www.opengovasia.com/enhancing-customer-journeys-and-improving-fraud-detection-through-machine-learning/

9) SoumikRoy,‘Howartificialintelligenceisfightingfinancialcrime’,June17,2019,https://www.fintechnews.org/how-artificial-intelligence-is-fighting-financial-crime/

10) SIswaran,‘SingaporeStatementbyMrSIswaran,MinisterforCommunicationsandInformation,attheG20DigitalEconomyMinistersMeeting’,July22,2020,https://www.mci.gov.sg/pressroom/news-and-stories/pressroom/2020/7/singapore-statement-by-minister-iswaran-at-the-g20-digital-economy-ministers-meeting

11) EuropeanCommission,‘OnArtificialIntelligence-AEuropeanapproachtoexcellenceandtrust’February2020,https://ec.europa.eu/info/sites/info/files/commission-white-paper-artificial-intelligence-feb2020_en.pdf

12) RussellT.Vought,‘GuidanceforRegulationofArtificialIntelligenceApplications’,TheWhiteHouse,https://www.whitehouse.gov/wp-content/uploads/2020/01/Draft-OMB-Memo-on-Regulation-of-AI-1-7-19.pdf

13) TimAdams,AndresPortilla,MatthewEkberg,MichaelShepard,RobWainwright,KatieJackson,TamsinBauman,ChrisBostock,AbuSaleh,PabloSapiainsLagos,‘TheglobalframeworkforfightingfinancialcrimeEnhancingeffectiveness&improvingoutcomes’October2019,IIFandDeloitte,https://www2.deloitte.com/global/en/pages/financial-services/articles/gx-global-framework-for-fighting-financial-crime.html

14)MonetaryAuthorityofSingapore,‘MASintroducesnewFEATPrinciplestopromoteresponsibleuseofAIanddataanalytics’,November12,2018,https://www.mas.gov.sg/news/media-releases/2018/mas-introduces-new-feat-principles-to-promote-responsible-use-of-ai-and-data-analytics#:~:text=The%20Monetary%20Authority%20of%20Singapore,and%20data%20analytics%20in%20finance

15)MonetaryAuthorityofSingapore,‘“FairnessMetrics”toAidResponsibleAIAdoptioninFinancialServices’,May28,2020,https://www.mas.gov.sg/news/media-releases/2020/fairness-metrics-to-aid-responsible-ai-adoption-in-financial-services

16)MonetaryAuthorityofSingapore,‘MASPartnersFinancialIndustrytoCreateFrameworkforResponsibleUseofAI’,November13,2019,https://www.mas.gov.sg/news/media-releases/2019/mas-partners-financial-industry-to-create-framework-for-responsible-use-of-ai

17) EricCharran,SteveSweetman,‘AIMaturityandOrganizations–UnderstandingAIMaturity’Microsoft,https://www.bastagroup.nl/wp-content/uploads/2019/01/AI-Maturity-and-Organizations-eBook.pdf

18) SvetlanaSicular,BernElliot,WhitAndrews,PieterdenHamer,‘ArtificialIntelligenceMaturityModel’,March2020,Gartner,https://www.gartner.com/guest/purchase/registration?resId=3885363

19) Deloitte,‘IIFandDeloitteWhitePaperOutlinesNeededReformstoImprovetheGlobalFrameworkforFightingFinancialCrime,October16,2019,https://www2.deloitte.com/global/en/pages/about-deloitte/press-releases/iif-deloitte-paper-on-fighting-financial-crime-pr.html


35


34

Contact us

Radish SinghFinancial Crime Compliance Leader and AML PartnerFinancial AdvisoryDeloitte Southeast Asia

[email protected]

Nicholas Alvin SebastianDirectorFinancial AdvisoryDeloitte Southeast Asia

[email protected]

Nick Lim HeadofAI,Analytics&AutomationGroup ComplianceUnitedOverseasBank

[email protected]

Eric AngHeadofComplianceAnalytics&InsightsGroup ComplianceUnitedOverseasBank

[email protected]

http://[email protected]

Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited (“DTTL”), its global network of member firms, and their related entities (collectively, the “Deloitte organization”). DTTL (also referred to as “Deloitte Global”) and each of its member firms and related entities are legally separate and independent entities, which cannot obligate or bind each other in respect of third parties. DTTL and each DTTL member firm and related entity is liable only for its own acts and omissions, and not those of each other. DTTL does not provide services to clients. Please see www.deloitte.com/about to learn more.

Deloitte Asia Pacific Limited is a company limited by guarantee and a member firm of DTTL. Members of Deloitte Asia Pacific Limited and their related entities, each of which are separate and independent legal entities, provide services from more than 100 cities across the region, including Auckland, Bangkok, Beijing, Hanoi, Hong Kong, Jakarta, Kuala Lumpur, Manila, Melbourne, Osaka, Seoul, Shanghai, Singapore, Sydney, Taipei and Tokyo.

This communication contains general information only, and none of Deloitte Touche Tohmatsu Limited (“DTTL”), its global network of member firms or their related entities (collectively, the “Deloitte organization”) is, by means of this communication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser.

No representations, warranties or undertakings (express or implied) are given as to the accuracy or completeness of the information in this communication, and none of DTTL, its member firms, related entities, employees or agents shall be liable or responsible for any loss or damage whatsoever arising directly or indirectly in connection with any person relying on this communication. DTTL and each of its member firms, and their related entities, are legally separate and independent entities.

© 2020 Deloitte Southeast Asia Ltd

Date post:	01-Feb-2021
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Adanced analtics and innoation in inancial rime omliance...helped to lessen the impact of COVID-19...

Documents