版權(quán)說明：本文檔由用戶提供并上傳，收益歸屬內(nèi)容提供方，若內(nèi)容存在侵權(quán)，請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡(jiǎn)介
AcademyofManagementReview
COMPETITIVEADVANTAGESTHROUGHARTIFICIALINTELLIGENCE:TOWARDATHEORYOFSITUATEDAI
Journal:
AcademyofManagementReview
ManuscriptID
AMR-2020-0205-Original.R3
ManuscriptType:
OriginalManuscript
TheoreticalPerspectives:
Resourcebasedview,Knowledge-basedview,Learning,Adaptation,Routines,andKnowledgeManagement
OtherTheoreticalPerspectives:
TopicAreas:
Business-levelresources/capabilities<BusinessandCompetitive
Strategy<BusinessPolicyandStrategy,Knowledgemanagement<StrategicManagementProcess<BusinessPolicyandStrategy,
Technologyevolution<TechnologyandInnovationManagement
OtherTopicAreas:
Abstract:
Howcanfirmsestablishcompetitiveadvantagesusingartificial
intelligence(AI)?AlthoughAIisbeginningtopermeatebusiness
activities,ourunderstandingofhowAIcanbeusedtocreateuniquevalueislimited.Toaddressthisvoid,weintroducetheconceptof
situatedAIandilluminateitsimportanceforestablishingAI-drivencompetitiveadvantages.Thepaperhighlightstheorganizational
activitiesinvolvedinsituatingAI—grounding,bounding,andrecasting.Italsoexplainstheconditionsinwhichthesesituatingactivitiesbetter
enablefirmstodevelopAI-drivencapabilitiesthatarefirm-specific,cost-effective,andappropriateforopportunitiesinthestrategicenvironment.Thus,thispaperprovidesanintegrativeframeworkforconnectinga
firm’sAIpursuitstocompetitiveadvantage.
60
Page1of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
COMPETITIVEADVANTAGETHROUGHARTIFICIALINTELLIGENCE:TOWARDATHEORYOFSITUATEDAI
AyendaKemp
PamplinCollegeofBusiness
VirginiaTechUniversity
ayenda@
Acknowledgments:Iamgratefultotheassociateeditor,AllanAfuah,andtothereviewersfor
theirinsightfulguidanceandsupportthroughouttherevisionprocess.ThispaperhasbenefitedimmenselyfromhelpfulcommentsfrommycolleaguesCynthiaDevers,DeviGnyawali,RichardHunt,AbrahamOshotse,KarenSchnatterly,andMaxStallkamp.
60
AcademyofManagementReviewPage2of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
COMPETITIVEADVANTAGETHROUGHARTIFICIALINTELLIGENCE:TOWARDATHEORYOFSITUATEDAI
Abstract
Howcanfirmsestablishcompetitiveadvantagesusingartificialintelligence(AI)?AlthoughAIis
beginningtopermeatebusinessactivities,ourunderstandingofhowAIcanbeusedtocreate
uniquevalueislimited.Toaddressthisvoid,weintroducetheconceptofsituatedAIand
illuminateitsimportanceforestablishingAI-drivencompetitiveadvantages.Thepaper
highlightstheorganizationalactivitiesinvolvedinsituatingAI—grounding,bounding,and
recasting.ItalsoexplainstheconditionsinwhichthesesituatingactivitiesbetterenablefirmstodevelopAI-drivencapabilitiesthatarefirm-specific,cost-effective,andappropriatefor
opportunitiesinthestrategicenvironment.Thus,thispaperprovidesanintegrativeframeworkforconnectingafirm’sAIpursuitstocompetitiveadvantage.
60
Page3of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
TheprospectofusingAItoestablishcompetitiveadvantagespresentsatheoreticalpuzzle.
Estimatespredictthatby2033,somewherebetween40and50percentofjobswillbeautomatedusingintelligentalgorithms(Frey&Osborne,2013),reflectingenhancedproductivityandlowercosts.ItisalsopredictedthatAImayleadtonewproducts(Barro&Davenport,2019;
Davenport&Kirby,2015),byallowingfirmstoembedAIintotheirproductsandbyigniting
innovationsinafirm’sproductdevelopmentprocesses(Gregory,Henfridsson,Kaganer,&
Kyriakou,2021;Cockburn,Henderson,&Stern,2019).Despitethispromise,agrowingbodyofresearchhighlightsthatAImaypresentsubstantialstrategicobstacles.AImaybemyopic
(Balasubramanian,Ye,&Xu,2022),incapableofperceivinginterdependencieswithinafirm
(Raisch&Krakowski,2021),andrecalcitranttomanagerialcontrol(Murray,Rhymer,&
Sirmon,2021).ThesefactorssuggestthatusingAItolowercostsandcraftdesirableproducts
maynotbeassimpleaspreviouslysuggested.Inaddition,AIisaformofexplicitknowledge
(Broussard,2018;Shrestha,He,Puranam,&vonKrogh,2021),andresemblesageneral-purposetechnology(Teece,2018).SoevenwhenAIleadstovaluecreationwithinafirm,theactivitiesunderpinningtheseoutcomesmaybereplicablebyafirm’srivals.Thus,whileAIholdspromiseforpromptingcompetitiveadvantages,itisunclearhowthispromisecanberealized.
ThispaperbeginstoresolvethispuzzlebydevelopingatheoryofsituatedAI—AIwhoseagencyiscircumscribedinafirm’sexperiential,structural,andrelationalsystems.Wegroundourframeworkintheorganizationalcapabilitiesliterature,whichholdsthatcompetitive
advantagesemergeprimarilywhenfirmsdeploytheirstrategicassetsusingorganizational
capabilitiesthatareidiosyncratic(Barney,1991),inexpensivetodevelop(Winter,2000),andalignedwiththefirm’sinternalandexternalenvironment(Mahoney&Pandian,1992;Sirmon,Hitt,&Ireland,2007).WearguethatachievingtheseoutcomesismadedifficultbyAI’s
60
AcademyofManagementReviewPage4of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
propensitytoactwithagency(Murrayetal.,2021),whichmaybecounterproductivewhennotproperlycontextualizedwithinthefirm.WealsoacceptthatuncontextualizedagencymaybeAI’sbaselinestate(Balasubramanianetal.,2022).
Weaddressthestrategiclimitationsofartificialintelligencebyexplaininghowfirmsmay(1)circumscribeAI’sagencyinthefirm’suniqueexperiencesandsystemsand(2)embedthis
transformedAIinthefirm’sorganizationalcapabilitiesthroughthreesituatingactivities:
grounding,bounding,andrecasting.Groundinginvolvesorchestratingwhichexperiencesone’sAIwillbeallowedtolearnfromacrosstheorganization.Boundinginvolveseffortstoshapetheexperiencesanchoringacompetitor’sAI.Recastinginvolvesorchestratingthecontinual
adaptationofalgorithmsandtheirsurroundingroutinestoenhanceAI’salignmentwith
interdependentactivitiesinafirm.Wealsoconsiderhowtechnologicalconstraintsand
environmentaldynamisminfluencethebenefitsofsituatingAI.Thus,thispaperacknowledgesAI’sstrategiclimitationswhileexplaininghowfirmscanovercometheselimitationstobetterrealizeAI’spotentialasanewfoundationforcompetitiveadvantage.
CONCEPTUALBACKGROUNDThePromiseofAIforCompetitiveAdvantage
Artificialintelligence(AI)broadlyreferstomachinesthatcancompletecognitivetasks
previouslypossibleonlyforhumans(Davenport,2018).Whilethereisalonghistoryof
machinesdisplacinghumanworkers,theriseofAIisuniqueinthatmachines,forthefirsttime,can“l(fā)earn”andperformtheirworkwithagency(Faraj,Pachidi,&Sayegh,2018).Withprevioustechnologies,machinescompletedtheirworkbyfollowingintricateif-thenstatements
programmedbyhumanactors.Themachinehadnoagencytospeakof;itsactionswereadirectreflectionoftheknowledgeofitsprogrammers(Dreyfus&Dreyfus,2005;Norman,2017).In
60
Page5of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
contrast,withAI,thecomputerisprovidedasetofinputdata,alearningobjective,anerror
function,andamathematicalalgorithmforminimizingthaterrorfunction(Chenetal.,2020;
Alpaydin,2016).Armedwiththisbasicdescriptionofaproblem,thecomputerthenlearnsits
own“rules”forlinkingtheinputdatatothedesiredoutcomes.Whatiscriticalabouttheserulesisthattheyarenotcreatedbyhumanactorsand,inmanycases,cannotevenbeexplainedby
humans(Castelvecchi,2016).Thus,AIcanbethoughtofaspossessingadistinctformofagenticrationalitythatincreasinglyallowsmachinestoperformcognitivetasksatalevelequalingor
surpassinghumanperformance(Murrayetal.,2021).
ThepowerofAIhasledmanytobelievethatAIwillrevolutionizeeconomicproductionbymakingfirmsmoreefficientthroughintelligentautomationandbyassistinghumansin
solvingnovelproblemsthatmayleadtovaluecreationthroughthedesignofnewproductsandtheimprovementofoldones(Barro&Davenport,2019;Brynjolfsson&Mitchel,2017;Frey&Osborne,2013).Indeed,wearebeginningtoseeprocessandproductimprovementswithAI
acrossmultipleindustries(Tarafdar,Beath,&Ross,2019).Asoneexample,DBSBankrecentlyimplementedAIthatpredictswith85percentaccuracywhetheranemployeewillleavewithinthreemonths.ThefirmisnowusingAItopoweradigital-onlybankinIndiathatemploys90
percentfeweremployeesthanatraditionalbank(Davenport,2018).Asasecondexample,
fragrancedesignersnowuseAIduringproductdevelopmenttoproduceperfumesthatappealtoconsumersmorethanfragrancescreatedbyhumanexpertsalone(Goodwinetal.,2017).
Despiteimprovementsinoperationsandproductdesign,however,firms’investmentsinAImayfailtomaterializeasprofits.InarecentsurveybytheBostonConsultingGroup,nineoutoftentopmanagersreportedthatAIrepresentsalargebusinessopportunityfortheirfirms
(Ransbothametal.,2019),and43%ofexecutivesreportedhavingimplementedAIintheir
60
AcademyofManagementReviewPage6of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
organizations.Thereportalsonoted,however,that“mostcompanieshaveahardtimegeneratingvaluewithAI.”Lui,Lee,andNgai(2022)offerempiricalsupportforthisconcern,findingthatmarketspenalizeAIadoptionsatthefirmlevel.Thus,whileAIisbeginningtoleadtoproductandprocessimprovements,firm-levelbenefitsofAI,suchasimprovedmarketperformanceorcompetitiveadvantage,maybemoredifficulttoachieve.
ThreeStrategicLimitationsofAI
Whatexplainsthisdisconnect?WhileexistingtheorydoesnotexplainhowfirmscansystematicallyleverageAItodevelopcompetitiveadvantages,recentresearchshedssomelightonobstaclestodoingso.Weexaminetheseobstaclesfromanorganizationalcapabilities
perspectiveandidentifythreereasonswhyfirmsmaystruggletoestablishcompetitive
advantageswiththeirAIinvestments.WefocusonAI’sgeneric,explicit,andmyopicnature.WhileAIundoubtedlyraisesmanyothernewchallengesforfirms,thesethreelimitationstakeprominenceinourtheoryduetotheiradverseeffectsoncapabilitydevelopment,whichwewillargueiscentraltoestablishingcompetitiveadvantageswithAI.
ThefirststrategicchallengeofAIisitsgenericnature.ThelogicemergingfromanAIalgorithmisgenerallynotuniquetotheuserapplyingthatalgorithm.Instead,itcanbe
“rediscoveredbyanyoneusingthesameprocedure”(Shresthaetal.,2021:4).Forexample,allelsebeingequal,aneuralnetworkalgorithmwillarriveatthesamelogicforconnectingasetofinputstooutputsregardlessofwhetherSpotifyorPandoraoperationalizestheneuralnetwork.ThissuggeststhatAImaydisplaycommonbehavioralpatternsacrosscompetingfirms.This
pointregardingthegenericnatureofAIiscriticalbecauseAIisregardedasageneral-purposetechnologyakintoelectricity,thesteamengine,ortheinternet(Brynjolfsson&Mitchell,2017;Frey&Osborne,2013;Lynch,2017).General-purposetechnologiesarelikelytobewidely
60
Page7of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
adoptedamongcompetingfirms(Bresnahan&Trajtenberg,1995).Asaresult,these
technologiestendtogenerateeconomy-widebenefitsratherthanprivaterents(Bresnahan&Trajtenberg,1995;Teece,2018).Thus,whileAImayhelpafirmtodevelopbetterandcheaperproducts,theex-facieexpectationisthatAIwillhelpafirm’scompetitorstodothesame.
ThesecondstrategiclimitationofAIisthatAImanifestsasaformofexplicitknowledge.AIalgorithms,andthedatathatdrivethem,mustbeavailabletothecomputerintheformof
explicitinstructionsormathematicalformulas(Broussard,2018).Consequently,the
organizationalknowledgethatdrivesafirm’sAIprocessesmaybeextractedfromcyber-attacksandmaybehighlyportableduringemployeeturnover(Tramér,Zhang,Juels,Reiter,&
Ristenpart,2016).Thisobservationisconsistentwiththegeneralideathatexplicitknowledgediffusesrelativelyquicklyacrossorganizationalboundaries,makingitchallengingtobuild
competitiveadvantages(Grant,1996;Nickerson&Zenger,2004).ThischallengeisespeciallysalientforAIbecauseintellectualpropertylawsdonot(currently)allowforpatenting
mathematicalformulasandprocedures(Gaudry&Hayim,2018;Liyange&Berry,2019).Thus,identifyingmechanismsforpreventingthespreadofAIassetsacrossfirmboundariesiscritical.
ThethirdstrategiclimitationofAIismyopia.AIismyopicinthesensethatAI
algorithmslackcontextualawarenessofactivitiesandeventsbeyondthescopeoftheirassignedtasks(Balasubramanianetal.,2022;Dreyfus,2012;Raisch&Krakowski,2020).AsingleAI
algorithmcantypicallyexecuteonlyasmallsubtaskwithinanentireorganizationalroutine
(Davenport,2018).Thus,AI-drivenroutineswillnormallyemploycollectivesofAIalgorithms(see,forexampleKumar,Venugopal,Qiu,&Kumar,2018).Yet,analgorithm’sabilityto
recognizeinterdependenciesbetweenitstaskandothertaskswithinthefirmislimited.Thismayresultinexpensivetechnicalandoperationalfailures(Balasubramanianetal.2022;Dreyfus,
60
AcademyofManagementReviewPage8of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
2012).Moreover,becauseAIcompletesitstaskwithhighdegreesofagency,managersfinditdifficulttocorrecttheseerrorswhentheyoccur(Murrayetal.,2021).
ArelatedconsequenceofAI’smyopiaisthatAIwilllackasophisticatedunderstandingofafirm’sstrategy.Thus,evenwhenAIbehavesinamannerthatisoptimalforcarryingoutatask,thereisnoguaranteethatthiswillresultinbehaviorthatisappropriateforthekindsof
marketopportunitiesafirmispursuing(orshouldbepursuing)(Balasubramanianetal.,2022).Forinstance,abudgetairlinemaydevelopAIthatcorrectlyidentifiesthatacustomercanpay$2,000forashort-haulflight.Still,thisAImaybeincapableofunderstandingthatmakingsuchanofferisinconsistentwiththefirm’smarketidentityandlow-costproviderstrategy.Inotherwords,theultimatevalueofAItothefirmdependsnotonlyonitstaskeffectivenessbutalsoonitsfitwiththefirm’soverallstrategy.Thus,organizationalmechanismsforovercomingAI’s
myopiaarenecessaryforestablishingcompetitiveadvantages.
Thegeneric,explicit,andmyopicnatureofAIallendangeritsfirm-levelbenefits,butareuniquelydifficulttosurmountbecauseAIisbothamachineandagentic.Forexample,AI’s
genericnatureisakintogeneralhumancapital.Generalhumancapitalcannottypicallyunderlyinterfirmadvantagesbecauseafirm’scompetitorscanusuallyacquireanddeploythat
knowledgeinwaysthatcloselymimicthefocalfirm(Barney&Wright,1998).UnlikewithAI,however,generalhumancapitalisconvertiblewithinafirmthroughsocializationandcanbe
madecontext-specificashumanemployeesengageinrichinformalsocialinteractionswithinthefirm(Coff,1997).ThisoptionisunavailableforAIbecauseitisamachine.Ontheotherhand,AI’scapacitytoactwithagencyremovesthetypicalmodesofcontrollingtechnologyinan
organization.Consider,forexample,thecaseofexpertsystems,whichwereheavilyusedbeforetherecentriseofAI(Cholletetal.,2022).LikeAI,thesesystemsweremyopicbecausethey
60
Page9of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
coulderroneouslyoverlookorganizationalinterdependencies.However,whensuchproblems
arose,theycouldbeaddressedbysupplyingthemachinewithmorerules.ThisoptionisnotasfeasiblewithAI,whichdependmoreondatathanrulesasbehavioralconstraints.Thus,AIaltersthefirm’sknowledgeproductionfunction,makingpathstocompetitiveadvantageelusive.
ConceptualBuildingBlocksforTheorizingAI-drivenCompetitiveAdvantages
WedevelopatheoryofsituatedAItoexplainhowfirmscanovercomethesestrategiclimitationstocraftcompetitiveadvantages.Weuseasourconceptualfoundationsthe
organizationalcapabilitiesliterature(Eggers&Kaplan,2013;Nickerson&Zenger,2004),andtheworkonhumanagency(Emirbayer&Mische,1998;Westphal&Zajac,2013).Webrieflydescribeeachconceptualbuildingblockbelow.
CapabilitiesandCompetitiveAdvantage.Anorganizationalcapabilityisacollectionofroutinesthat,togetherwiththeirimplementinginputflows,conferuponanorganization’s
managementasetofdecisionoptionsforproducingsignificantoutputsofaparticulartype
(Winter,2003).Theorganizationalcapabilitiesperspectiveviewsfirmperformanceasafunctionofsystematicandrandomfactors(Winter,2000).Marketsaremodeledascollectivesof
competingfirmssolvingarelatedproblemundertechnicalandbehavioraluncertainty(Afuah&Tucci,1997;Nickerson&Zenger,2004).Profitsareviewedasephemeralintheabsenceof
competitiveadvantagesgrantingsomefirmsastructuraledgeoverothers(Barney,1986).Andaafirm’sprimaryconcernliesinidentifyingandorchestratingpatternsoforganizationalactivitiesthatcanbereliablyleveragedtocreateandcapturevalue(Winter,2003).
Thecapabilitiesperspectivefocusesonthreecoresourcesofcompetitiveadvantage:
firm-specificity,capabilitydevelopmentcosts,andenvironmentalfit.Firm-specificcapabilitiesarethoseproducedusingco-specializedknowledge,leadingthemtohavegreatervalueinsidea
60
AcademyofManagementReviewPage10of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
firmthanexternally(Helfat,1994;Mahoney&Pandian,1992).Also,craftingfirm-specific
capabilitiesismorelikelywhenatleastsomeoftheco-specializedknowledgeneededtoproduceordeployacapabilityistacitorsociallycomplex(Grant,1996;King&Zeithaml,2001;
Nickerson&Zenger,2004).Second,becausecraftingcapabilitiesrequiresorganizationaleffortandresources,thecostofdevelopingacapabilitymustnotsupersedethevalueearnedfrom
deployingthecapability(Argyersetal.,2019;Winter,2000).Finally,customersarelikelyto
respondpositivelytoafirm’sofferingsonlywhenthefirm’scapabilitiesareadequatelymatchedtotheirneeds(Sirmonetal.,2007).
Situatedagency.Organizationsactwithagencytoinfluencetheircapabilitiesandthe
environmentsthatbindthem(Gavetti,Helfat,&Marengo,2017;Gavetti&Torras,2021;Nayak,Chia,&Canales,2020).Agencygenerallyinvolvesfreechoicewithconstraints(Emirbayer&Mische,1998;Giddens,1979).Ourbasicargumentisthat,whilefirmscannotalwayslimitAI’sagencydirectly(andmightnotwantto),firmscanbalanceamachine’sagencywithhuman
agency,bystrategicallystructuringthecontextinwhichAImakessenseofproblemsandappliessolutions.Agencyhasthreedimensionsthatinformhowanactor’sbehaviorisconstrainedandhowfreechoicemaymanifest(Emirbayer&Mische,1998).Theiterationaldimensionofagencyinvolvesactionsanchoredinanorganization’spriorexperiences.Thepracticalevaluative
dimensionofagencyconsidersactionanchoredinanorganization’spresentsocialcontext.Theprojectivedimensionofagencyregardsactionsbasedinanactor’sabilitytoreimaginetheir
organization’spresentarrangementstomeetfuturegoals.Agencyisconsideredtobesituatedwhenconstraintsonagencyoriginatepredominantlyinthesamecontextinwhichtheagentacts(Botti,1998;Westphal&Zajac,2013).WebuildonthisworkbyconceptualizingsituatedAIasthetechnologicalanalogofsituatedagency.
60
Page11of42AcademyofManagementReview
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
Capabilitydevelopment.WearguethatfirmscansituateAIduringcapability
development.Capabilitydevelopmentinvolvesorchestratingorganizationalactionacrossfour
steps:bundlingstrategicassets,embeddingassetsinroutines,assemblingroutinesascapabilities,andmatchingcapabilitiestoopportunitiesintheenvironment(Collis,1994;Eggers&Kaplan,2013;Sirmonetal.,2007).Withinthescopeofthismodel,wetreatinputdataasthemajor
strategicasset(Gregoryetal.,2021),andwereplacethefocusontraditionalroutinesinpriorcapabilitymodelswithafocusonconjoinedroutines,whichreferto(partially)automated
organizationalroutineswheretheroutine’sdesignandexecutioninvolvesamixofhumanandnonhumanagency(Murrayetal.,2021).
We,therefore,defineanAI-drivencapabilityasacollectionofconjoinedroutineswhich,alongwiththerequiredinputdata,allowafirmtoexecutespecificvaluechainactivitiesina
repeatableandreliablemanner(Gregoryetal.,2021;Helfat&Winter,2011;Winter,2003).OurmodelintroducesthreesituatingactivitiesthatfirmsmayleveragetoorchestratethedevelopmentofAI-drivencapabilities:grounding,bounding,andrecasting.
AcorepartofourtheorizinginvolvesaccountingforhowsituatedAIisadaptedinafirm’sAI-drivencapabilitiesovertime.Therefore,webuildonpriorresearchviewing
organizationaladaptationasincrementallychangingafirm’scorestructuresandstrategies
throughexperimentationandproblem-drivensearch(Ethiraj&Levinthal,2004).Wefocusonthelearningliteraturecenteredonthecognitiveunderpinningsofadaptivecapabilityformation(Gavetti&Levinthal,2000;Tripsas&Gavetti,2000).
Afinalconsiderationadoptedfromtheorganizationalcapabilitiesperspectiveisattentiontodynamisminthefirm’senvironment.Asafirm’sstrategicenvironmentbecomesmore
dynamic,newmarketopportunitiesmayemerge,existingopportunitiesmayevaporate,andthe
60
AcademyofManagementReviewPage12of42
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
technologiesandcapabilitiesneededtocaptureopportunitiesevolve(Sirmonetal.,2007;Teece,2007;Tripsas&Gavetti,2000).EnvironmentaldynamismhassignificanceasitrelatestoAI-
drivencapabilitiesbecauseAIshiftstheextenttowhichafirm’sroutinesareresponsivetochange(Balasubramanianetal.,2022;Murrayetal.,2021).We,therefore,considerhowenvironmentaldynamisminfluencesthebenefitsofsituatingAIforcompetitiveadvantage.
AIcharacteristics.OurtheoryisintendedtoaccountforthelargeandgrowingformsofAItechnologies.WefollowthemachinelearningliteraturebycharacterizingAIalgorithms
basedontheirtrainingparadigms(supervisedversusunsupervisedlearning)andtheirdegreeofexplainability.AnAIalgorithmismoreexplainablewhenitiseasierforhumanoperatorsto
describethelogicthroughwhichthealgorithmlinksinputstooutputs(Aryaetal.,2019;Gilpinetal.,2019).Thismayinvolvedescribingwhichfactorsanalgorithmweighsheavilywhen
arrivingatasolutionorprovidingsomeintuitionforhowthealgorithmtreatstheinteractionbetweendifferentfactors(Gilpinetal.,2019;Hendricksetal.,2019).Whilesomealgorithmsallowforahighdegreeoftransparencyregardingtheirinnerworkings,otherAIalgorithmsdonot.Figure1showshowsomestandardAIalgorithmsfitthistaxonomy.
***Figure1abouthere****
ThelearningparadigmtellsushowAIisinstructedtomakeinferencesfromthedata.
Withsupervisedlearning,AIisprovidedwithtrainingdatainwhichthe“rightanswers”fora
problemhavebeenlabeled(Dike,Zhou,Deveerasetty,&Wu,2018;Murphy,2012).Labelsmaybeprovidedbyhumanactorsormaybeinferredusinghumanknowledgeofthedatasource.Forexample,afirmmaytrainAItowritecomputercodebyscanningsiteslikeforuser-providedcodingquestionsandthenusingthehighest-ratedinputasthecorrectanswer.Incontrast,unsupervisedlearningprovidesAIwithdatabutnotwith“ans
溫馨提示
- 1. 本站所有資源如無特殊說明，都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等，如果需要附件，請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙，網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽，若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間，僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理，對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯，并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容，請(qǐng)與我們聯(lián)系，我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 國(guó)慶節(jié)團(tuán)建主題活動(dòng)方案
- ktv國(guó)慶節(jié)的朋友圈活動(dòng)方案
- 2024-2025學(xué)年新教材高中語文 第三單元 7.1 青蒿素：人類征服疾病的一小步（1）說課稿 部編版必修下冊(cè)
- 2024-2025學(xué)年高中語文 第二單元 七 仁義禮智我固有之說課稿5 新人教版選修《先秦諸子選讀》
- 2025變更勞動(dòng)合同范文
- 2025智能化施工合同
- Unit 12 Weather（說課稿）-2024-2025學(xué)年滬教牛津版（深圳用）英語四年級(jí)上冊(cè)
- 門診手術(shù)策劃方案
- 出資比例 英語合同范例
- 云杉買賣合同范例
- 2025年華能新能源股份有限公司招聘筆試參考題庫含答案解析
- 《中國(guó)心力衰竭診斷和治療指南（2024）》解讀完整版
- 《檔案管理課件》課件
- 2024年度中國(guó)共產(chǎn)主義共青團(tuán)團(tuán)課課件版
- 2025年中考物理終極押題猜想（新疆卷）（全解全析）
- 脛骨骨折的護(hù)理查房
- 房顫手術(shù)后護(hù)理流程
- 抽水蓄能電站項(xiàng)目建設(shè)管理方案
- 北郵工程數(shù)學(xué)期末試卷B卷
- 超長(zhǎng)結(jié)構(gòu)及大體積混凝土專項(xiàng)施工方案
- 初中 初一 數(shù)學(xué) 絕對(duì)值 課件
評(píng)論
0/150
提交評(píng)論