Capgemini-配備Gen AI的渦輪增壓軟件-組織如何充分發(fā)揮生成性人工智能在軟件工程中的潛力 Turbocharging software with Gen AI 2024

Turbocharging

softwarewithGenAI

HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

RESEARCHINSTITUTE

#GetTheFutureYouWant

2

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

16

04

Tableof

Content

Chapter1:OrganizationsareΓeapi∩gsig∩i?ca∩tbe∩e?ts

ExecutiveSummary

10

fromleveraginggenerativeAIforsoftwareengineering.

Introduction

CapgeminiResearchInstitute2024

3

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

28

Chapter2:GenerativeAI

adoptionisatanearlystagebutwillacceleratesharply

54

Chapter4:Howcanorganizations

44

harnessthefullpotentialofgenerativeAIforsoftwareengineering?

Chapter3:LackofprerequisitesandunofficialusageofgenerativeAIposesignificant

functional,security,andlegalrisks

CapgeminiResearchInstitute2024

4

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Executive

Summary

OrganizationsarereapingmultiplebenefitsfromleveraginggenerativeAIforsoftwareengineering.

?Theleadingbenefitsfororganizationsareenabling

moreinnovativework,suchasdevelopingnewsoftwarefeatures/services(observedby61%ofsurveyed

organizations),improvingsoftwarequality(49%),andincreasingproductivity(40%).

?OrganizationsusinggenerativeAIhaveseena7–18%

productivityimprovement1inthesoftwareengineeringfunctionasperearlyestimates.Thisishighestfor

specializedtaskssuchascodingassistance2(34%asthemaximumpotentialfortimesavingswith9%onaverage)andcreatingdocumentation(35%asthemaximum

potentialfortimesavingswith10%onaverage).This

researchanalyzedtimesavingsinvarioussoftware

engineeringtasksusinggenerativeAItoolsandnotcostsavingswhichcanbesignificantlydifferent.

?Organizationsareutilizingtheseproductivitygainson

innovativeworksuchasdevelopingnewsoftwarefeatures(50%)andupskilling(47%).Veryfewaimtoreduce

headcount(4%).

?GenerativeAIishavingapositiveimpactonsoftwareprofessionals’jobsatisfaction.

?69%ofseniorsoftwareprofessionalsand55%ofjuniorsoftwareprofessionalsreporthighlevelsofsatisfactionfromusinggenerativeAIforsoftware.

?78%ofsoftwareprofessionalsareoptimisticaboutgenerativeAI’spotentialtoenhancecollaborationbetweenbusinessandtechnologyteams.

CapgeminiResearchInstitute2024

5

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

GenerativeAIadoptionisatanearlystagebutwillacceleratesharply.

Executive

Summary

?AdoptionofgenerativeAIforsoftwareengineeringisstillinitsearlystages,with9in10organizationsyettoscale.

?27%oforganizationsarerunninggenerativeAIpilots,and11%havestartedleveraginggenerativeAIintheirsoftwarefunctions.

?Threeinfour(75%)largeorganizations(annualrevenuegreaterthan$20billion)haveadopted(piloted/

scaled)generativeAIcomparedto23%oftheirsmallercounterparts(annualrevenuebetween$1–5billion).

?Adoption(includingpilots)isexpectedtoincrease

significantlyinthenexttwoyearsfrom46%ofsoftwareworkforceusinggenerativeAItoolstoday(foranykindoftraining,experimenting,piloting,andimplementing,withauthorizedorunauthorizedaccess)toanestimateof85%in2026.

?GenerativeAIisexpectedtoplayakeyroleinaugmentingthesoftwareworkforcewithbetterexperience,toolsandplatforms,andgovernance(assistinginmorethan25%ofsoftwaredesign,development,andtestingworkby2026).

?Codingassistanceistheleadingusecase,butgenerative

AIalsofindsapplicationsinothersoftwaredevelopment

lifecycle(SDLC)activities(testcasegeneration,

documentation,codemodernization,UXdesignassistance,etc.)

?Mostusecaseshaveyettobeadoptedbyamajorityof

organizations(39%arefocusingoncodingassistanceand37%onUXdesignassistanceastopadoptedusecases).

CapgeminiResearchInstitute2024

6

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Executive

Summary

LackoffoundationalprerequisitesandunofficialusageofgenerativeAIposesignificantfunctional,security,and

legalrisks.

?27%oforganizationshavetheplatforms&tools,and32%havetalentprerequisitesinplace,toimplementgenerativeAIforsoftwareengineering.

?Over60%lackgovernanceandupskillingprogramsforgenerativeAIforsoftwareengineering.

?OfthosesoftwareprofessionalswhousegenerativeAI,63%useunauthorizedtools.

?Nearlyathirdoftheworkforceisself-trainingon

generativeAIforsoftwareaslessthan40%ofemployeesarereceivingtrainingfromtheirorganizations.

?Usingunauthorizedtoolswithoutpropergovernanceandoversightexposesorganizationstofunctional,security,andlegalriskslikehallucinatedcode,codeleakage,and

IPissues.

CapgeminiResearchInstitute2024

7

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Executive

Summary

HowcanorganizationsharnessthefullpotentialofgenerativeAIforsoftwareengineering?

?Selectandprioritizehighbenefitusecases.

?Mitigaterisksaroundsecurity,IP/copyrightissues,and

codeleakageusingathoroughriskmanagementapproach.

?TransformyoursoftwareorganizationtoensureoptimalusageofgenerativeAI:

?AugmentyoursoftwareteamswithagenerativeAI

assistant.Amajorityofjunior(53%)aswellassenior

professionals(58%)believethatgenerativeAItools

willaugmenttheirday-to-dayworkwithinthenext

twoyears.Forinstance,generativeAItoolscanhelpjuniorprofessionalslearnfasterandcomeuptospeedquickly,whiletheyallowseniorprofessionalstofocusongroomingjuniorsbyensuringtheirlearningand

retention,solvingcomplexissues,andcollaboratingwithbusiness.

?Identifyrequirementsfornewcapabilitiesandsourcethem.

?PrepareforgenerativeAIusebydeliveringtechnologyprerequisites:

?Buildarepositoryofplatformsandtoolsforaseamlessandaugmentedsoftwareengineeringexperience.

?PrivatelyandsafelycontextualizegenerativeAIassistantswithorganization’sowncontent.

?AdoptameasurementprotocolforgenerativeAIimpactmonitoringandusecaseprioritization.

?Putpeopleattheheartofthistransformationbycreatingalearningcultureatyourorganization.

?Provideupskillingandcross-skillingopportunities.

?Addressemployees’workdisplacementconcerns.

CapgeminiResearchInstitute2024

8

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Whoshouldreadthis

reportandwhy?

Thisreportprovidesinsightsintotheuseof

generativeAIforsoftwareengineeringand

offersrecommendationsthatwillbeusefultoorganizationsacrossindustriesinharnessing

1000+

organizationswithannualrevenue

greaterthan$1billion,representedbyaminimumofonesoftwareprofessionalandonesoftwareleader,arepartofthisresearch.

thefullpotentialofgenerativeAIforsoftwareengineering.

Businessleadersintechnology,IT,product,

strategy,R&D/engineering,generalmanagement,andinnovationwhohaveresponsibilityfor–

andoversightof–theirorganization’ssoftware

engineeringfunctionwillfinditparticularlyuseful.

Thisreportdrawsoninsightsfroma

comprehensivemulti-sectoralsurveyof1,098seniorexecutives(directorlevelandabove)

and1,092softwareprofessionals(including

architects,developers,testers,andproject

managers)fromorganizationswithover$1billioninannualrevenue.Thereportcoversthemajor

considerationsforimplementinggenerativeAIinsoftwareengineeringandincludesin-depthqualitativeinsightsfrom20industryleaders,professionals,andentrepreneurs.

CapgeminiResearchInstitute2024

9

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

ThisreportisapartofCapgeminiResearchInstitute’sseriesonGenerativeAI

GenAIinorganizations-annualresearch

GenAIforsoftwareengineering

GenAIformanagement*

GenAIformarketing

GenAIinR&Dandengineering

GenAIand

consumers

GenAIinsupplychain*

GenAIinmanufacturing*

GenAIincustomerservice*

GenAIandsustainability

GenAIandethics/ trust*

GenAIand

cybersecurity*

*

GenAIinbusinessoperations*

Datamastery*

SpecialeditionofourpremiumjournalConversationsfortomorrowonGenAI*

Tofindoutmore,pleasegoto

/insights/research-institute/

*Upcomingreports

CapgeminiResearchInstitute2024

10

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Introduction

Today,byleveragingthepoweroflargelanguagemodels

(LLMs),generativeAIcanenhancedevelopers’productivity,improvesoftwarequality,andacceleratetimetomarket.

MarcoArgenti,ChiefInformationOfficeratGoldmanSachs:

“GoldmanSachsisusingartificialintelligencetoturnsoftwaredevelopersandothersintosuperhumans.”4

IngenerativeAI,thesoftwareworkforcehasatoolto

acceleratekeytasks(suchasdesign,coding,migrating,

testing,deploying,supportandmaintenance)withminimaleffortandaminimallearningcurve.

Sincethedawnofthemoderncomputerage,therehasbeenadisconnectbetweennaturallanguageand

machinelanguage.Withhardwareandsoftwareadvances,

programminghasevolvedinwavesovertimeandthisgaphasbeguntoclose(seeFigure1).

Thisevolutionnowappearsnearcomplete,asnatural

languagebecomesthelinguafranca.Withrecentrapid

advancesinAIandhigh-performancecomputing,wecan

nowsimply“chat”withcomputersand–throughhuman

supervisionandaccountability–lettheAIassistantaugmenttasksrangingfromprogramming,generatingtestcases

anduserstories,todocumenting,amongothers.AsAndrej

Karpathy,oneofthefoundersofOpenAIandformerdirectorofAIatTesla,famouslyquippedfollowingtheintroductionofChatGPT:"ThehottestnewprogramminglanguageisEnglish”.3

CapgeminiResearchInstitute2024

11

Introduction

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Figure1.

Increasinglevelsofvaluecreationfromevolutionofsoftwaredevelopmentlanguagesandplatforms

Valuecreation

GenerativeAIboost

Evolvementofsoftwaredevelopmentlanguages&platforms

“Thehottestnew

programming

languageisEnglish”

Lowcode/Nocode

Cloudnative

DevOpsAutomation

Python

Java

C++

C-Programming

Cobol

IBM704Assembler

ENIAC

1940195019601970198019902000201020202030

Machine&assemblylanguageHigh-levelprogramminglanguageObject-orientedlanguageDevelopmentplatforms

Source:Capgeminianalysis

CapgeminiResearchInstitute2024

12

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Introduction

However,generativeAIbringsrisksandchallenges.

Uncontrolledusecanleadtohallucinatedcode,IPissues,

privatedataleakages,andsecurityvulnerabilities.Software

engineeringorganizationsneedanewstrategyand

implementationapproachtoharnessthepotentialofgenerative

AIwhilemanagingitsrisks.WiththisresearchweattempttoassesstheimpactofgenerativeAIonthesoftwareengineeringfunction,coveringsuchquestionsas:

?HowwillgenerativeAIimpactthevariousstagesofsoftwaredevelopmentlifecycle(SDLC)?

?HowcanorganizationsquicklyadoptandscalegenerativeAItodriveproductivityandinnovation?

?HowwillgenerativeAIimpactsoftwareengineers’waysofworking?

?WhatarethechallengesforsoftwareengineeringandhowbestcanwemanagetherisksassociatedwithgenerativeAI?

?Howcanorganizationscontinuouslymeasureandoptimize

impactofgenerativeAIontheirsoftwareengineeringfunction?

CapgeminiResearchInstitute2024

13

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Whatdowe

meanby

“GenerativeAIforsoftwareengineering”?

Definingtheterm“software”

Softwareisastrategiccapability,transformingtheway

businessesdesigntheirproductsandservices,function

overall,compete,andprovidevaluetocustomers.Softwareisvitaltomodernbusiness,whetherasaproductitselfor

integratedintoenterpriseappsorproducts.

Therearethreemaincategoriesofsoftware:

?Businesssoftware:Usedbyorganizationstorun,

scale,andoptimizeday-to-daybusinessfunctionsandprocessesand/orinteractwiththeircustomersand

partners.

Therearetwobroadtypesofbusinesssoftware:

?Packagedsoftware:Third-partystandardprogramsgroupedtoprovidedifferenttoolsfromthesame

familyinapackage,commerciallyavailableunderthelicensor’sstandardterms,payablewitheitheraone-offorannualfee.

?Customsoftware:Specific,advancedprogramsdevelopedforaspecificpurposeforanindividualorcompany,whichcanbemodifiedorchanged.Customsoftwareisnotcommerciallyavailable

butisbuiltandoperatedforinternalpurposes.

?Consumersoftware:Solddirectlytoendusers,

consumersoftwareincludesapps,webportals,

andinformationtoolssuchasmaps,financialdata,news,games,andmusicplayers.

?Embeddedsoftware:Apieceofsoftwareto

programhardwareornon-PCdevicestofacilitate

functioning.Thesearespecializedenvironments

andapplicationsforaspecifichardwarestackwithperformance,power,andfunctionalityrequirementandconstraints.

CapgeminiResearchInstitute2024

14

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

GenerativeAIhaspotentialforall

categoriesofsoftware,butthisresearchfocuseslargelyonsoftware

engineeringforcustom,embedded,orconsumersoftwarewhichgoes

throughtheentire

softwaredevelopmentlifecycle.

GenerativeAI’spotentialforsoftwareengineering

Softwareengineeringhasshiftedstronglytowardsgreaterautomationandsimplification,particularlywiththeadventofgenerativeartificialintelligence(generativeAI).Theriseoflargelanguagemodels(LLMs)hasbeenkey.LLMsare

deep-learningAIalgorithmsthatcanrecognize,summarize,translate,predict,andgeneratecontentbybuildingon

verylargedatasets.Theyhavefacilitatedtheincreasingadoptionbyconsumersandorganizationsofsoftwareengineering.

GenerativeAIhasthepotentialtotransformthesoftwareengineeringprocess,asitcanbeintegratedintotechstackstounlocknewfeaturesandupdatesforsoftwarecurrentlyinuse.ManyleadersarestrivingtointegrateAI-enabled

plug-insorincorporateAI-poweredtechnologyintotheirownenterpriseandsoftwareengineeringplatforms.OurpreviousresearchshowsthatgenerativeAIwillassistinwritingoneoutofeveryfivelinesofcodeinthecomingyear.5

GenerativeAI’simpactontheSDLC

Withtheincreasingproliferationofsoftwareinproducts,services,operations;softwareteamsareunderpressuretodelivermore,better,faster.GenerativeAIhasthe

potentialtoyieldbenefitsacrosstheSDLC.Figure2

showssomeofthetasksandactivitiesinSDLCthatcanbenefitfromtheuseofgenerativeAItools.Itisworthnotingthatitisasubsetofallactivitiesencompassing

SDLC.Itcanbeintegratedatanystage–frombusinessneedsanalysisandwritingagileuserstoriestosoftwaredesign,coding,documentation,packaging,deployment,testing,andoperations–augmentingtheworkof

softwareengineersandhelpingincreaseefficiency,improvequality,andenhancejobsatisfaction.

GenerativeAIalsotouchestherolesofmanydata

analysts,businessanalysts,platform/softwaredesigners,andsoftwareengineers,developers,andtester.

CapgeminiResearchInstitute2024

15

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Figure2.

PotentialapplicationareasofgenerativeAIintheSDLC

(DevOps)SoftwareLifecycle

Businessdemand/requirementoanalysisandwriting。

Usecasemodeling

UserstoriesgenerationReverseengineering

Codingassistance(codegeneration,completion)

Unittestsgeneration

Legacycodemodernization

(migration,conversion,etc.)

Codeexplanation

CodedocumentationCodevulnerabilities

Platformprovisioning&con?guration

Softwareobservabilitywith

analysisandrecommendations

Businessdemand

D...e...s...i.g...Co.d...i.ngB...ul.dT...e...s...tRe...l.e...a...seDe...p...l.oyOp...e...r..a...teM.onitor

UX/UIdesignSoftwarearchitectureSoftwarerefactoringSoftwarepackages

con?guration

TestCasegenerationTestDatasets

SoftwarepackagesassemblyReleasenotes

IncidentsresolutionTicketsassistance

(Agile)ProductTeams/(Waterfall)DevelopmentTeams

BacklogandroadmapplanningE?ortestimations

Productvaluestreamperformancerecommendations

Teame?ectivenessanalysisandimprovementTeamcommunicationandcollaboration

Processfacilitation(plannings,retrospective,burndown,etc.)

IndustrializedSoftwareEngineeringPlatform

oAgileProcessManagement/ALMoDeveloperworkplace(IDE)oDevOpsautomationoTestsautomationoGenerativeAIfoundations

Source:CapgeminiResearchInstituteanalysis.

CapgeminiResearchInstitute2024

16

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

01

Organizationsarereapingsignificantbenefitsfrom

leveraginggenerativeAIforsoftwareengineering.

CapgeminiResearchInstitute2024

17

Augmentinginnovationandimprovingsoftwarequalityaretheleadingbenefits.

ThΓeei∩?veoΓga∩izatio∩sseei∩∩ovativewoΓk-foΓexample,developi∩g∩ewfeatuΓesa∩dseΓvicesusi∩gsoftwaΓe-asthebiggestbe∩e?tofge∩eΓativeAlusei∩softwaΓee∩gi∩eeΓi∩g(seeFiguΓe3).OfsoftwaΓe

pΓofessio∩alssuΓveyed,80%believethat,byautomati∩gsimpleΓΓepetitivetasks,ge∩eΓativeAlwillfΓeeuptimefoΓthemtofocuso∩i∩∩ovatio∩a∩dvalue-addi∩gtasks,fosteΓi∩ggΓeateΓcΓeativity.

Ase∩ioΓtech∩icalleadeΓfΓomamulti∩atio∩aldigital

commu∩icatio∩stech∩ologycompa∩yelaboΓates:“OneofthebiggestdriversofgenerativeAIadoptionisinnovation.

Notjustontheproductsidebutalsoontheprocessside.WhileseniorprofessionalsareleveraginggenerativeAIcombined

withtheirdomainexpertiseforproductinnovation,junior

professionalsseevalueinAIprocessandtoolinnovation,andinautomationandproductivityoptimization.”

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Figure3.

O∩ei∩twooΓga∩izatio∩sadopti∩gge∩eΓativeAlseesimpΓoveme∩tsi∩e∩abli∩gi∩∩ovativewoΓka∩dqualityofsoftwaΓe.

Percentageoforganizationsseeingbene?tsthroughtheadoptionofgenerativeAI,asmentionedbysoftwareleaders

Enablinginnovativework(e.g.,developingnewfeatures,servicesetc.)

Qualityofsoftware ProductivityCollaboration

SecurityTimetomarket/reductioninlead-time CostofsoftwaredevelopmentTechnicaldebt

Complianceandriskmanagement

49%

41%

36%

34%

33%

25%

12%

9%

61%

SouΓce:Capgemi∩iReseaΓchl∩stitute,Ge∩eΓativeAli∩SoftwaΓeE∩gi∩eeΓi∩g,Se∩ioΓExecutiveSuΓvey,ApΓil2024,∩=

412softwaΓeleadeΓsthathavescaledupoΓaΓeΓu∩∩i∩gpilotswithge∩eΓativeAli∩softwaΓee∩gi∩eeΓi∩g.

CapgeminiResearchInstitute2024

18

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

GenerativeAIalsoenablesimprovementsinsoftware

quality.Itcanhelpdeliverhigher-qualitycodewithfewer

errorsandimprovementsintestcoverageandquality.Bothfactorsgiveorganizationsaproductivityboostatteamandorganizationallevels.Forexample,EmiratesNBD,alarge

bankinggroupintheMiddleEast,notonlyaccelerated

developerproductivitybyupto20%incomplextasks,butalsoimprovedthecompany’scodequalityby20%byusingGitHubCopilot’scodesuggestions.6

HeadofAIataleadingAustraliantelco,explains:“Withuseofge∩eΓativeA/foΓsoftwaΓee∩gi∩eeΓi∩g,the∩umbeΓoftestcasescouldbei∩cΓeasedby30%,gΓeatlye∩ha∩ci∩gtestcoveΓagea∩dquality.”

CapgeminiResearchInstitute2024

19

FoΓtelecombusi∩esses,ge∩eΓativeAlca∩playasig∩i?ca∩tΓolei∩thedevelopme∩tofsuchdata-poweΓed,i∩∩ovativeapplicatio∩sas∩etwoΓkma∩ageme∩ta∩dmai∩te∩a∩ce

aswellascustomeΓseΓvice/salesappso阡eΓi∩ghypeΓ-

peΓso∩alizatio∩.BTGΓoup’sDigitalu∩ithasa∩Al-poweΓed

pΓoductlifecyclema∩ageme∩tstΓategy.Withi∩fouΓmo∩thsofdeployi∩gAmazo∩’sCodeWhispeΓeΓ,ithadautomated

∩eaΓly12%ofΓepetitivewoΓk,allowi∩gthepilotwoΓkfoΓcetofocuso∩moΓestΓategicgoals.7

SimilaΓly,theΓetaili∩dustΓyisleveΓagi∩gge∩eΓativeAl

togatheΓa∩da∩alyzecustomeΓpΓefeΓe∩ces,competitoΓ

i∩sights,pastsaleshistoΓy,etc.,a∩dcΓeateΓobusta∩dpΓeciseΓequiΓeme∩tsdocume∩tatio∩asthebasisofe∩gagi∩g

customeΓ-faci∩gapps.WayfaiΓ,ahomegoodscompa∩y,is

co∩sideΓi∩gusi∩gge∩eΓativeAltoΓeducethetech∩icaldebtaccumulatedi∩theiΓsoftwaΓestackoveΓyeaΓs.8

TurbochargingsoftwarewithGenAI:HoworganizationscanrealizethefullpotentialofgenerativeAIforsoftwareengineering

Figure4.

Telecoma∩dΓetailsectoΓsseee∩ableme∩tofi∩∩ovativewoΓkasatopbe∩efitfΓomge∩eΓativeAl.

Percentageoforganizationsbysector,whohaveactiveinitiativesandseeenable-mentofinnovativeworkasatopbene?t,aspersoftwareleaders

TelecommunicationsRetail

LifesciencesandhealthcareConsumerproductsGlobal

Hightech

Energytransition&utilitiesBanking

Aerospace&defenceAutomotive

PublicservicesInsurance

76%

74%71%

61%

58%56%

55%

53%52%52%

45%

86%

SouΓce:Capgemi∩iReseaΓchl∩stitute,Ge∩eΓativeAli∩SoftwaΓeE∩gi∩eeΓi∩g,Se∩ioΓExecutiveSuΓvey,ApΓil2024,∩=412se∩ioΓexecutivesthathavescaledupoΓΓu∩∩i∩gpilotswithge∩eΓativeAli∩softwaΓee∩gi∩eeΓi∩g.

CapgeminiResearchInstitute2024

20

OrganizationswithactivegenerativeAIinitiativeshaveseenanaverage

7–18%improvementinproductivityacrossthe

SDLC.

ThoseorganizationsactivelyusinggenerativeAIinsoftware

engineeringhaveseenanaveragetotalproductivity

improvementof7–18%acrosstheSDLCtoday,comparedtonon-usageofgenerativeAI.Theincreasingmaturityoftoolsandprocessesalongwithgrowingprofessionalexperience,meansproductivityislikelytocontinuetoimprove.

Wealsofoundthatproductivityadvantageincreaseswithorganizationsize(seeFigure5).

TurbochargingsoftwarewithGenAI:Howorganizationscanrealizethefull