Insights

Producedinpartnershipwith

boom

TomeettheirAIambitions,

organizationsmustshiftfrompilotsandexperimentstoenterprise-widedeployment.

Aplaybookfor

craftingAIstrategy

2MITTechnologyReviewInsights

Preface

“AplaybookforcraftingAIstrategy”isanMITTechnologyReviewInsightsreport

sponsoredbyBoomi.Toproducethisreport,MITTechnologyReviewInsights

conductedaglobalsurveyofC-suiteandseniordataexecutivesacrosscountries

andindustries.Thereportalsodrawsonin-depthinterviewsconductedwithbusinessleadersondataandAI.

AdamGreenwastheauthorofthereport,TeresaElseywastheeditor,andNicolaCrepaldiwasthepublisher.Theresearchiseditoriallyindependent,andtheviewsexpressedarethoseofMITTechnologyReviewInsights.

Wewouldliketothankthefollowingexecutivesandexpertsfortheirtimeandinsights:KevinCollins,FounderandChiefExecutiveOfficer,CharliAI

AmyMachado,SeniorResearchManager,IDCMattMcLarty,ChiefTechnologyOfficer,Boomi

SPSingh,SeniorVicePresidentandGlobalHead,EnterpriseApplicationIntegrationandServices,Infosys

Aboutthesurvey

ThesurveyformingthebasisofthisreportwasconductedbyMITTechnologyReview

InsightsinMarch2024.Thesurveysampleconsistsof205executivesanddataand

technologyleaders.Elevenindustriesarerepresented:financialservices,manufacturing,ITandtelecommunications,consumergoodsandretail,pharmaceuticalandhealthcare,government,travelandhospitality,professionalservices,energyandutilities,transportandlogistics,andmediaandmarketing.

Nearlyallsurveyrespondents(88%)comefromtheC-suite.Theseincludechief

executiveofficers(20%),chiefinformationofficers(18%),chieftechnologyofficers

(19%),andchiefdataofficers(15%).Therespondents’organizationsareheadquarteredinNorthAmerica(31%);Europe,theMiddleEast,andAfrica(25%);Asia-Pacific(26%);andCentralandSouthAmerica(18%).AllrespondentsworkatorganizationswithmorethanUS$500millioninglobalannualrevenue,with73%representingorganizations

generatingmorethanUS$1billion,and34%morethanUS$10billion.

CONTENTS

MITTechnologyReviewInsights3

01Executivesummary 4

02Partneringforsuccess 6

Selectingavendor 7

Finance-friendlyAI 7

03Countingthecost 8

Spendingexpectations 9

Measuringreturnoninvestment 10

04Buildingadatacore 11

Datamanagement:Tipsandtactics 11

Reckoningwithlegacyinfrastructure 13

Datalineageandliquidity 13

Metadata 13

05Accelerationversuscaution 14

Hallucinations,errors,andbias 14

Cyberrisk 14

Dataprivacyandprotection 14

Risingregulatorytide 15

Compliancechallenges 16

06Conclusion 17

4MITTechnologyReviewInsights

Executivesummary

G

iddypredictionsaboutAI,fromits

contributionstoeconomicgrowthtotheonsetofmassautomation,arenowasfrequentasthereleaseofpowerfulnewgenerativeAImodels.Theconsultancy

PwC,forexample,predictsthatAIcouldboostglobal

grossdomesticproduct(GDP)14%by2030,generatingUS$15.7trillion1

Fortypercentofourmundanetaskscouldbeautomatedbythen,claimresearchersattheUniversityofOxford,

whileGoldmanSachsforecastsUS$200billioninAIinvestmentby2025.2,3“Nojob,nofunctionwillremainuntouchedbyAI,”saysSPSingh,seniorvicepresidentandglobalhead,enterpriseapplicationintegrationandservices,attechnologycompanyInfosys.

Whiletheseprognosticationsmayprovetrue,today’sbusinessesarefindingmajorhurdleswhentheyseektograduatefrompilotsandexperimentstoenterprise-wideAIdeployment.Just5.4%ofUSbusinesses,for

example,wereusingAItoproduceaproductorservicein2024.4

MovingfrominitialforaysintoAIuse,suchascode

generationandcustomerservice,tofirm-wide

integrationdependsonstrategicandorganizational

transitionsininfrastructure,datagovernance,and

supplierecosystems.Aswell,organizationsmustweighuncertaintiesaboutdevelopmentsinAIperformance

andhowtomeasurereturnoninvestment.

IforganizationsseektoscaleAIacrossthebusinessincomingyears,however,nowisthetimetoact.

Thisreportexploresthecurrentstateofenterprise

“Nojob,nofunctionwillremainuntouchedbyAI.”

SPSingh,SeniorVicePresidentandGlobalHead,EnterpriseApplicationIntegrationandServices,Infosys

AIadoptionandoffersaplaybookforcraftinganAI

strategy,helpingbusinessleadersbridgethechasmbetweenambitionandexecution.Keyfindingsincludethefollowing:

AIambitionsaresubstantial,butfewhavescaled

beyondpilots.Fully95%ofcompaniessurveyed

arealreadyusingAIand99%expecttointhefuture.Butfeworganizationshavegraduatedbeyondpilot

projects:76%havedeployedAIinjustonetothree

usecases.ButbecausehalfofcompaniesexpecttofullydeployAIacrossallbusinessfunctionswithintwoyears,thisyeariskeytoestablishingfoundationsforenterprise-wideAI.

AIreadinessspendingisslatedtorisesignificantly.Overall,AIspendingin2022and2023wasmodest

orflatformostcompanies,withonlyoneinfour

increasingtheirspendingbymorethanaquarter.Thatissettochangein2024,withnineintenrespondentsexpectingtoincreaseAIspendingondatareadiness(includingplatformmodernization,cloudmigration,

anddataquality)andinadjacentareaslikestrategy,culturalchange,andbusinessmodels.Fourinten

expecttoincreasespendingby10to24%,andone-thirdexpecttoincreasespendingby25to49%.

MITTechnologyReviewInsights5

DataliquidityisoneofthemostimportantattributesforAIdeployment.Theabilitytoseamlesslyaccess,combine,andanalyzedatafromvarioussources

enablesfirmstoextractrelevantinformationandapplyiteffectivelytospecificbusinessscenarios.

Italsoeliminatestheneedtosiftthroughvastdata

repositories,asthedataisalreadycuratedandtailoredtothetaskathand.

DataqualityisamajorlimitationforAIdeployment.Halfofrespondentscitedataqualityasthemost

limitingdataissueindeployment.Thisisespecially

trueforlargerfirmswithmoredataandsubstantial

investmentsinlegacyITinfrastructure.Companies

withrevenuesofoverUS$10billionarethemostlikelytocitebothdataqualityanddatainfrastructureas

limiters,suggestingthatorganizationspresidingoverlargerdatarepositoriesfindtheproblemsubstantiallyharder.

Governance,security,andprivacyarethebiggest

brakeonthespeedofAIdeployment,citedby45%ofrespondents.

CompaniesarenotrushingintoAI.Nearlyall

organizations(98%)saytheyarewillingtoforgobeingthefirsttouseAIifthatensurestheydeliveritsafely

andsecurely.Governance,security,andprivacyarethebiggestbrakeonthespeedofAIdeployment,citedby45%ofrespondents(andafull65%ofrespondents

fromthelargestcompanies).

6MITTechnologyReviewInsights

Partneringforsuccess

F

ewcompaniesareflyingsolointheAIage.Thecostandcomplexityofcreatinglargelanguagemodels(LLMs)andgenerativemodelsfrom

scratchisprohibitiveandanabundanceof

platformsandtoolsarealreadyonthemarket.

“Formostorganizations,buildingtheirownlargelanguagemodelsisveryexpensiveandthevalueistime-limited,”

saysKevinCollins,founderandCEOofCharliAI,anAI

solutionsprovider.“Ifyoudon’thavethenecessary

expertiseandresourcestomakeasignificantinvestment,it’sbettertofine-tuneandoptimizeoff-the-shelfmodels.”

AIisalsoincreasinglybeingintegratedintoexistingsoftwareplatforms,includingthosefromgiant

technologyproviderssuchasMicrosoftandAdobe.

“WeareinfusingAIintoeverythingwedowithinour

organizationandsimilarlytakingthatsamemindsettoourclients,”saysSPSinghofInfosys.Companies,ofcourse,canalsopaytouseproprietarymodelsfromproviderssuchasOpenAIorfine-tunethosemodelsasneeded.TheycanalsobuildtheirowngenerativeAItoolsusingopen-sourcemodels.

Figure1:CompaniesseekAIusecasestailoredtoindustryneeds

IsyourorganizationdevelopingthefollowingtypesofAIusecases?

(OrganizationsthataredevelopingAIusecases.)

Usecasesthatarecommonacrossindustries(e.g.,generativeAIfortextgenerationorchatbots)

Usecasesthatarespeci?ctoourindustry(e.g.,AIfordrugdiscovery)

Usecasesthatareuniquetoourbusiness

OverallCompanieswithannualrevenueof$500millionto$1billion

57%

77%

38%

Companieswithannualrevenueof$1billionto$10billionCompanieswithannualrevenuemorethan$10billion

48%

75%

77%

Source:MITTechnologyReviewInsightssurvey,2024

84%

63%

53%

52%

79%

61%

MITTechnologyReviewInsights7

MattMcLarty,CTOatthesoftwarecompanyBoomi,

concursthatbuildingLLMswillnotbethepathforwardformostcompanies.“Mostorganizationsarenotgoingtobuildlargelanguagemodels,whichareexpensive

andrequireamassiveamountofinfrastructure,”

hesays.“Instead,theyneedtobecomeexpertsinapplyingthenewtechnologiesintheirownbusinesscontext.Thecompaniesthatdon’tputtheAIcart

beforethebusinessproblemhorsearegoingtobebetterpositioned.”

Oursurveyresultssuggestthatbusinessesaretakingthatadvice,seekingAIusecasesthataddresstheir

uniquebusinessproblems.Whilemanyfirmsare

deployinggeneral-purposeusesofAI(generativeAI–poweredchatbotsforcustomerservice,forexample),alargersharesaytheyaredevelopingindustry-specificusecases,orevenusecasesuniquetotheirparticularbusiness(seeFigure1).

Sectoralnuances—includingdifferingdataformats,

regulatoryrequirements,andusages—supportthecaseforindustry-tailoredapproachestoAI.FromAItools

designedtodocumentpatient-doctorconversationstothoseleveragedtooptimizedrywallfinishing,offeringstailoredtotheneedsofspecificindustriesorbusinessesaremostlikelytooffergame-changingresults.5,6

Selectingavendor

CompanieshaveplentyofAIvendorstochoosefrom;ifanything,thechoicecanbedauntingforexecutiveswhomustbalancecost,security,anddiversification.

“Itfeelsoverwhelmingforcompanies—thereisalotofnoiseandquestionsaroundwhototrust,”saysAmyMachado,seniorresearchmanageratIDC,amarketintelligencefirm.

Forprivacyreasons,manyarewaryofrelyingon

publiclyavailableAItoolsthatmayretainandreuse

thedatausersenter,withsomecompaniesrestrictingemployeeuseofcertainLLMs.7,8“Ifyou’reshuffling

yourdataofftoageneral-purposeAI,youhavetodoyourduediligenceonhowthey’retreatingyourdata

andwhatthey’redoingwithitbehindthescenes,”saysCollins.MorespecializedLLM-basedtoolsaremore

likelytoprotectorganizations’proprietarydata,comewithsupport,andbeconsistentlyupdatedtoaddressvulnerabilities.

Ultimately,executivesmayfindthatstitchingtogethera“multi-AI”environmentbestmeetstheirbusiness

needsandcapturesthedistinctivevaluepropositionsofdifferentproviders.

Financialservicesfirmsoperateinahighlyregulated“NewAIhastheabilitytounderstand‘locked-up’

Finance-friendlyAI

environmentwithhighstandardsfordataprivacy,contentfarbetterthaneverbefore,sowe’reunlocking

compliance,andsecurity.“Inthefinancialservicesagoldmine,”saysCollins.“Ourcustomersrefertous

sector,trustandsecurityareparamount,”says

KevinCollinsatCharliAI.“Ourcustomersaremajor

astheautomatedanalystsforWallStreet—wecandoin30minuteswhatitmaytakeananalyst80hours

ofporingovera500SECfilingtodo.”

investmentbanks,hedgefunds,andbigbanksthatareanalyzingS&P500companies.Everythingwedoneedstobetrustedandfact-checked.”

Infinancialservices,valuableinformationisoftenlockedawayincontracts,legaldocuments,PDF

files,spreadsheets,andPowerPointpresentations.

Whilecompaniesareawareofthepotentialvalueoftheirdata,manystruggletoleverageiteffectively.“They’reverydataliterate,buttheyalsounderstandhowchallengingitistogetthatdataintoaformatthat’ssuitableforanalyticsandAI,”saysCollins.

Extractingandintegratingthisdataforanalysiscan

beadauntingtask.SpecializedserviceslikeCharliOursurveyfoundthatfinancialservicesrespondents

AIhaveopenednewpossibilitiesforunlockingthiswereleastlikelytobecurrentlydevelopingAIfor

trappeddata:thecompany’stoolsmineandcurateindustry-specificusecases,butcompanieslike

arangeoffinancialdata,freeinguptimeforanalystsCharliareaimingtochangethis.

toperformhigher-valuetasks.

8MITTechnologyReviewInsights

Countingthecost

F

romtheelectricityrequirementsforbuildingandrunningAImodelstothesoftware,

services,andsalariesneededtomanage

solutionsatscale,costisamajor

considerationforAI.Google’sUS$8billionof

spendinginthethirdquarterof2023alonewas

overwhelminglydrivenbyAI.9TrainingOpenAI’sGPT-4requiredUS$78millionofcompute,whileGoogle’s

GeminiUltrahadacomputecostofUS$191million10

Hardwareisamajorsourceofpriceuncertainty,

includingdemandforgraphicsprocessingunits

(GPUs).“AIisveryexpensive,notjustforthe

developmentandtrainingofmodels,butalsoforthe

operationsandhardwarecosts,particularlyGPUs,”

saysCollins.NewerGPUsaremoreexpensive,hardertoobtain,andhavealimitedlifespanrequiringa

recurringcapitalcost.Headds,“AIisnotcheap.A

lotofourcustomersareshockedwhentheyrealizehowmuchinfrastructuretheyhavetohavetogettheperformancethattheyneed.”

Whiletechcompaniesarelargelyshouldering

hardwareandcomputationspendingintheirracefor

marketsupremacy,firmsinevitablyfacetheirown

costsinareaslikedatamanagementandmonitoring,

whicharerarelybuiltintolower-costgeneral-purposeLLMs.Theymustalsoinvesttoensurecompliancewith

Figure2:ChangesinAIreadinessspendingfrom2022to2023

Howdidyourorganization’sinvestmentinAIreadinesschangeinthelastyear(from2022to2023)?

Data-relatedinvestments

(platformmodernization,cloudmigration,pipelines,quality,etc.)

1%

6%

Increasedby25%一49%

Increasedby10%一24%StayednearlythesameDecreased

34%42%

Source:MITTechnologyReviewInsightssurvey,2024

Investmentsinotherareas

(developingAIstrategy,culturalchange,buildingmodels,etc.)

Increasedbymorethan100%Increasedby50%一100%

40%

20%

34%

18%

6%

1%

MITTechnologyReviewInsights9

Figure3:LargecompaniesboostedAIspendingmoreaggressively

Companieswithrevenuegreaterthan$10billion:Howdidyourorganization’sinvestmentinAIreadinesschangeinthelastyear(from2022to2023)?

Data-relatedinvestments

(platformmodernization,cloudmigration,pipelines,quality,etc.)

1%

11%

Increasedby50%一100%

35%

26%

42%

27%

Source:MITTechnologyReviewInsightssurvey,2024

Investmentsinotherareas

(developingAIstrategy,culturalchange,buildingmodels,etc.)

Increasedby25%一49%

Increasedby10%一24%

StayednearlythesameDecreased

Increasedbymorethan100%

20%

29%

9%

1%

firm-specificregulations,fundmaintenancetoensurethesystemstheybuildremainup-to-dateandrelevanttotheusecase,andpayfortheincreasedenergy

usedbyAI11TalentcostsincludeeitherhiringAI-skilledworkersorupskillingtheexistingworkforce.

Spendingexpectations

Pollsshowthatlargerfirms,byrevenue,aremore

likelytohaveAIinproduction,suggestingthatfinancialconstraintsmaybeslowingsmallerandmedium-sizedcompanies12AnalysisfromCCSInsightssuggestscostwillcontributetoaslowdowningenerativeAIadoptionthroughout202413Asthehypecools,decision-makersmaybecomemorewary,especiallyastheproductivitygainsofgenerativeAImaynotbeapparentforyears14

Suchfindingsarecorroboratedbyoursurvey,whichfoundalinkbetweencompanysizeandspending.

Budgetaryconstraintswerethebiggestobstacleto

speedofAIdeploymentformedium-sizedfirms(annualrevenueofUS$500millionto$1billion),with47%

citingthisissuecomparedtoasurveyaverageof22%.Thisindicatesa“squeezedmiddle”ofcompaniesthatneedtoadoptAItostaycompetitiveintheirmarketsbutforwhomthecostsareprohibitive.

Todate,mostfirmshavekeptAIreadinessspendingflat,withlargercompaniesmorelikelytohaverampedup(seeFigures2and3).

However,nearlyallcompaniesexpecttoboost

spendinginthecomingyear,with9in10expecting

toincreaseinvestmentbyatleast10%andonethird

anticipatingspendingupto49%morebothindataandadjacentareaslikestrategyandculturalchange(seeFigure4).

“AIisnotcheap.Alotofourcustomersareshockedwhentheyrealizehowmuchinfrastructuretheyhavetohavetogettheperformancethattheyneed.”

KevinCollins,FounderandChiefExecutiveOfficer,CharliAI

0

10MITTechnologyReviewInsights

Figure4:AIreadinessspendingsettorisesignificantlyinto2024

Howwillyourorganization’sinvestmentinAIreadinesschangeinthenextyear(from2023to2024)?

Investmentsinotherareas

(developingAIstrategy,culturalchange,buildingmodels,etc.)

Data-relatedinvestments

(platformmodernization,cloudmigration,pipelines,quality,etc.)

1%

1%

Willincreasebymorethan100%

9%

16%

10%11%

35%

37%

Willincreaseby50%一100%Willincreaseby25%一49%Willincreaseby10%一24%Willstaynearlythesame

41%

Willdecrease

40%

Source:MITTechnologyReviewInsightssurvey,2024

AscompaniesdialintheirAIinvestments,AIcost

dynamicswillofcoursealsoevolve.Innovationwill

doubtlessbringdowncoststhroughefficiencygains.

Thecostcurveforcomputationwillalsofallasstartupsandlargecompaniesaddresstheshortcomingsof

existinghardwareandmodels.

Measuringreturnoninvestment

ToweighthefinancialoutlaysofAI,organizations

needtodeveloprobustreturnoninvestment(ROI)

methodologiescapturingnotjusttheefficiencygainsofautomating“businessasusual”tasks,butalsothenewvalueAIcancreate.

CompaniesarealreadyquantifyingAI’scostsavings.

MotorolahasdevelopedaframeworkthattrackshourstakentocompleteataskwithandwithoutgenerativeAI15SwedishfintechcompanyKlarnahasalready

calculatedgenerativeAI’soutputasequivalentto700customerserviceagents16But“themindsetisshiftingtousingAIasanenablerforrevenuegrowth,notjustcostsavings,”saysMachado.

Asemployeesbecomemoreproductive,new

opportunitiesemergethatmaybehardtoquantify.Forexample,AIallowsdatascientiststoexplorenewideasandexpandintonewproducts.Theautomationof

routinetasksbycopilotscanalleviateworkloadsand

“ThemindsetisshiftingtousingAIasanenablerforrevenuegrowth,notjustcostsavings.”

AmyMachado,SeniorResearchManager,IDC

reduceburnout,allowingemployeestodedicatemoretimetomorecreativeandinnovativeprojects.

AsenterpriseuseofAIspreads,employeeexpectationswillshiftandAI-enhancedworkplaceexperienceswill

becomenecessarytoattracttalent.AIprofessionals

alreadyhavedifferentexpectationsthantraditionaljobseekers,rankinginterestingjobcontentandworkon

cutting-edgeprojectshigheramongtheirjobpriorities17

TheconsultancyPwCcompares“hardROI,”suchastimesavedandproductivityboosts,with“softROI,”

includingimprovedemployeeexperienceandability

torespondflexiblytonewopportunities18AccuratelyassessingAI’sROIrequiresmetricsbeyondtraditionalfinancialmeasures.Forinstance,Collinspointsout

thattoolslikeCharliAI’sunlockhugevaluebyhelpingcompaniesmakesenseoftheirdatamoreefficiently.

MITTechnologyReviewInsights11

Buildingadatacore

D

?Trackandmanagethelineageofyour

organization’sdatatomaintaindataqualityandintegrityasitmovesfromoneanalyticalmodel

ataquality,infrastructure,andgovernanceareallessentialtodeliverAIworkloads

efficientlyandatscale.WhileAI

technologycanbetransformative,withoutthedatafoundationsinplace,

organizationswillstruggletobothcollecttherightdataandderiveinsightfromit.“Peoplestillhavethe

perceptionthatAItodayismagicorsuperintelligent,

andthat’sfarfromthetruth,”saysCollins.“AIisa

scienceandatool.Youstillhavetodoallofthehard

workarounddatagovernanceandfiguringouttherightapplicationofthesetools.”

Organizationsthatoverlookthesesupporting

dimensionsmaystruggletodeployAIsuccessfully.

Dataqualityproblems,citedbyhalfofrespondentsastheirmostlimitingdataissue,canseverelyhinderAIperformance.“Goingforward,theconversation

willbearoundgettingdataAI-readyandintherightformattoactuallyutilizeiteffectively,”saysMachado.Datainfrastructureorpipelines,dataintegration

tools,andcloudmigrationwerealsomentionedby

manyrespondentsasmajorbarrierstodeployment,indicatingtheimportanceofarobustITarchitecture.

TheseissuesareparticularlyprominentforcompanieswithrevenuesofoverUS$10billion,whichare

themostlikelytocitebothdataquality(52%)and

datainfrastructure(55%)asobstacles,compared

withoverallsurveyaveragesof49%and44%(see

Figure5).Organizationspresidingoverlargerdata

repositoriesandlegacyITinfrastructuremaybe

findinggreatercomplexityandcostsintransitioningtoanAI-readyarchitecture.

Datamanagement:Tipsandtactics

orapplicationtothenext.

?Optimizecommunicationacrossthe

organizationonAI,andondataaggregationandrequirements,ensuringcross-functionalteamscollaborateonbusinesscases.

?Insteadoffocusingondatacentralization,adoptastrategythatprioritizesdatacontextualization,suchasextractingmetadataandstoringitinanappropriatemetadatalanguage.

?Identifyyourorganization’scorecapabilities

andcompetencies,ensuringyouareabletoadapttorapidchangesandarenotoverly

reliantonlegacyinfrastructure.

“AIisascienceandatool.Youstillhavetodoallofthehardworkarounddata

governanceandfiguringouttherightapplicationof

thesetools.”

KevinCollins,FounderandChiefExecutiveOfficer,CharliAI

12MITTechnologyReviewInsights

Figure5:DatabottlenecksinAIdeployment

Whichaspectsofyourorganization,sdataaremostlimitingthespeedtodeployAI?

Overall

Companieswithannualrevenueof$500millionto$1billion

Companieswithannualrevenueof$1billionto$10billion

Companieswithannualrevenuemorethan$10billion

Dataquality

49%

38%

38%

52%

Datainfrastructureorpipelines

44%

26%

49%

55%

Dataintegrationtools

40%

27%

41%

50%

Incompletecloudmigrationstatus

36%

26%

41%

40%

Organizationalcultureorapproachtodata

26%

25%

34%

18%

Dataarchitectureorplatform

24%

38%

17%

18%

Regulatoryconstraintsondatahandling/use

21%

38%

10%

18%

Source:MITTechnologyReviewInsightssurvey,2024

MITTechnologyReviewInsights13

Reckoningwithlegacyinfrastructure

TransitioningawayfromlegacyITinfrastructurecanbedauntingfororganizationsofallsizes.“Legacy

systemsandsunkcostsaredefinitelyaninhibitor

forallkindsofdatamanagementinitiatives,”says

Machado.“Manyorganizationsdon’twanttotouchcertainlegacysystemsbecause‘it’skindofworking’andonlyonepersonknowshowtomaintainthat

oldertechnology.”Theselegacysystemsareoftenbasedonfragmented,outdatedarchitecturesandprogramminglanguagesthatarehardtointegratewithmodernAIsolutions.

ToreapthebenefitsofAI,organizationsmustdevelopanITarchitecturecapableofaccommodatingboth

structuredandunstructureddataacrosstheentire

lifecycle,fromthesourcetoprocessing,analytics,

andstorage.UnstructureddatacanyieldconsiderableinsightforAIsystems,butitsrawinformationcanbe

difficulttoassimilateintoexistingsystems.Thatcanbeproblematicsince,accordingtoMachado,90%ofthedatainanenterprisecontentmanagementsystemis

unstructured.

Acentralizedrepositorysolutionintheformofadatalakecanbeuseful,butitmustbewellorganizedandstructuredtopreventitbecomingfilledwithirrelevantandunusabledata.“It’snotenoughtojusthaveabigdatalake,becausethat’sgoingtobecomeadata

swamp;youhavetohavegoodgovernanceoverthedatalineageandhowitflowsfromonemodeltothenext,”saysCollins.

Nearlyhalf(46%)ofsurveyrespondentsalsocite

manual,non-digitized,ornon-automatedprocesses

fordatahandlingandmanagementasalimitingfactorforAIdeployment.Withoutstrategiestofullydigitizeprocessesacrossthebusiness,organizationswill

leavevaluabledatasourcesuntappedandlimitAI’sautomationpotential.

Datalineageandliquidity

Decidingwhichdatatoleveragerequiresittobe

recombinedandcontextualized.Implementingdatalineagetrackingandquality-assurancemeasurescansupportthiscontextualizationprocess,preserving

theintegrityofdataasitundergoesprocessingandanalysis.

AccordingtoMcLarty,dataliquidity,ortheabilitytoseamlesslyaccess,combine,andanalyzedatafromvarioussources,isoneofanorganization’smostimportantAIassets.Dataliquidityempowersorganizationstounderstandtherelationships,

dependencies,andnuanceswithintheirdata,

enablingthemtoextractrelevantinformationandapplyiteffectivelytospecificbusinessscenarios.

“Organizationswithhighdataliquidity