英文【科技棱镜】2025聚焦科技引领的商业变革_第1页
英文【科技棱镜】2025聚焦科技引领的商业变革_第2页
英文【科技棱镜】2025聚焦科技引领的商业变革_第3页
英文【科技棱镜】2025聚焦科技引领的商业变革_第4页
英文【科技棱镜】2025聚焦科技引领的商业变革_第5页
已阅读5页,还剩72页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

△△△

LookingGlass

Bringingtech-ledbusinesschangesintofocus

/thoughtworks

strategyDesign.Engineering·2025

Introduction3

OperationalizingAIforbusinessimpact4

Strengtheningthedatavaluechain11

ReimaginingresponsibletechfortheeraofgenerativeAI18

Enablingricherexperiencesthroughmultimodalinteractions26

Unlockinggreatervaluefromphysical-digitalconvergence34

Glossary42

△△

Introduction

WelcometotheLookingGlass2025.Unlikemanytechtrendreports,

Thoughtworks’LookingGlassisnotintendedtoshinealightonthe

latestbuzzwords.Instead,wetakealongtermlookatthetechnology

horizonsandexplorewhatthatmeansforbusinesses.Whatarethethingsyouneedtoknowaboutnow?Andwhat’slikelytobeimportantinthe

longerterm?TheLookingGlassenablesyoutounderstandandinterpretemergingtechnologiessoyoucanmakesound,strategicchoicesfor

yourorganization.

Therelentlessspeedoftechnologicaladvancementmakesithardertopredictwhat’scomingandwhereyourinvestmentswillpayoffthemost.BreakthroughsinareassuchasagenticAIpromisetoupendhowwe

thinkabouttechnology.Buthowquicklyshouldyoupreparetoadapt?Here’swhereThoughtworks’LookingGlasscomesin.

Inthisedition,weexploremorethan90trendsthroughfivedistinct

perspectivesthatdefinetheevolvingtechlandscapeinbusiness.

Someofthesetrendsarealreadytransformingoperations,whileothersremainjustoverthehorizon,sparkinginterestanddebatebutstill

unfolding.Forbusinessleaders,keepingabroad,strategicperspectiveonthesedevelopments—bothcurrentandfuture—isessential.

LookingGlassoffersexactlythat:aframeworktogainacomprehensiveunderstandingofkeytrends.

Thefivelensesprovideclarityandfocus,helpingensureyourorganizationremainsadaptable,resilientandreadytoharnessorrespondtothe

inevitableshiftsintechnologythatshapeourmodernworld.

RachelLaycock

ChiefTechnologyOfficer,Thoughtworks

©Thoughtworks,Inc.AllRightsReserved.4

OperationalizingAIforbusinessimpact

ThemainstreamingofAI—andgenerativeAIinparticular—iscontinuingapace.ButasAIproliferates,it’smoreevidentthatsuccessfullyoperationalizingAImodelsandbringingthemtoproductionremainsachallenge.Fromquestionableoutputtounintendedconsequences,thereareahostofrealand

projectedscenariosthatpreventorganizationsfromleveragingAItoitsfullpotential.

Enterprisescontinuetostrugglewithdataquality,dataaccessibilityandthechallengesofdataat

scale,allofwhichremainfoundationaltorobust,effectiveAI.Asourdataplatformlensexplores,

carefuldatacuration,andeffectivedataengineeringandarchitectureareessential.Theimportanceofsyntheticdata,particularlyinresearchcontexts,asatooltoavoidprivacyanddataintegrityissuesisalsobecomingmoreandmoreapparent.

OrganizationsalsoneedtodevelopbetterapproachestotheevaluationandcontrolofAIsystems.

Forward-lookingenterprisesareadopting‘evals’—testsofAIoutputtodeterminereliability,accuracyandrelevance—andguardrails,programmedpolicylayersthatmitigatetheinherentunpredictabilityofgenerativesystems.

Asadoptionincreases,

improvingthemechanisms

throughwhichAIsystems

areconnectedwithenterpriseapplicationsgrowsmore

important.ProxyservicesareemergingtohelpdeveloperslinkAImodelswiththe

applicationstheybuild.

©Thoughtworks,Inc.AllRightsReserved.5

OperationalizingAIforbusinessimpact

AIagentsaresometimespositionedasthenextstepintheevolutionofAI,duetotheircapacitytomimichumanreasoning.However,thetechnologyremainsrelativelynew,andfindingapplicationsforagentsrequiresdomainexpertise,aswellastheabilitytopreciselymapandmodelcomplex

processesandinteractions.TobuildasustainableandproductiveAIpractice,it’svitalthatthe

organizationdoesn’tresorttoshortcuts,acquirestherequisiteskillsandkeepsinnovationrootedinbusinessrealities.

“Thelessonsfromautomationendeavorsinthe

‘80scouldhelptobuildtherightlevelofhuman-AIagenthandovers.Wemustfocusonaugmentinghumansratherthantryingtosubstitutetheir

currenttaskscompletely.”

SrinivasanRaguraman

TechnicalPrincipal,Thoughtworks

Signals

Theemergenceofsmalllanguagemodels,suchasMicrosoft’sphi-3,andAMD’sAMD135.

ThesemakeitpossibletorunAImodelsattheedgeofnetworksondeviceslikemobilephones,andbecausetheyarerelativelylightweight,focusedandefficient,havearangeofpositive

business,securityandsustainabilityimplications.LLMsalsocontinuetoevolve,withAnthropic’sClaude3.5SonnetLLM,whichhassetindustrybenchmarksintermsofperformance,recentlyupgradedtoincludecomputerusecapabilities.

Researchshowingthatformanyorganizations,AIinvestmentsandadoptionarentnecessarilytranslatingintodeploymentorbusinessimpact.Whileinterestin(andspendingon)AIsolutionsremainshigh,businessesarebeginningtopaymoreattentiontothecostofAIprojects,and

steppingupeffortstoensuretheydelivervalue.

ThecomingintoforceoftheEuropeanUnionsAIAct,whichsetsaninternationalbenchmarkbylayingoutobligationsarounddatagovernance,documentation,humanoversightandsecurityforbusinessesadoptingAIsystems.

Sustained,massiveinvestmentindatacenters,withGoogleeventurningtonuclearpower

togeneratethevastamountsofpoweritsAIofferingsarelikelytorequire.ThisindicatesAIisa

long-termbetthatwillcontinuetogainmomentuminthebusinesscontext,andinsocietyasawhole.

ThegrowthoftoolssimplifyinghowengineersandothersinterfacewithAImodels,suchasLiteLLMandLangchain.

RenewedfocusontacklingLLMhallucinationsandfabrications,withnoveltechniqueslike

‘semanticentropy’beingappliedtorootouterrors,andLLMspolicingtheoutputofotherLLMs.

RisingawarenessofshadowAI,ortheuseofunsanctionedAItoolsintheenterprisecontext,

whichcouldposesignificantproblemsforcompaniesifsensitiveinformationisleakedtoLLMsbyemployees.InonerecentsurveyathirdoforganizationsadmittedtofindingithardtomonitortheillicituseofAIamongtheirteams.

©Thoughtworks,Inc.AllRightsReserved.6

OperationalizingAIforbusinessimpact

Trendstowatch

e

e

s

o

t

g

n

i

n

n

i

g

e

B

48

47

49

O

h

t

n

h

e

i

r

o

43

42

44

n

o

z

41

45

46

40

36

37

32

35

31

51

20

30

29

34

19

14

25

w

24

o

n

50

23

18

13

9

28

33

39

g

12

8

5

22

27

38

17

n

i

e

21

16

11

7

4

26

3

e

S

10

6

1

2

15

AnticipateAnalyzeAdopt

Strategicrecommendation

Seeingnow

Adopt

1.Accessibilityinmultimodalexperiences

2.Agent-basedsimulation

3.AIagents

4.AIasaservice

5.AIinsecurity

6.AI-assistedsoftwaredevelopment

7.Automatedcompliance

8.Collaborationecosystems

9.Datamesh

10.Edgecomputing

11.Ethicalframeworks

12.EvaluatingandmanagingAIoutputs

13.Evolutionaryarchitectures

14.ExplainableAI

15.GenerativeAI

16.Integrateddata

andAIplatforms

17.InterfacingwithAI

18.LLMOps

19.MLOps

20.Modeltrainingoptimization

21.Onlinemachinelearning

22.Platformsasproducts

23.Privacyfirst

24.Software-definedvehicles

25.Vectordatabases

Analyze

26.AImarketplaces

27.AIsafetyandregulation

28.AI-generatedmedia

29.Automatedworkforce

30.Autonomousrobots

31.ChangingperceptionsofAI

32.Easingaccessto

generativeAI

33.Federatedlearning

34.MultimodalAI

35.Personalizedhealthcare

36.Syntheticdata

Anticipate

37.Understandableconsent

Beginningtosee

Adopt

38.AI-readydata

39.Finegraineddata

accesscontrols

Analyze

40.AIObservability

41.Datalineage

42.GenAIcomputercontrol

43.Intelligentmachineto

machinecollaboration

44.Productionimmunesystems

45.Smalllanguagemodels

46.Talktodata

Anticipate

47.Adversarialmachinelearning

48.Affective(emotional)computing

49.AIinrobotics

Onthehorizon

Adopt

Analyze

50.AIavatars

Anticipate

51.AGIresearch

OperationalizingAIforbusinessimpact

©Thoughtworks,Inc.AllRightsReserved.7

Theopportunities

Bygettingaheadofthecurveonthislens,organizationscan:

EnhanceknowledgemanagementandtransferbyadoptingGenAItohelpemployeessiftthrough,summarizeandanalyzestoresofenterprisedata,whetherstructured

orunstructured.Awiderangeofproductsareemergingtofacilitatetheretrievalanddisseminationofimportantinformationinindustrieslikeproperty.

HarnessAItoaccelerateprocesseslikelegacymodernizationandcoding.ThoughtworksisalreadysuccessfullyapplyingGenAItoassistteamswithoneofthemostdifficultaspectsofmodernization:understandingandunpackingtheintricatewebofconnectionsthat

typicallyunderpinlegacysystemsandcodebases.AIassistantscanalsosignificantly

boosttheproductivityofsoftwaredevelopmentandotherteamsbytakingoverfrequent,repetitivetasks.

ExploreAIagentstoelevateautomation,potentiallytransforminghowemployeesperformtaskslikeschedulingandcustomersupport,andraisingthebarforengagementand

personalizationincustomerinteractions.

BoostthespeedatwhichLLMsarebroughtintoproduction,andtheireffectiveness

whendeployedthroughemergingpracticesandtoolslikeLLMOps,whichacceleratemodeldevelopment;retrieval-augmentedgeneration(RAG),whichcanenhancemodels’reliability;andAIgatewaysorsmartendpointstoconnectAIsystemstoapplications.

Developandcommunicateajoined-upAIstrategythatempowersemployeestoexperimentwithAIinastructuredway,whilepreventingtheemergenceof‘shadowAI’thatcouldposeathreattotheorganization’sintellectualpropertyorreputation.

LeveragesmalllanguagemodelstobringAIinnovationstoedgedevices,offering

opportunitiesforeverythingfromoperationalanalyticstopersonalization—without

compromisingprivacy,sincedatadoesn’thavetobemovedtothecenterofanetwork.

LeadthewayintermsofcomplianceandethicalAIpractices.WeurgeourclientsnotjusttofollowbutembraceregulationsliketheEUAIAct,assuchlegislationoftenreflectswidersocietalsentimentandconcerns—andpotentialcustomerstakenoticeofbusinessesthatareresponding.

©Thoughtworks,Inc.AllRightsReserved.8

OperationalizingAIforbusinessimpact

Whatwe’vedone

PEXA

ThoughtworkspartneredwithdigitalpropertytechnologycompanyPEXA,AWSandRedactiveto

developaninnovativeandversatileAIassistantthathasboostedtheproductivityofPEXA’semployeesbyprovidingpersonalizedanswerstoqueriesandaugmentingtaskslikeinformationdiscovery.

SeamlesslyintegratedwithPEXA’sinternalsystems,thesolutionalsometrobustrequirementsfordatasecurityandprivacybyequippingtheassistantwithpermissionsawareness,ensuringemployeesareonlyabletoaccessinformationclearedforsharing.

OperationalizingAIforbusinessimpact

©Thoughtworks,Inc.AllRightsReserved.9

Actionableadvice

Thingstodo(Adopt)

•IdentifyAIchampionswhocanhelpguideandteachyourorganizationaboutthepotentialuse

casesforemergingsolutions—butunderstandthatAIcanandwillbeappliedindifferentways

inalmosteverypartoftheenterprise,whichmeansthesechampionsneedtokeepanopenmind.Havingpeoplewithaclearideaofwhat‘good’lookslikecanreducerisksandensureAIinitiativesfocusonmeaningfulbusinessresults.

•ImplementaholisticandcomprehensiveAIstrategyforyourorganizationthatincludesguidelinesonpermittedtoolsandthecontextsinwhichAIcanbeused,tominimizetherisksofshadowAI.

•Adoptretrieval-augmentedgeneration(RAG)whendevelopingAIsystems,togivereliabilityanupliftandpositionmodelstocreatemorespecificoutputs.Integratingevalsandobservabilitycanfurtherenhancetheresilienceofsystemsoverthelongterm.

•EmbedAIthroughoutthesoftwaredevelopmentlifecycle.Maximumresultsareachievedwhen

theroleofAIisn’tjustlimitedtocoding,butassistswithprocessesliketestinganddocumentation.

•ApplydatameshanddataproductthinkingtoensureAIapplicationsarebuiltontherobustdatafoundationneededtoensuretheydeliverbusinessorcustomervalue.Disciplineslike

datacuration,whichcreates,organizesandmanagesdatasetssothey’retransparentandeasilyaccessible,alsocontributetothesuccessofAI.

•UseproxiestosimplifythewayteamsinteractandleverageAImodels,pavingthewayfortheenhancementofapplicationstheydevelopwithAIfeaturesandcapabilities.

OperationalizingAIforbusinessimpact

©Thoughtworks,Inc.AllRightsReserved.10

Thingstoconsider(Analyze)

•Avoidwhat’sknownasthe‘substitutionmyth’—theideathatAIcansimplydirectlyreplacea

human.Instead,buildandimplementsystemsthataugmentrolestomaketeamsmoreproductiveandengaged,whileacknowledgingthecontinuedimportanceofhumanjudgementandoversight.

•BecognizantofvariedexpectationsaroundAI.ResearchsuggestspeoplemayapproachAIdifferentlydependingonculturalbackground,withsomewantingahighdegreeofcontrolandothersprioritizingasenseofconnection.Thesedifferences,aswellasvariancesincontextorsituation,needtobeunderstoodandacknowledgedwhenplanningandimplementingAI.

•Paycloseattentiontocosts,andtrytoidentifytheapproachesmostlikelytomeetyourneeds

whilegeneratingreturnoninvestment.RunningAImodelscanbeexpensive,especiallyifexpenseslikeemployeecompensationarefactoredin.Keepingspendingincheckrequiresactivefinancial

monitoring(i.e.FinOps)andconsiderationofthingslikesmalllanguagemodels.

•MonitorAIregulationandfuturepolicydevelopments,particularlyhowtheseintersectwith

privacylaws,whichcouldhaveamassiveimpactonthedataresourcesavailableforAIprojects.MultipleUSstates,andcountriesfromCanadatoIndiaandJapan,areplanningtoenhanceorrolloutlegislationthatwillsetguardrailsaroundAIuseanddevelopment.

Thingstowatchfor(Anticipate)

•QuestionsaroundlegalliabilityandaccountabilityforthenegativeconsequencesofAIuse.AsissuessuchasAImisleadingcustomersandtheassociatedlegalchallengesemerge,authoritiesliketheEUaremovingtomakeorganizationsmoreculpable.

•ThepotentialgrowthofAIcompanions,designedtoprovideemotionalsupport,friendshiporevenintimacy.Whilethesecouldhelpcombatlonelinessandisolation,theymayalsohavetroublingimplicationsforhumaninteraction,requiringbusinessestothinkcarefullyabouttheintroductionofAIwithcompanion-likefeatures.

©Thoughtworks,Inc.AllRightsReserved.11

Strengtheningthedatavaluechain

LeveragingdataplatformsandAI

AsenterpriseadoptionofAIgainspace,there’srisingawarenessofdata’sroleasadifferentiator,andasourceofcompetitiveedge.Developingthecapabilitiestoleveragedataatspeedandscale,and

becometrulydata-driven,hasbecomeanemergingpriority.Treatingdataasaproductrepresentsoneofthemosteffectivemeanstoachievethisgoal,andthebestwaytobuildanddistributedataproductsisthroughdataplatforms.

Theprinciplesthatunderpinhigh-performancedataplatformsremainthesame—decentralization

andfederateddataownership—butnewtrendsandopportunitiesinthespacearepresenting

challengesthatorganizationsneedtobepreparedfor.Inparticular,theriseofgenerativeAI(GenAI),andtheimportanceofunstructureddatainit,requiresteamstothinkdifferentlyabouthowdatais

managedandprocessed.It’sbecomingcriticaltotreatunstructureddataasafirstclasscitizen,notasstructureddata’spoorercousin.

It’salsoimportanttonotetherisingneedforbetter—andideallyautomated—governanceofdataproducts.

Dataproducts—reusabledata

assetsengineeredtodeliver

trusteddatasetsforspecific

purposes—existindynamic

environmentswheretheneedsofteamsandthewiderorganizationareconstantlyevolving,andit’s

importantthattheyalsodevelopinawaythatdeliversvalue.

Maintainingthecapacityforcompetitiveandsustainablechangerequiresintentionaldesign

ofcohesivecentralizedanddecentralizedcapabilities.Someorganizationsarenavigatingawayfromcreatingconsensus-based‘singlesourcesoftruth’toformingintegrated‘contextualtruths’.

©Thoughtworks,Inc.AllRightsReserved.12

Strengtheningthedatavaluechain

Equallyessentialisensuringdataproductsarebuiltwithaclearlinetobusinessadoption.Platformandproductthinkingcanhelp,butthere’saneedtomovebeyondexistingparadigmsandtooling,andconsiderapplyinghuman-centereddesignformoreeffectivewaysfordatatobeconsumed

andleveragedbybusinessusers.GenAIandtrendslike‘talktodata’andgraph-baseddiscoveryarecreatingpromisingopportunitiesinthisspace,transformingthewayteamsinteractwith

andconsumedata.

“AnopenandevolvingdataandAIplatformallowsorganizationstoembraceuncertaintyinrhythmwithchangingdemands,fosteringacultureof

continuouslearning.”

NimishaAsthagiri

TechnicalPrincipalandDataMeshLeader,Thoughtworks

Signals

•Unstructureddatamovingfromasupportingtoastarringrole.

There’sgrowingfocus

onthe

useofunstructureddata(suchastext,video,imagesandaudio)tobuildbetterAItrainingmodels,whichrequiresintegratingandworkingacrossdifferenttypesofdatainasfrictionlessawayas

possible.

Startupsinthisspacearegainingsignificantinvestment

andthelikesofIBMareunveiling

newproducts

specificallydesignedtohelpenterprisesunleashthepotentialofunstructureddatainanalyticsandAI.

•EnterprisesapplyingGenAItobetterleverageunstructureddata.GenAI’sabilitytoparseandsummarizevastquantitiesoftheinformationcontainedineverythingfrommeetingrecordingstoPowerPointpresentations,andtosupportnaturallanguageinteractions,is

transformingthe

wayteamsaccessandusedata

andenhancingknowledgemanagement.However,thistrendisalsoraisingquestionsastowhetherAIandGenAIplatformsshouldbeintegratedwithotherdataplatformsorkeptdistinct,which,insomecases,isleadingtoplatformproliferation.

•Moreorganizationsgrapplingwiththechallengesoftreatingdataasaproduct,asitbecomesabusinessimperative.

Researchshowsthevastmajorityofbusinessesseeclearbenefits

fromsuchanapproach,includingimproveddatasharingandstrengtheningtheconnectionbetweendataandbusinessgoals.However,theyareconfrontingmultiplebarriersalongtheway,fromfragmented

systemstouncertaintyaboutdataprovenance.

•Therisingimportanceofdatadiscoverability.Byempoweringuserstobetterdiscover,

understandandusedataassets,datacatalogscanplayanimportantroleindataplatforms

andadataproductapproach.Buttheycanalsocausemoreissuesthantheysolveiftheiruserexperiencesorcapabilitiesarelimited,impedingthediscoveryprocess.Therecent

introduction

ofknowledgegraphs

todataplatformsisaddressingtheserisks,makingitpossibletodrawoutrelationshipsandnuancesindatathataretypicallylostintheprocessofabstraction.

•Morepressurebeingputondatateamsto

demonstrateROIandmanagecostsmoreeffectively

.

Theincreasinglyestablishedlinkbetweendatastrategyandenterpriseperformance

alsomeanstheseteamscannolongerworkinisolation;insteadstrategiesshouldbeco-developedwith,

andcreateplatformsthatdeliverresultsfor,thebusiness.

©Thoughtworks,Inc.AllRightsReserved.13

Strengtheningthedatavaluechain

Trendstowatch

e

e

s

o

t

g

n

i

n

n

i

g

e

B

42

h

t

n

O

h

e

z

i

r

o

39

40

n

o

38

41

34

30

33

29

25

32

28

24

20

w

o

19

31

23

18

13

27

9

14

n

5

35

17

12

8

37

22

36

g

n

3

21

16

11

7

26

i

e

4

e

S

15

10

6

2

1

AnticipateAnalyzeAdopt

Strategicrecommendation

Seeingnow

Adopt

1.AIasaservice

2.Automatedcompliance

3.Collaborationecosystems

4.Datacatalog

5.Datafitnessfunctions

6.Datamesh

7.Dataproductspecification

8.Developerexperienceplatforms

9.Digitaltwin

10.Edgecomputing

11.Ethicalframeworks

12.ExplainableAI

13.FinOps

14.Greencomputing

15.IntegrateddataandAIplatforms

16.Knowledgegraphs

17.MLOps

18.Modeltrainingoptimization

19.Onlinemachinelearning

20.Platformsasproducts

21.Privacyfirst

22.Privacy-enhancingtechnologies(PETs)

23.Securesoftwaredelivery

24.Smartsystemsandecosystems

25.Vectordatabases

Analyze

26.Autonomousrobots

27.Autonomousvehicles

28.Datacleanroom

29.Datamarketplaces

30.Decentralizeddataarchitectures

31.Federatedlearning

32.Semanticrepresentationaltechnologies

33.Syntheticdata

Anticipate

34.Understandableconsent

Beginningtosee

Adopt

35.AI-readydata

36.Datacontract

37.Finegraineddataaccesscontrols

Analyze

38.Datalineage

39.Integrating

unstructureddata

40.Intelligentmachinetomachinecollaboration

41.Talktodata

Anticipate

42.Decentralizedpersonaldatastores

Onthehorizon

Adopt

Analyze

Anticipate—

Strengtheningthedatavaluechain

©Thoughtworks,Inc.AllRightsReserved.14

Theopportunities

Bygettingaheadofthecurveonthislens,organizationscan:

ConsolidatedataandAIplatformcapabilities,enablingAIasaservicetoembedthis

newtechnologyandempoweruserstoleverageitsuccessfullythroughouttheorganization.SurveyshaveshownthatdespiteconcernsaboutthewiderimpactsofAI,adoptionhas

positiveimplications

forteams’collaboration,efficiencyandperformance.

UseAI(andGenAI)tobuildandmaintaindataproductsmoreeffectively.EmergingAItoolshavethepotentialtocontributetodataproductsina

numberofways

,fromsynthesizingandanalyzinginformationgarneredinend-userresearchortesting,toacceleratingcodingand

creatingdocumentationthatcansmooththepathtoeffectiveadoption.

Enhancecontrolovercosts.Withdatamanagement

oftendominatingenterprisetechnology

spending

,introducingnewtoolingtotrackdatalineageandanalyzetheimpactofcomplex

datainitiativescanhelpteamsdetermineanddemonstrateROIwithgreaterprecision.

FinOps

thinking

cancontributesignificantlytothisprocessbystrengtheningthelinksbetweentechandbusinessteamsandensuringinvestmentscomewithfinancialaccountability.

Strengthendatagovernancebyintroducingemergingbestpracticesandstructures.Theseinclude

datacleanrooms

,secure,self-containedenvironmentswhereenterprisescanblendproprietaryandthird-partydatatoimproveanalyticsandpersonalizationwhileprotecting

customerprivacy;and

datacontracts

,whichbysettinggroundrulesfordatausersand

consumers,canimprovetransparencyandtrustwhensharingdataacrossanorganization.

CombineknowledgegraphsandGenAI,whichcanenhanceunderstandingoflarge,

complexdatasetsbymappingtherelationshipsamongentitieswithinthem.Thisopens

thepossibilityofmoresemanticapproachestointegration,whichinturncanhelpcreate

abetteruserexperiencefordataconsumers.Inaddition,combiningknowledgegraphsandGenAIcanalsodeliverbetterLLMresponsesbecausewe’retakingexplicitknowledgefromknowledgegraphsandcombiningitwithimplicitstatisticalknowledgefromLLMs.

©Thoughtworks,Inc.AllRightsReserved.15

Strengtheningthedatavaluechain

Whatwe’vedone

Pfizer

Thoughtworksisworkingactivelywiththeseleadingpharmaceuticalcompaniestocreatedatameshplatformsthatenhancetheirabilitytocreateanddelivertransformativedataproducts.WithPfizer,

wehelpeddevelopcutting-edgelayeredplatformsservingAI-powereddataproducts,graph-basedsemanticinteroperability,andLLM-basedagentsthatdrivethefirm’soncologyresearch,supportingearlydrugdiscovery.

Gilead

ForGilead,wesupportedthedesignandimplementationofGileadDnA,ascalableenterprise-widedataplatformthatprovidesdataengineersandresearcherswithasecureself-serviceenvironmentfordataprocessing,completewith‘talktodata’functionality.

Strengtheningthedatavaluechain

©Thoughtworks,Inc.AllRightsReserved.16

Actionableadvice

Thingstodo(Adopt)

•Laytherightfoundationsforcreatingeffectivedataproductsbyimplementinga

datamesh

,

whichplacesdatawithinthereachofteamsthatneeditmostandreducesfrictionbetweendataproducersandconsumers.

•Automatedatagovernanceasmuchaspossibletoensurepoliciesareimplementedconsistentlyandwithminimalimpactondatausageandconsumerexperience.

Fitnessfunctions

andmore

rigorousmonitoringofservicelevelindicators(SLIs)canbegoodplacestostart.

•Starttreatingunstructureddataasafirstclasscitizenthatisgiventhesameattentionand

prominenceasstructureddatainyourdataplatform,anddrawonitspotentialtoimproveanalyticsandAImodels.

•Investinasuperiordataproductdevelopmentexperiencetoaccelerateadoption.Mapping

decisionjourneyscanhelptheorganizationbetterunderstandandtracehowtomovefromusecases

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论