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
WORKINGPAPER|ISSUE03/2023|13MARCH2023
AIADOPTIONINTHEPUBLIC
SECTOR:ACASESTUDY
LAURANURSKI
Thiscasestudyillustratesthedriversofandbarrierstoartificialintelligenceadoptionbyorganisations,andacceptanceofAIbyworkersinthepublicsector.Severalfactorswerecrucialinthesuccessfuladoptionofahuman-centredapproachtoAI,includingafastdiscoveryphasethatinvolvedworkers(orendusers)inthedevelopmentearlyon,andaligninghumanresources,informationtechnologyandbusinessprocesses.Subsidysupportmechanismswerealsospecificallytargetedandacquiredtosupporttheadoption.
However,makingAIsupportavailabletoworkersprovedinsufficienttoensureitswidespreadusagethroughouttheorganisation.TheslowadaptationofexistingworkprocessesandlegacyITsystemswasabarriertotheoptimalusageofthetechnology.Moreover,theusefulnessofthetechnologydependedonboththetaskroutinenessandworkerexperience,therebynecessitatingarethinkingoftheworkdivisionbetweentechnologyandworkers,andbetweenjuniorandseniorworkers.
Successfulhuman-centredroll-outofAIinEuropewillthereforedependontheavailabilityof,orinvestmentsin,complementaryintangibleorganisationalcapital.Verylittleiscurrentlyknownabouttheseinvestments.
TheauthorisgratefultoTomSchraepen(Bruegel)forresearchassistance,toMiaHoffmann(Georgetown’sCenterforSecurityandEmergingTechnology)forcommentsonearlierversions,andtothecontactsatthecaseorganisations,whoprovidedtheircooperationandinputtothestudy.
LauraNurski(laura.nurski@)isaResearchFellowatBruegel
Recommendedcitation:
Nurski,L.(2023)‘AIadoptioninthepublicsector:acasestudy’,WorkingPaper03/2023,Bruegel
1
Tableofcontents
1Introduction 2
1.1Productivityandtechnologyacceptance 2
1.2Theorganisationsinthiscasestudy 3
1.3Selectionofthecase 3
1.4Methodology 4
2AIadoptionbytheorganisation 5
2.1Adoptionprocess 5
2.2Driversandbarrierstoadoption 9
3AIacceptancebystaffmembers 13
3.1Studiedalgorithm:AI-assistedquestionanswering 13
3.2Frameworkforuseracceptanceandactualuse 14
3.3Barrierstotheuseofthealgorithm 15
4Impactandsupport 21
4.1Impactonworkdivisions,learningandsocialrelationships 21
4.2PathtowardsincreasingAIacceptance 22
5Conclusionandrecommendations 26
References 27
Annex:Listofcasestudymaterials 30
S
1Introduction
1.1Productivityandtechnologyacceptance
VL〔!:!c!9l!u〔Gll!8GucG(Vl)!29uGM8GuGL9lbnLbo2G〔Gcpuolo8l(ebl)GxbGc〔Gq〔opL!u8bLoqnc〔!^!〔l89!u2,pn〔9l2o〔o!山b9c〔〔pGu9〔nLG9uqdn9l!〔lo:MoLk.lpG!u〔Loqnc〔!ouo:bLG^!on2ebl2,!uclnq!u8GlGc〔L!c!〔l9uqco山bn〔GL2,p922poMu9lou8l98pG〔MGGu〔pG9qob〔!ouo:uGM〔Gcpuolo8!G2!u〔pGbLoqnc〔!ouo:8ooq29uq2GL^!cG2,9uqM!qG2c9lGop2GL^9plG!ucLG92G2!u〔o〔9l:9c〔oLbLoqnc〔!^!〔l.lp!2bnSSl!u8op2GL^9〔!ou!2o:〔GuqnppGq〔pG“productivityparadox”(「9uq9nGL,Taa2),!G〔pGbpGuo山Guou〔p9〔l9L8G!u^G2〔山Gu〔2!uuGMll〔Gcpuolo8!G2pl:!L山2p9^Guo〔pGGu9cco山b9u!Gqpl2np2GdnGu〔 !ucLG92G2!uu9〔!ou9lbLoqnc〔!^!〔l2〔9〔!2〔!c2.lMocLnc!9l:9c〔oL2!uGxbl9!u!u8〔p!2b9L9qox9LG〔pG2loM9q9b〔9〔!ouo:pn2!uG22bLocG22G29uq〔pGnuqGLn〔!l!29〔!ouo:〔Gcpuolo8lplMoLkGL2(DG^9L9)9uqKopl!,S003).
MpGuoL89u!29〔!ou9lbLocG22G29uq!ucGu〔!^G29LGuo〔9l!8uGqM!〔p〔Gcpuolo8ln2G,MoLkGL2M!lluo〔:nlll9qob〔uGM〔Gcpuolo8!G2!u〔pG!LMoLk(V〔k!uetal,S0TS).Vl9bbl!c9〔!ou2Gdn9lll2n::GL:Lo山2ncpnuqGLn〔!l!29〔!ou,922〔nq!G2o:〔pGp9uk!u8(Xn9uqSpn,S0ST),LG〔9!l(K9M98ncp!,S0S0)9uqpG9l〔pc9LG(19nketal,S0ST)2Gc〔oL2qG山ou2〔L9〔G.EnL〔pGL山oLG,!〔〔9kG2〔!山G,山ouGl9uqM!ll!u8uG22:oLpn2!uG22山oqGl29uqoL89u!29〔!ou9l山oqGl2〔o9q9b〔〔o〔pGn2Go:uGM〔Gcpuolo8l.2ncp9q9b〔9〔!ouLGdn!LG2!u^G2〔山Gu〔!ubGoblGbL9c〔!cG2oLpn山9uLG2onLcG2(H匕)bL9c〔!cG2,!uclnq!u8
〔L9!u!u8,bGL:oL山9ucGG^9ln9〔!ou29uqp!L!u8(Bloo山etal,S0TS).lu^G2〔山Gu〔!29l2ouGGqGq!uoL89u!29〔!ou9lbL9c〔!cG2,2ncp92pn2!uG22bLocG22LGGu8!uGGL!u8,qGcGu〔L9l!29〔!ouoLoL89u!29〔!ouLGqG2!8u(BLG2u9p9uetal,S00S).
lo9u9ll2G〔pGqL!^GL2o:9uqp9LL!GL2〔oVl9qob〔!ouploL89u!29〔!ou2,9uq9ccGb〔9ucGo:VlplMoLkGL2,MG!u^G2〔!89〔G92bGc!:!cc92G!u〔p!22〔nql.lpG2〔nqlqoG2uo〔2GL^G929‘2〔9Lc92G,(Mp!cp
qG山ou2〔L9〔G2LGbLoqnc!plGpG2〔bL9c〔!cG2)oL929‘LG2G9Lcpc92G,(Mp!cp9!山2〔o!山bLo^GGcouo山!c〔pGoL!G2o:c9n29l!〔l)(B9kGL9uqe!l,S0T3).lu2〔G9q,〔p!2!29‘〔G9cp!u8c92G,〔p9〔!lln2〔L9〔G22c!Gu〔!:!c〔pGoL!G2!u9bL9c〔!c9lGx9山blG9uqpL!q8G22G^GL9lq!2c!bl!uG2(!uclnq!u8ll,山9u98G山Gu〔,oL89u!29〔!ou9lpGp9^!onL,b2lcpolo8l9uqGcouo山!c2)pll!uk!u8〔pGoL!G2:Lo山LG2bGc〔!^Gqo山9!u2!uouGc92G.lpG8o9l!2〔o!qGu〔!:lb!〔:9ll2!u〔pGbLocG22o:〔Gcpuolo8l9qob〔!ou9uq〔obLo^!qG2o山G lG22ou2:oLpo〔pbol!cl9uqpn2!uG22.lp!2c92G2〔nql!2b9L〔o:〔pGEn〔nLGo:MoLk子lucln2!^GeLoM〔p
3
project
1
atBruegel,whichaimstoidentifytheimpactoftechnologyonthenature,quantity,andqualityofwork.
1.2Theorganisationsinthiscasestudy
WeanalyseAIadoptionbyFlandersInvestmentandTrade,apublicorganisation,whichwasassistedbyRadix,aprivatefirm
2
.Notethatthroughoutthepaperwerefertoalistofcase-studymaterialsthroughnumeralsshowninsquarebrackets.Theannexliststhecase-studymaterials.
FlandersInvestmentandTrade(FIT)isthetradepromotionorganisation(TPO)ofFlanders,aregionofBelgium.TPOsarefacilitativeagenciesthatpromoteandstimulatetradebyprovidinginformation,linkages,technicaladvice,marketingandpolicyadvocacy(Giovannucci,2004).Theiractivitiescanbegroupedintofourbroadcategories:productandmarketidentificationanddevelopment;tradeinformationservices;specialisedsupportservices;andpromotionalactivitiesabroad(Jaramillo,1992).FIT’smissionistointernationalisetheeconomyofFlandersbyassistingFlanders-basedcompaniesintheirexporteffort(‘trade’)andbyattractingforeigncompaniesandinvestmenttotheregion(‘invest’).Alongsidedeliveringtradeandinvestmentservices,FITengagesinpromotionalanddevelopmentactivitiesincludingthehostingofeventsandpublicationofmarketinsights.FIThassixregionalofficesinFlandersandBrussels(employingabout150people)and100localofficesabroad(employingabout180people).
RadixisaBelgianAIsolutionprovider,foundedin2018.Ithasateamof40engineersandsolutionleadsacrosstwoofficesinFlandersandBrussels.RadixprovidesaportfolioofAIsolutionstoimproveoperationsinarangeofindustries,includingmanufacturing,transportation,financialservicesandthepublicsector.
1.3Selectionofthecase
ThecasewasfoundthroughthewebsiteoftheAIdeveloper(Radix),whichshowcasesclientstories.SeveralRadixclientstorieswererelevantfortheFutureofWorkandwerethereforeconsidered.Amongthemweretwoclientsinthehumanresourcesandpublicemploymentsectors:anAI-supportedorientationtestdevelopedfortheFlemishpublicemploymentagency,andanAI-poweredjob-matchingalgorithmdevelopedforaprivateHRservicescompany.AIwilllikelyplayamajorroleinmatchingjobseekerstojobvacanciesinthelabourmarketsofthefuture.Boththeopportunitiesand
1See
/future-work/future-work-and-inclusive-growth-europe
.
2See
/
and
https://radix.ai/
.
4
potentialdangersofthisapplicationarecurrentlybeingstudiedanddebatedwidely,withspecificfocusontheriskofincreasingdiscriminationinthelabourmarket.However,inthisparticularcasestudy,thegoalistostudyAInotinthejob-matchingprocess,butintheproductionprocessitself.
FITwashighlightedasaRadixclientthatadoptedAIinoneoftheircorebusinessactivities:answeringtrade-relatedquestionsfromFlemishcompanieslookingtodotradeabroad.OtherclientcasesofthisAIdeveloperwithapplicationsintheproductionprocessincluded:aproductionplanningalgorithmthatimproveson-time-deliveryofproductionorders,takinglesstimethanahumanplanner;analgorithmthatimprovesvaccinedevelopmentbycountingandreportingcolonyformingunits;andanalgorithmthatautomaticallytagsnewarticlesofanewssupplierwithtopicalhashtags.Weselectedthequestion-answeringalgorithmforFITovertheseotherexamplesbecauseitfittedthecurrentnarrativeofAIreplacingroutinecognitivetasksofknowledgeworkers.AnotherreasonwasthatthedevelopernotedintheirFITclientprofilebothproductivityincreases(27percenttimesavings,36percentmorequestionsanswered)andjobsatisfactionimprovements(focusonmorecomplexcasesandotherpartsoftheirjobs)[8–seetheannex],whichfittedourgoalofstudyingbothproductivityandjob-qualityeffects.
1.4Methodology
Thecasewasstudiedthroughthecollectionandanalysisofseveraldatasources.First,deskresearchwasperformedontheexistingscientifictheoriesandevidenceonAIadoptionandacceptance.Thisdeskresearchresultedinthepublicationofseveralblogpostsandpapersonthesetopics(seeforexampleHoffmanandNurski,2021a,2021b).Second,deskresearchwascarriedoutonpubliclyavailableinformationonthecases,mostnotablytherespectivewebsitesofFITandRadix.Inathirdstep,interviewguidesweredevelopedonthetopicsontechnologyadoptionandacceptanceforseveralintervieweetargets.InterviewswereconductedwithFIT’sAIleadandHRlead(see[4],[10],[13])andwithfour‘endusers’ofonespecificAIapplicationatFIT,alsoknownas‘casehandlers’(see[12]).Thefourendusers(twomenandtwowomen)werestationedinfourdifferentoffices:France,Germany,ItalyandtheUSA.Dependingontheinternalorganisationoftheoffice,someoftheintervieweesspecialisedincertainregionsoftheircountry,whileothersspecialisedincertainindustriesinthatcountry.Afinaldatasourceconsistedofcollecteddocuments,includingslidedecks,screenshotsandtrainingmaterials.Thefulllistofcasestudymaterialscanbefoundintheannex.
5
2AIadoptionbytheorganisation
2.1Adoptionprocess
2.1.1Timeline
Aspartofitsdigitalinnovationstrategy(seesection2.2.1),FITisadoptingAIacrossarangeofactivitiesinitsprimaryservices,namelythetradeandinvestservices.Overfouryears(2017to2021),FITwentthroughthreeAIprojectcyclesto:(1)experimentwithproof-of-concepts(POCs),(2)buildanAIstrategy,and(3)set-upthenecessarydatainfrastructure.
Table1:SummaryofphasesintheAIadoptionprocess
Year
Phase
Goal
2017-2019
AIproof-of-concepts
QuickPOCstoexperiment,learnanddiscover
opportunities
2020
AIstrategy
Assessingcurrentas-isAImaturityanddevelopingaroadmaptowardsthedesiredto-bestateofAIadoption
2020-2021
Datainfrastructure
Installrequiredinfrastructureforcentralisingandprocessingallinternalandexternaldatasources.
Source:Bruegelbasedon[4].
2.1.2Phase1:DevelopingAIproof-of-concepts(2017-2019)
Inthefirstphase,FITfamiliariseditselfwithAItechnologytodiscoveropportunitiesandinvestigatewhetheritwouldbeusefultoexplorefurther.AnexternalAIagency(Radix)setupa‘fastdiscoveryworkshop’forFIT’sAIleadtoscreenFIT’sbusinessprocessesforpotentialAIopportunities[7].ThisworkshopconsistedofaseriesofbrainstormingexercisesbetweentheAIdeveloperandtheorganisationlookingtoadoptAI.Firstalonglistofideaswasassembledbygatheringideasfromdifferentstakeholders;nexttheideaswereanalysedandprioritisedinlightoftheirtechnicalfeasibilityandbusinessvalue;finally,effortandvalueestimationsweremadefortheselectedopportunities[14].
Thisprocessgeneratedfiveproof-of-concepts(POCs)forusingAItosupportFIT’scorebusinessservices,namelythetradeandinvestservices.Theyrangedfrominformationgatheringonforeigncompaniesthroughwebscraping,leaddetectionofpotentialclientsthroughsociallistening,andpredictivemodellingformarketingbasedonlikelihoodstoinvestandtrade[4].Thislistofopportunitieswasprioritisedaccordingtotheirbusinessvalueandtechnicalfeasibility(effortandcomplexityofimplementation)(seeFigure1).ThePOCthatcameoutasa‘quickwin’(highvalue,low
6
Value
complexity)wasaquestion-answeringalgorithmforFIT’s‘tradecases’
3
,aimedatpartlyautomatingtheprocessofansweringtradequestionsfromFlemishcompaniesaboutforeignmarkets.Usingnaturallanguageprocessing,trainedonalargedatasetofpasttradequestionsandanswers,thealgorithmwasdesignedtoretrievepastanswerstofrequentlyaskedroutinequestions.The‘highvalue’wasestimatedbecauseofthelargesharethistasktakesupintheworkloadofcasehandlers(namely,60percentto70percentoftheirworkload).The‘lowcomplexity’wasestimatedduetotheavailabilityofhighquality‘offtheshelf’naturallanguageprocessing(NLP)modelsthatcouldbetrainedonFIT’slargehistoryoffiveyearsofpreviouslyansweredquestions(about10,000peryear).Finally,analgorithmwasdesignedtoretrievepastanswerstoroutinequestions,sothatFITadvisorscouldspendmoretimeonthecomplexquestions.TheapplicationactsasanAI-poweredsearchengine,notjustcomparingindividualwords,butinterpretingtheentirebodyofthequestionandfindingthemostrelevantpastanswer.
Figure1:Value-complexitymatrixforprioritisingAIopportunities
Highvalue,lowcomplexity
Quickwins
Highvalue,
highcomplexity
Strategic
initiatives
Lowvalue,lowcomplexity
Lowvalue,
highcomplexity
Complexity
Source:[7].
The‘tradecases’question-answeringPOCwasfurtherdevelopedintoacompleteAIproductbyintegratingthealgorithms’recommendationsintoFIT’sexistingCustomerRelationshipManagementsoftware(CRM),MicrosoftDynamics.Toevaluateandimprovethequalityofthisfirstminimumviableproduct(MVP),thedeveloperconducted10interviewsacrossseveralofFIT’sinternationalofficesandassessedtheresultsfor175newtradequestionsthatwerehandedtotheAI.Ineachcase,thealgorithmsuggestedfivepreviousanswers,meaningabout875AI-suggestedanswerswereevaluated.
3A‘tradecase’isaquestionfromaFlemishcompanyaboutaforeignmarket,thatconcernsservicesofFIT,forexampleinquiriesaboutthesizeorcustomsofalocalmarket,potentialforeignbusinesspartners,traderegulationsorbarriers,subsidies,ormarketopportunities.Seesection
3.1
belowformoredetailonthebusinessprocessandAIsupport.
7
Thedeveloperusedstaffmembers’personalmemoriesofpastcasesbyaskingthemifabetteranswerfromthepastexisted,andthenanalysedwhythealgorithmdidnotretrievethemostrelevantanswer.Justasworkerslearnhowtoimprovetheiranswersovertime,thealgorithmwasretrainedbasedonthecorrectionsofFITstaff.Reasonsformissingbetteranswersfromthepastincluded:unrecognisedsynonyms(sametopicbutdifferentwords),wronglanguage(sametopicbutdifferentlanguage,egEnglish,Dutchorotherlanguage),unclearlink(sametopicbutnotexplicitlymentioned),wrongfocus(AIdidn’tfocusonrightwords),andout-of-vocabulary(AIdidn’tknowcertainwords).Bytakingintoaccountstaffmemberfeedback,thehitrate(casesinwhichtheAIfoundarelevantanswertoaquestion)increasedfrom51percentto62percent[7].Involvingusersinthedesignofthealgorithmthusimproveditsquality(andthereforeuseability,see3.3.2)substantially.
2.1.3Phase2:BuildinganAIstrategy(2020)
ThefirstphaseshowedthatitwaspossibleandopportunetoexpandtheadoptionofAIinawiderrangeofFIT’sprocesses.Inthesecondphase,theytookastepbackfromtheoriginalfivePOCsandtookamorestructuralapproachtoAIbybuildinganAIvisionandstrategy(orAI‘blueprint’)fortheorganisation.Withthehelpofthreeexternalexperts,anAImaturityassessmentwasdone,followedbythedesignofafuturevisionandaroadmaptomovefromtheas-issituationtothedesiredto-bestate[4].
ThemethodologyforbuildingtheAIstrategyconsistedofthreebuildingblocks.First,anenterprisearchitecturewasdrawnup,mappingthecurrentbusinessprocessesonapplications,datalayersandtechnicalsystems.Second,anAImaturityassessmentwasconductedtoassessthe‘as-is’stateofAImaturityandtodevelopanAIroadmapofpotential‘to-be’statesofAIadoption.ThethirdpartoftheAIstrategyrelatedtotrainingandhumanresources.ItincludedsettingupanAIunitresponsibleforAIimpactanddisseminationatFIT,trainingeveryoneatFITonbasicAIliteracy,andspecifictrainingforthedigitalmarketingteamondata-drivenmarketingstrategiesandtools.
Theexternalexpertsclassifiedtheas-isstateofFIT’sAImaturityat‘AIready’,whichisthesecondlevelofmaturityintheirassessmentframework:
•AINovice:AInoviceshavenottakenproactivestepsontheAIjourneyand,atbest,areinassessmentmode.
•AIReady:SufficientlypreparedtoimplementAIintermsofstrategy,organisationalset-upanddataavailability.
8
•AIProficient:AreasonabledegreeofpracticalexperienceandunderstandingofhowtomoveforwardwithAI.Therearestillgapsandlimitations.
•AIAdvanced:AgoodlevelofAIexpertiseandexperience,withaproventrackrecordacrossarangeofusecases.Goodoperationalproceduresinplace.
TheAIroadmaptowardstheto-bestatewasdrawnuptomovethroughthreestates.Inafirststage,FITwoulduseself-servicebusinessanalytics
4
anddashboardingapps(suchasPowerBIandAzuredataservices)andready-madeAIsupportedinsights(forexampleOffice365workplaceanalytics)tobuildadatafoundationandsupportadata-drivendecision-makingculture.Inasecondstage,FITcouldusesolution-specificAIservicesandAI-basedcontentunderstanding(forexamplechatbotsandApplicationProgrammingInterfaces(APIs)toNaturalLanguageProcessing(NLP)models)tobuildanFITconversationalknowledgeplatform.Finally,inthethirdstage,FITcouldadoptadvancedcloudinfrastructuresandopenmachine-learningframeworks,aswellasdeveloptheirowncustomdatascienceanddeepAIcapabilitiestosupportthedigitalmarketingpipeline(forexampleontargetedads,leadsanddirectmarketing).
2.1.4Phase3:Settingupthenecessarydatainfrastructure(2020-2021)
Fromtheassessmentinphase2,itbecameclearthatFITlackedtherequiredinfrastructureforlarge-scaleAIprojectsthat,forexample,requiretheprocessingofunstructureddatainrealtime.Thefirststepintheroadmapthereforeconsistedofbuildingadatahub(ordatavault)forabsorbingdatafromdifferentinternaldatasources[4].TheseinternalsourcesincludedFIT’saccountingsystem,EnterpriseResourcePlanning(ERP)system,website,CRMsystemandtwooldlegacysystemsthatstillfedintotheCRM.Thedatahubwouldalsocentraliseandingestallpurchasesofexternaldata,likecompanydatabases.Ontopofthephysicalinfrastructureforstoringdata,anoperationaldatabaselayerwouldbebuiltaroundcustomers,products,accountsandtransactions.ThisdatalayerwouldfeedintoanAPIaccesslayerthatwouldgrantdifferentbusinessapplicationsaccesstoandmonitortheiruseofthedata.Thisset-upwouldserveasthebasisforallfuturedataconsumption(bothstructuredandunstructured),datasharingandexchange,datamonitoringandaccessmanagement.Bysupportingnearreal-timedataprocessingandreporting,itwouldserveasthefoundationforallfutureAIdevelopment.
4Self-serviceanalyticsisaformofbusinessintelligence(BI)inwhichline-of-businessprofessionalsareenabledandencouragedtoperformqueriesandgeneratereportsontheirown,withnominalITsupport.
(/en/information-technology/glossary/self-service-analytics
).
9
2.2Driversandbarrierstoadoption
Anorganisation’sdecisiontoadoptanewtechnologyisinfluencedbythetechnological,organisationalandenvironmentalcontext(Baker,2012;HoffmannandNurski,2021).AccordingtoaEurope-widecompanysurvey(EuropeanCommission,2020),themainreasonsforfirmstonotadoptAIarealackoffinancialmeans,humancapitalanddataavailability,bothwithinthefirmandfromtheexternalenvironment(HoffmanandNurski2021).Table2listsdriversandbarriersthatwereidentifiedinthiscasestudyineachofthethreecontexts,whilethefollowingparagraphsdivedeeperintoeachofthefactors.
Table2:IdentifieddriversandbarrierstoAIadoptionatFITinthetechnological,organisationalandenvironmentalcontext
Identifieddrivers&facilitators
Identified(overcome)barriers
Technological
context
Expectedproductivitygains
Dataavailability
Hightrialability
LackofcompatibleITinfrastructure
Organisational
context
Leadershipandmanagementsupport
Environmental
context
Competitivepressures
Accesstoskilledlabourand
externalfunding
Source:BruegelbasedonBaker(2012),interviews,documentsandwebsites(seetheannex).
2.2.1Maindriverofadoption:competitiveenvironment
Asasmall,openeconomy,internationalbusinessisakeyfactorintheeconomicdevelopmentofFlanders.In2021,Flandersimported€378.8billionworthofgoodsandservicesandexported€380.5billion,puttingtheFlandersregioninthetop20ofglobalexportercountries(WTOStatsdashboard).Topexportedproductsincludepharmaceutical,chemical,andmineralproducts,andmachinery,electronicandtransportequipment.ThemaintradingpartnersareneighbouringcountriesGermany,FranceandtheNetherlands,andintra-EUtraderepresentstwo-thirdsoftotalexportsfromFlanders[2].WhileseparatenumbersareunavailableforFlanders,exportfromBelgiumasawholesupports843,900jobsinBelgiumoutoffivemilliontotalemployment(Rueda-Cantucheetal,2021).
TPOsaroundtheworldcompeteforlocalinvestmentsbymultinationalcompaniesandneedsophisticatedapproachestoattract,andkeepforeigninvestors(Zanattaetal,2006).FITconsidersdigitalisationakeyfactorinitsstrategytostaycompetitiveinthisinternationallandscape[3].FIT
10
thereforeaimstobean‘earlyadopter’(Rogers,1983)indigitalisation.Theachievementofthisgoalisrecognisedbyitsenvironment,asFITisconsideredoneofthebestpracticesfordigitalisationandAIadoptionbytheEuropeanCommission[4and9].
ThedigitisationofFITreflectsthewiderdigitaltransformationoftheFlemishgovernmentandtheFlemishDigitalStrategy,buildingontheFlemishDataStrategythatwasapprovedon18March2022[5].Whilethedigitalstrategyisstillbeingbuilt,theFlemishgovernmentaimstoreachatop-fivespotintheEuropeanrankingofdigitalpublicservices,asmeasuredbytheDigitalEconomyandSocietyIndex(DESI)[6].
2.2.2Overcomingfinancialbarriers:externalfinancing
Foreachofthethreephases,externalprojectsubsidieswereacquiredforthespecificgoalofdigitalisationandAIadoption,eitherdirectlyorindirectlyfinancedbypublicfunds.Thefirststage(AIPOCs)andthirdstage(datainfrastructure)tookplacewithintheframeworkofFlandersAccelerates,whichisFIT’sinternationalisationstrategyfortheFlemisheconomy.TheexecutionofthisstrategyissupportedbyacombinationofEuropeanandregional(Flemish)funds.Fortheperiod2017-2022,FITreceived€1.8millionfromtheEuropeanRegionalDevelopmentFund(ERDF)and€1.6millionfromtheFundforAccompanyingEconomicandInnovationPolicy(HermesFund),managedbytheFlemishInnovationandEntrepreneurshipagency(VLAIO).BothfundswereawardedspecificallyforFIT’sdigitalisationstrategy.
Thesecondphase(buildingtheAIstrategy)wasspecificallyanddirectlysupportedbytheStructuralReformSupportProgramme(SRSP),managedbytheEuropeanCommission’sDirectorate-GeneralforStructuralReformSupport(DGReform),theEUbodythathelpscountriesdesignandimplementreformsaspartoftheireffortstosupportjobcreationandsustainablegrowth.TheCommissionprovidedsupportovera12-monthperiodintheformoftechnicaladvisoryservicesbyentitieswithsubstantialexperienceinthedevelopmentofblueprintsforAIforpublicadministrations[9].TheadvisoryservicessupportedthethreeelementsoftheAIstrategydiscussedabove,namely:(1)developinganAImaturityassessment;(2)recommendingafuturearchitectureandroadmapforAIdeployment;(3)proposingcurriculaforAI-relatedtrainingofFITstaff.DGreformfeaturestheprojectonitswebsiteasinspirationforotherEUcountries[9].
11
2.2.3Overcominghumanandorganisationalbarriers:hiringandtraining
FollowingFIT’sdigitalisationandinnovationstrategy(see2.2.1)themanagementteamdecidedthat“FITwantedtojointheAItrain”[10].Abusinessandinformationsystemsengineerwithsevenyears’experiencein
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。
最新文档
- 新教材新高考地理鲁教版一轮复习第十五单元第二节资源枯竭地区的发展以德国鲁尔区为例教案
- 新教材高中语文第三单元永遇乐京口北固亭怀古部编版必修上册教案
- 第十讲回归教案
- 部编人教版二年级语文下册《语文园地三》教案
- 部编版一上语文aoe教案资料(2025-2026学年)
- 小班社会朋友夹心饼干教案
- 五年级下册数学教案长方体与正方体的体积沪教版
- 2026年渤海船舶职业学院单招职业技能笔试备考试题及答案详解
- 部编统编二下语文语文园地二公开课试卷教案(2025-2026学年)
- 第十四课沟通中外文明的“丝绸之路”教案(2025-2026学年)
- 2025年大学《电子商务概论》期末试题及答案
- 2025呼和浩特市文化旅游投资集团有限公司招聘工作人员(职能类)20人考试参考题库及答案解析
- 后勤洗刷合同协议
- 2026年海南职业技术学院单招职业技能测试题库及参考答案详解1套
- 浙江省强基联盟2025-2026学年高三上学期二模英语试题(解析版)
- 2026春译林版新版八年级下册英语单词默写表
- 2025至2030中国网球行业市场发展分析与发展趋势及投资风险报告
- 运动员退役协议书
- 太阳能路灯可行性研究报告
- GB/T 7044-2013色素炭黑
- T∕CCCMHPIE 1.44-2018 植物提取物 淫羊藿提取物
评论
0/150
提交评论