CRITICAL THINKING ON FORENSIC ACCOUNTING AND AI-BASED FRAUD ANALYTICS

CRITICAL iTHINKING iON iFORENSIC iACCOUNTING iAND iAI-BASED iFRAUD iANALYTICS

Prof iA.Seshachalam, iDepartment iof iCommerce,

Ms iSukitha iNagraj iAdaki i– iMBA iPGCET

Ms iBhavitha iKinnera i– iMBA iPGCET

IFIM iCollege i( iAutonomous), iBengaluru: i560 i100.

1. iINTRODUCTION

Corporate ifraud irepresents ia imajor ithreat ito iinvestors, iemployees, icreditors, igovernments iand ifinancial imarkets. iTraditional iauditing iprimarily iprovides ireasonable iassurance iregarding ifinancial istatements, iwhereas iforensic iaccounting igoes ifurther iby iinvestigating isuspicious itransactions iand ideveloping ievidence isuitable ifor idisciplinary, iregulatory ior ilegal iproceedings.

The iincreasing ivolume iand icomplexity iof ifinancial iinformation ihave icreated ia istrong icase ifor itechnology-assisted iforensic iinvestigation. iAI ican ianalyse imillions iof itransactions, iidentify iabnormal ipatterns, iexamine itextual iinformation iand iconnect iapparently iunrelated ientities. i

A irecent isystematic ireview iof i76 istudies ipublished ibetween i2019 iand i2026 ifound iincreasing iacademic iinterest iin iAI iand imachine ilearning ifor iforensic iaccounting, iwith isupervised ilearning, iensemble imethods, ideep ilearning iand inatural ilanguage iprocessing iemerging ias iimportant itechniques. iThe icritical iquestion, itherefore, iis inot isimply i“Can iAI idetect ifraud?”, ibut irather i“Can iAI idetect ifraud ireliably, iexplainably iand ilegally, iwhile ipreserving ihuman iprofessional ijudgement?”

2. iREVIEW iOF iLITERATURE

Forensic iaccounting ihas itraditionally irelied iupon iratio ianalysis, iBenford's iLaw, itransaction itesting, iinterviews, idocument iexamination iand iinvestigative iaccounting. iThe iemergence iof ibig-data ianalytics ihas iexpanded ithese itechniques itoward icontinuous iand ipredictive ifraud idetection. iRecent iliterature ishows ia isubstantial iincrease iin iresearch iconnecting iAI iwith iforensic iaccounting. iA i2025 ibibliometric istudy iof i261 iScopus-indexed ipublications iidentified iforensic iaccounting, ifraud, iAI, iblockchain, ibig idata iand idata ianalytics ias iincreasingly iinterconnected iresearch ithemes. i

Guo iand iTang i(2025) iemphasise ithat iAI ican iimprove idetection iaccuracy, ireal-time imonitoring iand ianalysis iof iunstructured iinformation, iwhile ialso ihighlighting idata-quality, iexplainability iand iethical ichallenges. i

Similarly, iQafisheh i(2026) iargues ithat iAI ican iprovide ipredictive iaccuracy, iearly-warning isystems iand icontinuous imonitoring, ibut iremains ia icomplement irather ithan ia isubstitute ifor iprofessional iforensic iexpertise. i

Recent isystematic ievidence ifurther isuggests ithat iAI's ieffectiveness idepends ion idata iquality, imodel idesign, ifraud itype, ievaluation imethodology iand iinstitutional icontext. iThus, ithe iliterature ibroadly isupports ian iaugmented iintelligence iapproach, irather ithan ia icompletely iautonomous ifraud-investigation imodel.

3. iOBJECTIVES iOF iTHE iSTUDY

The istudy iaims ito:

• Examine ithe irole iof iAI-based ianalytics iin imodern iforensic iaccounting. i

• Analyse ihow iAI ican istrengthen ifraud idetection iand iinvestigation. i

• Critically iexamine ilessons ifrom ithree imajor icorporate ifraud icases. i

• Identify ilimitations iassociated iwith ialgorithmic ifraud idetection. i

• Propose ia ihuman–AI iframework ifor ieffective iforensic iaccounting. i

• Suggest ifuture iresearch idirections iin iAI-enabled ifraud ianalytics. i

4. iRESEARCH iGAP

Although iliterature ion iAI iand ifraud idetection iis iexpanding irapidly, iseveral igaps iremain. iFirst, imany istudies iconcentrate ion itechnical imodel iaccuracy irather ithan ithe icomplete iforensic-investigation iprocess. iSecond, irelatively ilimited iattention iis igiven ito iwhether iAI-generated ialerts ican ibe itranslated iinto ilegally idefensible ievidence. iThird, iexplainability iremains iproblematic ibecause isophisticated imodels imay iidentify ianomalies iwithout iclearly iexplaining iwhy ia itransaction iis isuspicious.

Fourth, ithere iis iinsufficient iintegration iof iprofessional iscepticism, iinvestigative ijudgement iand iAI ioutputs. iRecent iliterature ispecifically iidentifies iinterpretability, icontextual igeneralisability, ilegal iadmissibility iand iinterdisciplinary iintegration ias icontinuing igaps. iThis iresearch itherefore iadopts ia icritical iperspective: ihigh ipredictive iaccuracy ialone icannot iestablish ifraud.

5. iMETHODOLOGY iAND iLIMITATIONS

The istudy ifollows ia iqualitative, iexploratory iand icomparative icase-study iapproach, iusing isecondary iinformation ifrom iacademic iliterature, iregulatory ipublications iand idocumented icorporate-fraud iinvestigations.

The ithree icases iselected iare iEnron iCorporation, iSatyam iComputer iServices iLtd. iand iWells iFargo i& iCompany ibecause ithey irepresent idifferent iforms iof icorporate imisconduct iand idemonstrate ihow isophisticated ianalytics icould ipotentially iassist iforensic iinvestigation.

LIMITATIONS

The istudy iis ibased iprimarily ion isecondary iinformation iand idoes inot itest ian iactual iAI ifraud-detection imodel iagainst iproprietary icompany itransaction idatasets. iConsequently, ithe ianalysis icannot iestablish ithe iprecise ipercentage iimprovement ithat iAI iwould ihave iachieved iin ieach ihistorical icase. iFurther, iretrospective iapplication iof imodern iAI ito ihistorical ifrauds iinvolves icounterfactual iassumptions.

6. iANALYSIS iAND iINTERPRETATION: iTHREE iMAJOR iCORPORATE iCASES

6.1 iENRON iCORPORATION

Enron's icollapse iin i2001 ibecame ione iof ithe imost iprominent iexamples iof iaccounting imanipulation. iInvestigators iidentified icomplex itransactions, imanipulated ifinancial ireporting iand ihidden iliabilities. iThe iFBI ireports ithat iinvestigators iultimately icollected imore ithan ifour iterabytes iof idigitised ievidence iand ianalysed ithousands iof idocuments, iwhile ifinancial ianalysts iexamined ibank iand ibrokerage iaccounts. i

AI-based iforensic iinterpretation: iModern igraph ianalytics icould ipotentially imap irelationships ibetween iEnron, ispecial-purpose ientities, iexecutives iand ifinancial iintermediaries. iNatural-language iprocessing icould ianalyse iemails iand icorporate icommunications ifor isuspicious ilanguage, iwhile ianomaly-detection ialgorithms icould iidentify iunusual itransactions iand ichanges iin ifinancial irelationships. iThe icritical ilesson iis ithat itransaction icomplexity iitself ican ibecome ia ifraud iindicator.

6.2 iSATYAM iCOMPUTER iSERVICES iLTD.

The iSatyam iscandal irepresents ione iof iIndia's imost isignificant icorporate-accounting ifraud icases. iSEBI's iinvestigation iidentified ifalsification iand imanipulation iof ifinancial istatements, iincluding iinflated irevenues iand iliabilities ithat iwere iunderstated. iSEBI ialso idocumented imismatches ibetween ibank iinformation iand iaccounting irecords iand iidentified itwo isets iof imanagement iinformation ireports icontaining idifferent ifigures. i

AI-based iforensic iinterpretation: iAn iAI-enabled iforensic isystem icould icompare ibank iconfirmations iwith iledger ibalances, iidentify iduplicate ior ifabricated iinvoices, ianalyse irevenue itrends, idetect iabnormal idebtor ibalances iand ireconcile imultiple iinformation isources iautomatically.

The icase iillustrates ithe iimportance iof iindependent idata iverification. iAI iis ivaluable ionly iwhen iit ireceives ireliable iunderlying iinformation; itherefore, iexternal iconfirmation iand idata igovernance iremain iessential.

6.3 iWELLS iFARGO i& iCOMPANY

The iWells iFargo icase idemonstrates ithat ifraud ianalytics imust iextend ibeyond iconventional ifinancial-statement ianalysis iinto ibehavioural iand ioperational ianalytics. iThe iSEC ifound ithat ithe ibank's ipublicly ireported icross-sell imetric iwas iinflated iby iunused, iunnecessary ior iunauthorised iaccounts. iWells iFargo iagreed ito ia i$500 imillion iSEC isettlement ias ipart iof ia icombined i$3 ibillion iresolution iwith ithe iSEC iand iDepartment iof iJustice. i

AI-based iforensic iinterpretation: iMachine ilearning icould ipotentially iidentify iunusual ipatterns isuch ias iexcessive iaccount iopenings iby iparticular iemployees, iunusually ihigh isales iperformance, irepeated icustomer icomplaints, idormant iaccounts iand igeographical ior iemployee-level ioutliers. iThe icritical ilesson iis ithat ifraud iindicators imay iexist ioutside ithe igeneral iledger. iEffective iforensic ianalytics imust itherefore iintegrate ifinancial, icustomer, iemployee iand ibehavioural idata.

7. iCOMPARATIVE iINTERPRETATION

The ithree icases idemonstrate ithat iAI icould ipotentially ishift iforensic iaccounting ifrom ireactive iinvestigation ito iproactive irisk idetection. iHowever, ian iAI ialert iis ionly ia istarting ipoint. iThe iforensic iaccountant imust iinvestigate ithe iunderlying ifacts, iestablish iintent iwhere irelevant, ipreserve ievidence iand idetermine iwhether ithe iconduct iconstitutes ifraud ior ianother iform iof imisconduct.

8. iSUGGESTIONS

• Develop iHuman–AI iHybrid iForensic iTeams: iAccountants, iauditors, idata iscientists, ilawyers iand icybersecurity ispecialists ishould iwork itogether. i

• Adopt iExplainable iAI: iFraud-detection isystems ishould iexplain iwhy ia itransaction ihas ibeen iclassified ias isuspicious. i

• Establish iContinuous iMonitoring: iOrganisations ishould imove ifrom iperiodic ireview itoward icontinuous itransaction imonitoring. i

• Integrate iMultiple iData iSources: iGeneral iledgers ishould ibe ianalysed ialongside ibank idata, iinvoices, iemails, icustomer icomplaints, iprocurement irecords iand ioperational iinformation. i

• Strengthen iData iGovernance: iPoor-quality idata ican igenerate ifalse ipositives ior iconceal iactual ifraud. i

• Create iAI iAudit iTrails: iEvery ialgorithmic idecision ishould ibe idocumented ito isupport ireview iand ipotential ilegal iproceedings. i

• Develop iProfessional iAI iCompetencies: iForensic iaccountants ishould iacquire iskills iin idata ianalytics, imachine ilearning, icybersecurity iand idigital ievidence. i

• Maintain iProfessional iScepticism: iAI-generated ialerts ishould inever iautomatically ibe itreated ias iproof iof ifraud. i

• Strengthen iCorporate iGovernance: iBoards iand iaudit icommittees ishould iperiodically iassess ithe ieffectiveness iof iAI-enabled ifraud-risk isystems. i

• Develop iRegulatory iGuidance: iProfessional iand iregulatory ibodies ishould iestablish istandards ifor iexplainability, ievidence ipreservation, iprivacy iand iadmissibility iof iAI-generated iforensic ievidence. i

9. iCONCLUSION

Forensic iaccounting iis ientering ia inew itechnological iphase iin iwhich iartificial iintelligence iand iadvanced ianalytics ican isignificantly iincrease ithe ispeed, ibreadth iand isophistication iof ifraud idetection. iEvidence ifrom irecent iresearch iindicates ithat imachine ilearning, ideep ilearning, inatural ilanguage iprocessing iand iother iAI imethods ican iidentify ipatterns ithat iconventional imanual iprocedures imay ioverlook. i

The icases iof iEnron, iSatyam iand iWells iFargo idemonstrate ithree iimportant idimensions iof icorporate ifraud: icomplex ifinancial irelationships, imanipulated iaccounting iinformation iand ibehavioural/operational ianomalies. iAI-based ianalytics icould ipotentially iidentify isignals iacross ieach iof ithese idimensions.

Nevertheless, icritical ithinking irequires irecognition iof iAI's ilimitations. iA imodel imay igenerate ia ifalse ipositive, iinherit ibias ifrom ihistorical idata, ifail ito iexplain iits ireasoning ior iidentify ian ianomaly iwithout iestablishing ifraudulent iintent. iTherefore, iAI ishould iaugment—not ireplace—the iforensic iaccountant.

The ifuture iof iforensic iaccounting ilies iin ia ihybrid imodel iwhere iAI iprovides ianalytical iintelligence iand ithe iforensic iprofessional iprovides icontextual, iethical, iinvestigative iand ilegal ijudgement. iSuch iintegration ican icontribute ito istronger icorporate igovernance, iimproved ifinancial ireporting iquality iand igreater iconfidence iamong iinvestors iand istakeholders.

10. iSCOPE iFOR iFURTHER iRESEARCH

Future istudies imay:

• Develop iand itest iexplainable iAI imodels ifor iIndian icorporate ifraud. i

• Compare imachine-learning ialgorithms ifor idetecting ifinancial-statement imanipulation. i

• Study iAI's irole iin idetecting iGST, ibanking iand isecurities ifraud. i

• Examine ithe iadmissibility iof iAI-generated ievidence iunder iIndian ilaw. i

• Develop iAI imodels iintegrating ifinancial iand inon-financial ifraud iindicators. i

• Study ithe ieffectiveness iof iAI iin idetecting iESG iand igreenwashing ifraud. i

• Investigate iblockchain-AI iintegration ifor icontinuous iforensic iauditing. i

• Examine ithe icompetencies irequired iby ifuture iforensic iaccountants. i

• Conduct iempirical iresearch iusing ianonymised icorporate itransaction idatasets. i

• Develop ian iIndian iForensic iAccounting iAI iFramework ifor ilisted icompanies iand ifinancial iinstitutions. i

REFERENCES

[1].Adejumo, iA. iP., i& iOgburie, iC. iP. i(2025). iForensic iaccounting iin ifinancial ifraud idetection: iTrends iand ichallenges. iInternational iJournal iof iScience iand iResearch iArchive, i14(3), i1219–1232. ihttps://doi.org/10.30574/ijsra.2025.14.3.0815 i

[2].Ghozali, iD. iM., i& iHayati, iN. i(2026). iArtificial iintelligence i(AI), iaudit, iforensic iaccounting: iLiterature ireview. iCompetitive: iJurnal iAkuntansi idan iKeuangan, i9(1). ihttps://doi.org/10.31000/competitive.v9i1.10039 i

[3].Guo, iH., i& iTang, iZ. i(2025). iArtificial iintelligence iin iforensic iaccounting: iA iliterature ireview. iJournal iof iAccounting iand iFinance, i25(3). i

[4].Qafisheh, iA. i(2026). iIntegrating iartificial iintelligence iinto iforensic iaccounting: iOpportunities, ilimitations, iand iimplications ifor ifinancial ifraud idetection. iHorizons iIntermediary iJournal iof iBusiness iResearch, i1(1). ihttps://doi.org/10.67150/hijbr.v1i1.7 i

[5].Suffah, iF., iSabila, iR. iF., iAlfani, iU. iN., i& iIndriyani, iI. i(2026). iArtificial iintelligence iin iforensic iaccounting: iA ibibliometric ianalysis iof iglobal iresearch itrends iand ifuture idirections. iEKOMA: iJurnal iEkonomi, iManajemen, iAkuntansi, i5(5). ihttps://doi.org/10.56799/ekoma.v5i5.20273 i

[6].Zakaria, iZ. iN. iZ., iHussain, iN. iH. iC., iAwang, iY., i& iShuhidan, iS. iM. i(2025). iSmart iforensics: iTracing ithe irise iof iartificial iintelligence iin iforensic iaccounting, ifraud idetection iand irisk imanagement. iInformation iManagement iand iBusiness iReview, i17(2(I)S), i51–68. ihttps://doi.org/10.22610/imbr.v17i2(I)S.4565 i

[7].Securities iand iExchange iCommission. i(2020). iWells iFargo ito ipay i$500 imillion ifor imisleading iinvestors iabout ithe isuccess iof iits ilargest ibusiness iunit. i

[8].Securities iand iExchange iCommission. i(2015). iOrder iin ithe imatter iof iSatyam iComputer iServices iLtd. i

[9].Federal iBureau iof iInvestigation. i(n.d.). iEnron. iU.S. iDepartment iof iJustice. i