《凱捷：2024軟件工程中的人工智能研究報告（英文版）（88頁）.pdf》由會員分享，可在線閱讀，更多相關《凱捷：2024軟件工程中的人工智能研究報告（英文版）（88頁）.pdf（88頁珍藏版）》請在三個皮匠報告上搜索。

1、#GetTheFutureYouWantTurbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering#GetTheFutureYouWantHow organizations can realize the full potential of generative AI for software engineeringTurbocharging software with Gen AITable of Co

2、ntent2Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringExecutive Summary 04Introduction10Chapter 1:Organizations are reaping significant benefits from leveraging generative AI for software en

3、gineering.163Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringChapter 2:Generative AI adoption is at an early stage but will accelerate sharply28Chapter 3:Lack of prerequisites and unofficial

4、 usage of generative AI pose significant functional,security,and legal risks44Chapter 4:How can organizations harness the full potential of generative AI for software engineering?54Executive Summary Organizations are reaping multiple benefits from leveraging generative AI for software engineering.Th

5、e leading benefits for organizations are enabling more innovative work,such as developing new software features/services(observed by 61%of surveyed organizations),improving software quality(49%),and increasing productivity(40%).Organizations using generative AI have seen a 718%productivity improveme

6、nt1 in the software engineering function as per early estimates.This is highest for specialized tasks such as coding assistance2(34%as the maximum potential for time savings with 9%on average)and creating documentation(35%as the maximum potential for time savings with 10%on average).This research an

7、alyzed time savings in various software engineering tasks using generative AI tools and not cost savings which can be significantly different.Organizations are utilizing these productivity gains on innovative work such as developing new software features(50%)and upskilling(47%).Very few aim to reduc

8、e headcount(4%).Generative AI is having a positive impact on software professionals job satisfaction.69%of senior software professionals and 55%of junior software professionals report high levels of satisfaction from using generative AI for software.78%of software professionals are optimistic about

9、generative AIs potential to enhance collaboration between business and technology teams.78%4Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringExecutive Summary Generative AI adoption is at an

10、early stage but will accelerate sharply.Adoption of generative AI for software engineering is still in its early stages,with 9 in 10 organizations yet to scale.27%of organizations are running generative AI pilots,and 11%have started leveraging generative AI in their software functions.Three in four(

11、75%)large organizations(annual revenue greater than$20 billion)have adopted(piloted/scaled)generative AI compared to 23%of their smaller counterparts(annual revenue between$15 billion).Adoption(including pilots)is expected to increase significantly in the next two years from 46%of software workforce

12、 using generative AI tools today(for any kind of training,experimenting,piloting,and implementing,with authorized or unauthorized access)to an estimate of 85%in 2026.Generative AI is expected to play a key role in augmenting the software workforce with better experience,tools and platforms,and gover

13、nance(assisting in more than 25%of software design,development,and testing work by 2026).Coding assistance is the leading use case,but generative AI also finds applications in other software development lifecycle(SDLC)activities(test case generation,documentation,code modernization,UX design assista

14、nce,etc.)Most use cases have yet to be adopted by a majority of organizations(39%are focusing on coding assistance and 37%on UX design assistance as top adopted use cases).5Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generat

15、ive AI for software engineeringExecutive Summary Lack of foundational prerequisites and unofficial usage of generative AI pose significant functional,security,and legal risks.27%of organizations have the platforms&tools,and 32%have talent prerequisites in place,to implement generative AI for softwar

16、e engineering.Over 60%lack governance and upskilling programs for generative AI for software engineering.Of those software professionals who use generative AI,63%use unauthorized tools.Nearly a third of the workforce is self-training on generative AI for software as less than 40%of employees are rec

17、eiving training from their organizations.Using unauthorized tools without proper governance and oversight exposes organizations to functional,security,and legal risks like hallucinated code,code leakage,and IP issues.6Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizati

18、ons can realize the full potential of generative AI for software engineeringExecutive Summary How can organizations harness the full potential of generative AI for software engineering?Select and prioritize high benefit use cases.Mitigate risks around security,IP/copyright issues,and code leakage us

19、ing a thorough risk management approach.Transform your software organization to ensure optimal usage of generative AI:Augment your software teams with a generative AI assistant.A majority of junior(53%)as well as senior professionals(58%)believe that generative AI tools will augment their day-to-day

20、 work within the next two years.For instance,generative AI tools can help junior professionals learn faster and come up to speed quickly,while they allow senior professionals to focus on grooming juniors by ensuring their learning and retention,solving complex issues,and collaborating with business.

21、Identify requirements for new capabilities and source them.Prepare for generative AI use by delivering technology prerequisites:Build a repository of platforms and tools for a seamless and augmented software engineering experience.Privately and safely contextualize generative AI assistants with orga

22、nizations own content.Adopt a measurement protocol for generative AI impact monitoring and use case prioritization.Put people at the heart of this transformation by creating a learning culture at your organization.Provide upskilling and cross-skilling opportunities.Address employees work displacemen

23、t concerns.7Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringWho should read this report and why?This report provides insights into the use of generative AI for software engineering and offer

24、s recommendations that will be useful to organizations across industries in harnessing the full potential of generative AI for software engineering.Business leaders in technology,IT,product,strategy,R&D/engineering,general management,and innovation who have responsibility for and oversight of their

25、organizations software engineering function will find it particularly useful.This report draws on insights from a comprehensive multi-sectoral survey of 1,098 senior executives(director level and above)and 1,092 software professionals(including architects,developers,testers,and project managers)from

26、 organizations with over$1 billion in annual revenue.The report covers the major considerations for implementing generative AI in software engineering and includes in-depth qualitative insights from 20 industry leaders,professionals,and entrepreneurs.1000+organizations with annual revenue greater th

27、an$1 billion,represented by a minimum of one software professional and one software leader,are part of this research.8Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringGen AI in organizations-

28、annual researchGen AI for managementData masterySpecial edition of our premium journal Conversations for tomorrow on Gen AIGen AI in supply chainGen AI in manufacturingGen AI for marketingGen AI for software engineeringGen AI in R&Dand engineeringGen AI in business operationsGen AI in customer servi

29、ceGen AI andconsumersGen AI andsustainabilityGen AI andethics/trustGen AI andcybersecurityThis report is a part of Capgemini Research Institutes series on Generative AI*To find out more,please go to https:/ reports*9Capgemini Research Institute 2024Turbocharging software with Gen AI:How organization

30、s can realize the full potential of generative AI for software engineeringIntroduction Since the dawn of the modern computer age,there has been a disconnect between natural language and machine language.With hardware and software advances,programming has evolved in waves over time and this gap has b

31、egun to close(see Figure 1).This evolution now appears near complete,as natural language becomes the lingua franca.With recent rapid advances in AI and high-performance computing,we can now simply“chat”with computers and through human supervision and accountability let the AI assistant augment tasks

32、 ranging from programming,generating test cases and user stories,to documenting,among others.As Andrej Karpathy,one of the founders of OpenAI and former director of AI at Tesla,famously quipped following the introduction of ChatGPT:The hottest new programming language is English”.3Today,by leveragin

33、g the power of large language models(LLMs),generative AI can enhance developers productivity,improve software quality,and accelerate time to market.Marco Argenti,Chief Information Officer at Goldman Sachs:“Goldman Sachs is using artificial intelligence to turn software developers and others into sup

34、erhumans.”4In generative AI,the software workforce has a tool to accelerate key tasks(such as design,coding,migrating,testing,deploying,support and maintenance)with minimal effort and a minimal learning curve.10Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can

35、 realize the full potential of generative AI for software engineeringIntroduction Source:Capgemini analysisFigure 1.Increasing levels of value creation from evolution of software development languages and platforms ENIAC1940Evolvement of software development languages&platformsValue creation19501960

36、1970198019902000201020202030Generative AI boostIBM 704AssemblerCobolC-ProgrammingC+JavaPythonDevOps AutomationLow code/No code“The hottest newprogramminglanguage is English”Machine&assembly languageHigh-level programming languageObject-oriented languageDevelopment platformsCloud native11Capgemini Re

37、search Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringimplementation approach to harness the potential of generative AI while managing its risks.With this research we attempt to assess the impact of generativ

38、e AI on the software engineering function,covering such questions as:How will generative AI impact the various stages of software development lifecycle(SDLC)?How can organizations quickly adopt and scale generative AI to drive productivity and innovation?How will generative AI impact software engine

39、ers ways of working?What are the challenges for software engineering and how best can we manage the risks associated with generative AI?How can organizations continuously measure and optimize impact of generative AI on their software engineering function?However,generative AI brings risks and challe

40、nges.Uncontrolled use can lead to hallucinated code,IP issues,private data leakages,and security vulnerabilities.Software engineering organizations need a new strategy and Introduction 12Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potent

41、ial of generative AI for software engineeringWhat do we mean by“Generative AI for software engineering”?Defining the term“software”Software is a strategic capability,transforming the way businesses design their products and services,function overall,compete,and provide value to customers.Software is

42、 vital to modern business,whether as a product itself or integrated into enterprise apps or products.There are three main categories of software:Business software:Used by organizations to run,scale,and optimize day-to-day business functions and processes and/or interact with their customers and part

43、ners.There are two broad types of business software:Packaged software:Third-party standard programs grouped to provide different tools from the same family in a package,commercially available under the licensors standard terms,payable with either a one-off or annual fee.Custom software:Specific,adva

44、nced programs developed for a specific purpose for an individual or company,which can be modified or changed.Custom software is not commercially available but is built and operated for internal purposes.Consumer software:Sold directly to end users,consumer software includes apps,web portals,and info

45、rmation tools such as maps,financial data,news,games,and music players.Embedded software:A piece of software to program hardware or non-PC devices to facilitate functioning.These are specialized environments and applications for a specific hardware stack with performance,power,and functionality requ

46、irement and constraints.13Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringGenerative AI has potential for all categories of software,but this research focuses largely on software engineering

47、 for custom,embedded,or consumer software which goes through the entire software development lifecycle.Generative AIs potential for software engineeringSoftware engineering has shifted strongly towards greater automation and simplification,particularly with the advent of generative artificial intell

48、igence(generative AI).The rise of large language models(LLMs)has been key.LLMs are deep-learning AI algorithms that can recognize,summarize,translate,predict,and generate content by building on very large datasets.They have facilitated the increasing adoption by consumers and organizations of softwa

49、re engineering.Generative AI has the potential to transform the software engineering process,as it can be integrated into tech stacks to unlock new features and updates for software currently in use.Many leaders are striving to integrate AI-enabled plug-ins or incorporate AI-powered technology into

50、their own enterprise and software engineering platforms.Our previous research shows that generative AI will assist in writing one out of every five lines of code in the coming year.5Generative AIs impact on the SDLCWith the increasing proliferation of software in products,services,operations;softwar

51、e teams are under pressure to deliver more,better,faster.Generative AI has the potential to yield benefits across the SDLC.Figure 2 shows some of the tasks and activities in SDLC that can benefit from the use of generative AI tools.It is worth noting that it is a subset of all activities encompassin

52、g SDLC.It can be integrated at any stage from business needs analysis and writing agile user stories to software design,coding,documentation,packaging,deployment,testing,and operations augmenting the work of software engineers and helping increase efficiency,improve quality,and enhance job satisfact

53、ion.Generative AI also touches the roles of many data analysts,business analysts,platform/software designers,and software engineers,developers,and tester.14Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for softwa

54、re engineeringSource:Capgemini Research Institute analysis.Figure 2.Potential application areas of generative AI in the SDLC.Agile Process Management/ALM Developer workplace(IDE)DevOps automation Tests automation Generative AI foundationsProduct value stream performance recommendationsBacklog and ro

55、admap planningEffort estimationsTeam communication and collaborationTeam effectiveness analysis and improvementProcess facilitation(plannings,retrospective,burndown,etc.)(Agile)Product Teams/(Waterfall)Development Teams Industrialized Software Engineering Platform(DevOps)Software Lifecycle Business

56、demand/requirement analysis and writingUse case modelingUser stories generationReverse engineeringUX/UI designSoftware architectureSoftware refactoringSoftware packages configurationSoftware packages assemblyRelease notesIncidents resolutionTickets assistanceCoding assistance(code generation,complet

57、ion)Unit tests generationLegacy code modernization(migration,conversion,etc.)Code explanationCode documentationCode vulnerabilitiesTest Case generationTest Data setsMonitorPlatform provisioning&configurationSoftware observability with analysis and recommendations BusinessdemandCodingTestDeployOperat

58、eReleaseBuildDesign15Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringOrganizations are reaping significant benefits from leveraging generative AI for software engineering.0116Capgemini Resea

59、rch Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAugmenting innovation and improving software quality are the leading benefits.Three in five organizations see innovative work for example,developing new feat

60、ures and services using software as the biggest benefit of generative AI use in software engineering(see Figure 3).Of software professionals surveyed,80%believe that,by automating simpler repetitive tasks,generative AI will free up time for them to focus on innovation and value-adding tasks,fosterin

61、g greater creativity.A senior technical leader from a multinational digital communications technology company elaborates:“One of the biggest drivers of generative AI adoption is innovation.Not just on the product side but also on the process side.While senior professionals are leveraging generative

62、AI combined with their domain expertise for product innovation,junior professionals see value in AI process and tool innovation,and in automation and productivity optimization.”Source:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executive Survey,April 2024,n=412 software

63、 leaders that have scaled up or are running pilots with generative AI in software engineering.Figure 3.One in two organizations adopting generative AI sees improvements in enabling innovative work and quality of software.Compliance and risk managementTechnical debtCost of software developmentTime to

64、 market/reduction in lead-timeSecurityCollaborationProductivityQuality of softwareEnabling innovative work(e.g.,developing new features,services etc.)Percentage of organizations seeing benefits through the adoption of generative AI,as mentioned by software leaders61%49%41%36%34%33%25%12%9%17Capgemin

65、i Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringGenerative AI also enables improvements in software quality.It can help deliver higher-quality code with fewer errors and improvements in test coverag

66、e and quality.Both factors give organizations a productivity boost at team and organizational levels.For example,Emirates NBD,a large banking group in the Middle East,not only accelerated developer productivity by up to 20%in complex tasks,but also improved the companys code quality by 20%by using G

67、itHub Copilots code suggestions.6Head of AI at a leading Australian telco,explains:“With use of generative AI for software engineering,the number of test cases could be increased by 30%,greatly enhancing test coverage and quality.”18Capgemini Research Institute 2024Turbocharging software with Gen AI

68、:How organizations can realize the full potential of generative AI for software engineeringFor telecom businesses,generative AI can play a significant role in the development of such data-powered,innovative applications as network management and maintenance as well as customer service/sales apps off

69、ering hyper-personalization.BT Groups Digital unit has an AI-powered product lifecycle management strategy.Within four months of deploying Amazons CodeWhisperer,it had automated nearly 12%of repetitive work,allowing the pilot workforce to focus on more strategic goals.7Similarly,the retail industry

70、is leveraging generative AI to gather and analyze customer preferences,competitor insights,past sales history,etc.,and create robust and precise requirements documentation as the basis of engaging customer-facing apps.Wayfair,a home goods company,is considering using generative AI to reduce the tech

71、nical debt accumulated in their software stack over years.8Source:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executive Survey,April 2024,n=412 senior executives that have scaled up or running pilots with generative AI in software engineering.Figure 4.Telecom and retail

72、 sectors see enablement of innovative work as a top benefit from generative AI.InsurancePublic servicesAutomotiveAerospace&defenceBankingEnergy transition&utilitiesHigh techGlobalConsumer productsLife sciences and healthcareRetailTelecommunicationsPercentage of organizations by sector,who have activ

73、e initiatives and see enable-ment of innovative work as a top benefit,as per software leaders 86%76%74%71%61%58%56%55%53%52%52%45%19Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringSource:Cap

74、gemini Research Institute,Generative AI in Software Engineering,Senior Executive Survey,April 2024,n=412 software leaders that have scaled up or running pilots with generative AI in software engineering.Top and bottom productivity ranges are found by the 80th and 20th percentile respectively of indi

75、vidual productivity improvement data.Figure 5.Larger organizations have seen greater productivity improvement with generative AI.Productivity improvement range of a software professional,grouped by organization revenue size Global Average18%15%18%19%19%19%7%4%6%9%9%11%USD 1 billion to USD 5 billionU

76、SD 5 billion to USD 10 billionUSD 10 billion to USD 20 billionUSD 20 billion to 3 years)Improve job satisfactionJunior software professionals(having experience=3 years)Reduce attritionGenerative AI benefits in software engineering extend to job satisfaction and happiness.Our research shows that gene

77、rative AI has a positive impact on software professionals job satisfaction and reduces attrition rates(see Figure 8).Fabio Veronese,Head of ICT Industrial Delivery at ENEL Group:“We are more ambitious.For us,improving development productivity with generative AI is not just about lines of code.It is

78、also about developer experience.”69%Senior software professionals believe that generative AI will have a positive impact on job satisfaction 25Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineerin

79、gMost of the current workforce sees generative AI as a strong enabler and motivator 35%associate it with being“assisted and augmented,”and 24%feel“excited and happy”about its adoption(see Figure 9).While there is currently an emphasis on generative AIs utility in code completion and writing,three in

80、 four senior executives believe it will significantly transform their software engineering organization.Tommy MacWilliam,Engineering Manager for Infrastructure at Figma:“Personalized,natural language recommendations are at the fingertips of all our developers.Our engineers are coding faster,collabor

81、ating more effectively,and building better outcomes.”12Source:Capgemini Research Institute,Generative AI in Software Engineering,Software Professionals Survey,April 2024,N=1,092 software professionals Figure 9.Most of the workforce feels positive about generative AI tools for software engineering.Ho

82、w does the workforce feel as regards to the adoption of generative AI35%Assisted and augmented24%Excited or happy 22%Neutral11%Threatened or afraid6%Stressed2%Frustrated or helpless26Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential

83、of generative AI for software engineeringFabio VeroneseHead of ICT Industrial Delivery at Enel Grids“We are more ambitious.For us,improving development productivity with generative AI is not just about lines of code.It is also about developer experience.”27Capgemini Research Institute 2024Turbocharg

84、ing software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringGenerative AI adoption is at an early stage but will accelerate sharply0228Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full pot

85、ential of generative AI for software engineeringGenerative AI adoption in software engineering is at an early stage,with larger organizations showing higher rates of adoption.Currently,27%of organizations are running generative AI pilots,with 11%leveraging generative AI for software development task

86、s(see Figure 10).The remaining majority are at an earlier stage.Source:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2024,N=1,098 senior executives Figure 10.Only one in ten organizations currently uses generative AI for software engineering.11%We

87、are already using gen AI in our software engineering function34%We are assessing and evaluating potential gen AI use cases2%We have no plans to adopt gen AI in our software engineering functions at least over the coming year27%We are running pilots with gen AI27%We are aware of gen AI potential in s

88、oftware engineering and are strategizing our approach29Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAdoption rates show a strong correlation with revenue size(see Figure 11).Cost of tools

89、,lack of upskilling and training budget,privacy and security concerns,and the high costs of safeguarding against them13 are some of the common deterrents for smaller-sized organizations.Source:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2024,N=1,

90、098 senior executives.Figure 11.Larger organizations are further along their adoption journey than their smaller counterparts.Percentage of organizations that have adopted(piloted/scaled)generative AI,grouped by revenue size23%39%52%75%USD 1 billion to USD 5 billionUSD 5 billion to USD 10 billionUSD

91、 10 billion to USD 20 billionMore than USD 20 billion75%Organizations with more than USD 20 billion of annual revenue have adopted(piloted or scaled)generative AI for software engineering.30Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full pot

92、ential of generative AI for software engineeringSource:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2024,N=1,098 senior executives.Note:Todays data is representative of the survey results,while 2025 and 2026 is estimated on senior executives futur

93、e deployment plans and the trends in unofficial usageFigure 12.It is estimated that more than four in five software professionals will leverage generative AI tools and solutions by 2026 both officially and with unauthorized access Percentage of workforce leveraging generative AI tools in the workfor

94、ce29%17%Today2025202646%59%1.8x85%Estimated future adoptionWorkforce using generative AI tools that are not officially authorized by their organizationWorkforce that has access to generative AI tools officallyAdoption is expected to accelerate more than four in five software professionals will lever

95、age generative AI tools and solutions by 2026As organizations unravel the use cases of generative AI in software engineering and start realizing benefits,adoption is predicted to accelerate(see Figure 12).Nearly half of the workforce(46%)today is making use of generative AI tools for software engine

96、ering in any type of use training,experimenting,piloting,or implementing in real environments.It is expected to nearly double in the next two years.Business users will play a role in generative AIs growth in software engineering.Director at a leading biopharma company:“Not everyone is going to be an

97、 app developer.But generative AI will unlock the capabilities of business users to some extent and make them more independent,allowing them to self-create some code or apps as needed.”31Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potenti

98、al of generative AI for software engineering32Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAs we will discuss further,a considerable segment of the software workforce(63%of those using ge

99、nerative AI)currently leverages generative AI unofficially(see Chapter 3).Organizations overall are increasing their spending to keep up with the growing demand for generative AI(see Figure 13).Source:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2

100、024,N=1,098 senior executives Figure 13.Organizations are increasing their investment in generative AI for software engineering.Average yearly investment in gen AI for software engineering(USD Million)6.89.413.02023Today2025+37%+38%33Capgemini Research Institute 2024Turbocharging software with Gen A

101、I:How organizations can realize the full potential of generative AI for software engineeringCoding assistance is the leading use case,but generative AI is being used in other SDLC activities.Applications of generative AI in software engineering go beyond coding assistance,ranging from test case gene

102、ration to business requirements/demand analysis and writing to legacy code modernization.Of our respondents,70%agree that generative AIs potential extends beyond writing code.We list below some interesting uses cases and applications:Coding assistance:This is the most widely recognized generative AI

103、 use case due to the significant impact it can have on productivity and accuracy.The developer can share target functionality,and the tool uses LLMs to generate the code.Coding assistants such as Microsoft Github Copilot,AWS Amazon Q Developer,Google Gemini Code Assist,and Codeium can generate code

104、from natural language and assist in syntax completion,creation of code boilerplates and prototypes,etc.Based on past data,algorithms can detect source code patterns that are likely to introduce bugs and suggest code changes to help address this.14 UX design assistance:Generative AI allows developers

105、 to test UX designs better and faster.It can capture user inputs,provide recommendations,design customized and engaging user experiences,generate novel design elements,and automate the creation of prototypes to some extent.Yannis Paniaras,Principal Designer,Microsoft Digital Studio:“The AI transform

106、s into the conductor of the user experience.This enables our designers to move away from defining fixed flows and embrace non-deterministic design orchestrated by the AI.”15 Business requirements/demand analysis and writing:In addition to taking natural language inputs from analysts,generative AI ca

107、n analyze large amounts of data,including customer reviews,market research,and industry best practices,to identify user needs and preferences and translate them into functional and system requirements quickly.It can help to track changing requirements and improvise and validate requirements document

108、ation by analyzing it for completeness and ambiguity,providing feedback and suggestions.It can also assist in generating epics and stories,contextualized with organizations private documentation corpus.While this is not one of the top use cases being adopted(as we will see in the next sub-section on

109、 use case adoption)the availability of tools and industry examples indicate the potential of using generative AI tools for this use case.For instance,Siemens equipped field and shopfloor workers to document and report issues in real-time using natural speech which leverages AI to analyze it and rout

110、es it to the appropriate design,engineering,or manufacturing expertsin the specific language they require.16 POPaI is a user story and Jira issue generator leveraging generative AI to create and refine user stories with title,benefit hypotheses,and acceptance criteria.17 A Senior Architect at a Fort

111、une 500 industrial conglomerate:“We are piloting with generative AI to convert whole meetings into user(JIRA)stories with requirements and acceptance criteria and push them into our requirements tool.The added benefit is that when we give requirements for phase two,AI will have the requirements for

112、phase one and can craft/add to them accordingly.”34Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering Legacy code modernization:Generative AI can be a game changer for legacy migration project

113、s,significantly accelerating code migration and reducing downtime.It can automatically convert legacy code to natural language,making it simpler to understand the complex functionalities and business logic embedded in the code.Many legacy applications and systems run on decades-old code,and when the

114、 current workforce cannot understand or enhance the codebase,this leads to a pile-up of technical debt.Generative AI helps organizations start afresh.It can also help document current architecture and optimize target architecture by recommending optimized partitioning,refactoring codebases for cloud

115、,migrating data to cloud,and automating deployment.18 Tesing:In software testing,generative AI accelerates quality engineering and enhances test coverage(the share of code thats tested)and overall testing efficiency.Generative AI quickly understands user stories and requirements,almost instantly app

116、lying proven test design methodologies.It generates comprehensive test cases much faster than manual methods and reduces human error,ensuring thorough and precise testing.Utilizing various testing techniques,such as boundary value analysis,state transition testing,and equivalence partitioning,AI ens

117、ures all scenarios,including edge cases,are comprehensively covered.It adapts to specific project needs,generating highly relevant test cases and handling Fabio VeroneseHead of ICT Industrial Delivery at Enel Grids“Enel Grids uses generative AI to create user acceptance tests and specifically to exp

118、lain test scenarios to businesspeople whose IT knowledge might be limited.”35Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAvanthika RameshDirector of Product Management at Salesforce“Gene

119、rative AI can promote collaboration and cross-functional thinking amongst software engineers.AI can help them understand how their solution fits into the broader solution and how to integrate their code.projects of any size,from small to large-scale applications.Over time,it improves by refining out

120、puts based on feedback and new data,quickly updating test cases to reflect changes in the application.Additionally,generative AI can create a baseline of automation scripts,ensuring a strong foundation for automated testing processes.Fabio Veronese,Head of ICT Industrial Delivery at Enel Grids share

121、s,“Enel Grids uses generative AI to create user acceptance tests and specifically to explain test scenarios to businesspeople whose IT knowledge might be limited.”Deployment and maintenance:Generative AI can assist in activities in the deployment and rollout phase,including infrastructure configurat

122、ion,containerization,and orchestration tasks as well as such maintenance activities as incident and log analysis19(see Figure 15).Director Product Marketing at a leading technology major shared,“Before a product rollout,our company uses generative AI to assist in the detection of any vulnerabilities

123、 that might be there in the code.As a result,what used to take 24 hours earlier,is now accomplished in just one hour!”Reduced technical debt:Generative AI can help control growth of technical debt by generating optimized code/output and by assisting the clean-up and migration of legacy applications.

124、Wayfair relies on generative AI to reduce technical debt accumulated while trying to fix technology problems quickly.This includes legacy code in PHP and old database code in SQL,as well as code written by developers who have left the company.20 Enhanced collaboration:Enabling seamless communication

125、 between team members,as well between technology and business teams,whether communicating requirements or explaining what the code does in natural language.78%of software professionals are optimistic about generative AIs potential to enhance collaboration between business and technology teams.Avanth

126、ika Ramesh,Director of ProductManagement at Salesforce,the leading CRM software company:“Generative AI can promote collaboration and cross-functional thinking amongst software engineers.AI can help them understand how their solution fits into the broader solution and how to integrate their code.Gene

127、rative AI can go beyond coding assistance for developers.Established tools like Microsoft GitHub CoPilot 21,Google Gemini Code Assist22,and Amazon Q Developer23 are catering to quite a few use cases in the SDLC value chain like optimizing code,writing test cases,checking for vulnerabilities,moderniz

128、e legacy code,etc.e.g.,Microsoft is developing the GitHub suite beyond Copilot,with Copilot Enterprise and GitHub Copilot Workspaces that caters to multiple use cases across the SDLC.36Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potentia

129、l of generative AI for software engineeringMicrosoft Copilot Studio has usage around documentation and business analysis,benefiting from“low code”contextualization using organizations own content.Microsofts Sketch2Code has applications in design and modelling as it can translate handwritten user int

130、erface design to valid HTML markup code.24 AWS is extending its Amazon Q suite through Amazon Q Developer not only for coding assistance but also assisting in code modernization,changing application architecture,front end development,among many other use cases.In addition to existing AWS Bedrock ser

131、vices which addressing other SDLC stages through custom solutions.Google is progressively improving upon Gemini Code Assist for rich coding use cases including assisting in generating and debugging code,generating unit tests,troubleshooting code with issues,offering suggestions for optimization,impr

132、oving code readability,etc.in the existing GCP Vertex services which are addressing other SDLC activities through custom solutions.Additionally,software engineering focussed solutions like Atlassian,GitLab are continuously embedding generative AI in their respective solutions(Atlassians ALM,Gitlab D

133、evSecOps)for coding assistance,code modernization,etc.37Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringSDLC phaseBusiness Demand Analysis8bases Archie25 for ideation,blueprint generation,de

134、tailed requirements writing in addition to architecture design support.GPT powered tools like PRD wizard,and WritemyPRD26,etc.for product requirements documentationDesignErasers DiagramGPT to create data flow diagrams,architecture diagrams,etc.27 Mintlify to generate code documentation28TestingDiffb

135、lue for writing test cases.29 Deepcodes Synk for securing code30Deployment and maintenance Grit for generating release notes by analyzing commits,issues,and differences;dependency upgradesLegacy code modernizationBloop AI to help teams modernize,and understand their legacy code32DevOpsOpenText DevOp

136、s Aviator for faster application delivery33Illustrative list of startup solutionsSapient for generating test cases.31In addition,as illustrated in the table below,there are many upcoming specialized tools from startups that cater to specific use cases:38Capgemini Research Institute 2024Turbocharging

137、 software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringFigure 14.Generative AI can deliver a range of benefits throughout the SDLC.As Figure 14 shows,our survey highlights the areas where generative AI can deliver a high,medium,and low levels

138、of benefit.Business demand/requirement analysis and writingTestRelease DeployMonitorCoding Business demandDesignUser stories generationRFP preparation and proposal evaluationTest cases generation/Testing code generationUser acceptance testing Code review and quality assuranceCoding assistance(code g

139、eneration completion,etc.)Code explanationUnit tests generationLegacy code modernization(transition)migration,etc)Identification of software securicy vulnerabilitiesUX design assistanceArchitecture writing assistance and modellingApplication IntegrationPlatform provisioning and configuration(Infrast

140、ructure as code script writing support)Debugging and error predictionLog analysisIncident analysis and resolutionBuildHigh benefitMedium benefitLow benefit(DevOps)Software LifecycleSource:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executive and Software Professional Su

141、rvey,April 2024,N=1,092 organizations Note:This chart shows a selection of use cases that we evaluated in greater detail in the research.Figure 2 gives a longer list of all potential use cases for generative AI in software engineering.Both these figures represent a subset of all SDLC activities.39Ca

142、pgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringSource:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executive Survey,April 2024,N=1,098 senior executives;Software

143、Professionals Survey,April 2024,N=1,092 software professionals.Note:Adoption and implementation refers to any scale-from trial/pilot,to a scaled implementation.Figure 15.Top generative AI use cases for software engineering by rate of adoptionGlobal AverageCode review and quality assuranceDebugging a

144、nd error predictionUser story generationRFP preparation and proposal evaluationUX design assistanceCoding assistance(generation,completion)Percentage of organizations focusing on the following use cases for implementation(pilot or scaled)39%37%33%30%29%29%24%Most use cases are in early stages of ado

145、ption.The adoption of most use cases is at a nascent stage,and few organizations are implementing or piloting them.Coding assistance,the highest adopted use case,is at 39%,while test case generation stands at 26%(see Figure 15).Globally less than one in four organizations are focusing on any use cas

146、e on average.This may be attributable to the relatively recent rise of generative AI technology,people mindset,organizations being slow to provide the necessary tools,training and governance,and poor selection of use cases(more on this in the final chapter).40Capgemini Research Institute 2024Turboch

147、arging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringCoding assistance:77%of organizations agree that using generative AI for code assistance brings significant benefits.Duolingo saw a 25%increase in developer speed using CoPilot,GitHu

148、bs tool.It increased developer productivity by limiting context switching,reducing the need to manually produce boilerplate code,helping developers focus on complex tasks.34 Duolingo Chief Technology Officer Severin Hacker:“A tool like GitHub Copilot is so impactful because suddenly engineers can ma

149、ke changes to other developers code with little previous exposure.”Avanthika Ramesh from Salesforce:“Engineers have been saying that they dont need to search for certain code snippets as theyre already on hand.Theyre pleasantly surprised about how intelligent AI is!”Test case generation:61%of organi

150、zations agree that using generative AI for test case generation can give significant benefits.Goldman Sachs was able to increase legacy code test coverage from 36%to 72%in less than 10%of the time it would take to do so manually,thus ensuring higher application stability and faster integration.Avera

151、ge time to write each unit test came down from 30 minutes to 10 seconds.35Qualtrics,a Seattle-based company that builds experience management software,reported that unit tests,which previously took a full day to write,take 10 minutes with their generative AI tooI.36Legacy code modernization:The main

152、 benefit of generative AI is that it makes such large and previously unthinkable migration projects feasible.Peter Schrammel,CTO and Cofounder of Diffblue,a leading generative AI startup for software engineering says:“The big gain for application modernization and migration is being able to undertak

153、e such big projects.It gives you confidence in reaching the right test coverage at lower cost and faster,making it affordable for large enterprises and drastically reduces the time taken.”Companies like ADP,UK-based payroll and HR systems provider,and JPMorgan Chase already use generative AI for sys

154、tem modernization and migration.37Nicole Onuta,Manager Artificial Intelligence Lab at a large financial services company in Netherlands says,“I have seen a huge benefit in transitioning the code from a legacy programming language to a new one using generative AI.38 Similarly,tech company in the heal

155、th and security sector,Leidos,has reduced by 50%the time spent dealing with legacy code using generative AI tools.Migrating a certain piece of code from Oracle to PostgreSQL which would have otherwise taken a full sprint could be 8090%completed,with use of generative AI,in a matter of minutes.39Nico

156、le OnutaManager Artificial Intelligence Lab at a large financial services company in Netherlands“I have seen a huge benefit in transitioning the code from a legacy programming language to a new one using generative AI.41Capgemini Research Institute 2024Turbocharging software with Gen AI:How organiza

157、tions can realize the full potential of generative AI for software engineeringSource:Capgemini Research Institute,Generative AI in Software Engineering,Software Professional Survey,April 2024,N=1,092 software professionals.Figure 16.Generative AI has the potential to assist with more than a quarter

158、of software design,development,and testing work in two years.Percentage share of work that can be augmented or assisted by Generative AI by the next two years,grouped by role28%26%25%16%12%Software design(e.g.,software designers/architects,Business analysts)Software development(e.g.,developers,UI/UX

159、 developers)Testing and quality(e.g.,testers,quality engineers)Operations(e.g.,database operators,software maintenance engineers)Managerial(e.g.,project owners,product managers)Generative AI will play a larger role in assisting the software workforce within the next two yearsThe software workforce i

160、s optimistic.They expect over a quarter of all work in software design,development,testing,and quality to be augmented and assisted by generative AI in two years(see Figure 16).Senior director of software product engineering at a global pharmaceutical major:“Currently,assisting in code generation an

161、d writing test cases are number one priority,but use cases like bug fixing and documentation are fast emerging,with others like UX design,requirement writing,etc.just around the corner.”42Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full poten

162、tial of generative AI for software engineeringPeter SchrammelCTO and Cofounder of Diffblue“The big gain for application modernization and migration is being able to undertake such big projects.It gives you confidence in reaching the right test coverage at lower cost and faster,making it affordable f

163、or large enterprises and drastically reduces the time taken.”43Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringLack of prerequisites and unofficial usage of generative AI pose significant fu

164、nctional,security,and legal risks0344Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringFigure 17.Platforms and tools,along with talent,are the top prerequisites currently lacking for the imple

165、mentation of generative AI.People and talent(e.g.,literacy in the organization,trainings,skills and capabilities)Processes and workflows(e.g.,quality assurance and feedback mechanisms,feedback cycles,collaboration and support)Cybersecurity principles and governanceRegulatory complianceCybersecurity

166、principles and governanceProcesses and workflows(e.g.,quality assurance and feedback mechanisms,feedback cycles,collaboration and support)People and talent(e.g.,literacy in the organization,trainings,skills and capabilities)Platform and tools(e.g.,IDEs platforms,collaboration tools,automation tools,

167、testing tools,platforms,etc.)Percentage of software professionals who state that their organizations have these prerequisites in place27%32%38%40%44%44%49%68%To fully leverage the potential of generative AI,software organizations need to provide fundamental prerequisites,such as platforms and tools

168、and processes/workflows.Less than a third of organizations have the platforms and tools or people and talent needed to implement generative AI for software engineering.Only around a third of organizations are suitably equipped(see Figure 17).Most lack the prerequisites for implementing generative AI

169、 for software engineering aside from culture and leadership,which two in three organizations(68%)claim is in place.Senior director of software product engineering at a global pharmaceutical major:“One challenge towards generative AI adoption is the lack of tools which can be completely integrated in

170、to the software workflow.In the initial days of adoption this was a bigger concern,but as generative AI is maturing,its not long before vendors will be able to mitigate this issue satisfactorily.”Source:Capgemini Research Institute,Generative AI in Software Engineering,Software Professionals Survey,

171、April 2024,N=1,092 software professionals45Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringSustainability in generative AI for software is frequently overlooked.Environmental sustainability

172、is a key concern for organizations worldwide,but the impact of using generative AI in software engineering is underappreciated by most software professionals.40%believe generative AI use in software engineering will increase the carbon footprint of their organizations software function,whereas 35%di

173、sagree indicating confusion about the impact.Only 19%of software professionals rank carbon footprint and impact on sustainability as a top five challenges of implementing generative.Only 10%of organizations cover generative AIs carbon footprint in any reskilling program.64%of organizations have no g

174、overnance framework in place to cover sustainability and carbon footprint monitoring related to generative AI for software initiatives.This indicates a need to bring sustainability into leadership discussions,define proper safeguards for generative AI use and consider sustainability when selecting t

175、ools and approaches.46Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringFigure 18.More than three in five organizations have not implemented a governance framework for generative AI.Yes39%No61

176、%Has your organization decided on a governance framework for Gen AI for your software engineering function/processMore than three in five organizations lack a governance framework and upskilling program for generative AI.Most organizations(61%)lack a relevant governance framework and upskilling prog

177、ram(see Figure 18).A governance framework that defines standards and guidelines to ensure generative AI implementation aligns with organizational priorities and objectives is critical to success.Key aspects to address include code provenance,reducing bias,explainability of the model and output,IP/co

178、pyrights,dependency on external platforms,data/code leakage,access policies,etc.Lack of proper governance and upskilling programs can lead to poor-quality and hallucinated code,and can make the organization vulnerable to risks like code leakage,IP issues,exposure to malicious actors,etc.Source:Capge

179、mini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2024,N=1,098 senior executives47Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAmong the 39%of o

180、rganizations with a governance framework,there remain considerable gaps:61%have governance for cybersecurity issues.60%have governance for productivity and quality KPIs to assess and measure generative AI output and success.57%have governance for ethical issues,such as lack of transparency,bias,etc.

181、Most organizations(61%)also lack an upskilling/reskilling program(see Figure 19).Our research has shown that organizations with an upskilling program to prepare the workforce are better able to improve their workforce productivity.40Mousumi Bhattacharya from Centene shares,“Generative AI in software

182、 engineering is still a new field;getting used to the tools is most important so that people are not scared of it.You need to make sure that the workforce is properly trained on the tools as well as on issues of ethics and compliance.People need to understand that these trainings will make them perf

183、orm better and make them more needed in the workforce.”Figure 19.Nearly three in five organizations have no upskilling/reskilling program for generative AI.Yes39%No38%Not yet,but were working on it 22%Percentage of organizations that have developed a reskilling/upskilling program for generative AISo

184、urce:Capgemini Research Institute,Generative AI in Software Engineering,Senior Executives Survey,April 2024,N=1,098 senior executives48Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringMore th

185、an three in five software professionals using generative AI are doing so without organizational approval.More than three in five software professionals(63%)using generative AI,use it in an unauthorized manner(see Figure 20).Some large organizations deploy tools and processes to check such“shadow”use

186、.A senior technical leader from a multinational digital communications technology company share,“Organizations are deploying AI-specific software firewalls to check if software professionals are using their personal access to generative AI tools to generate code.Other include allocating API keys,whi

187、ch give access to the Microsoft Azure OpenAI,for example,and procuring a pool of licenses to be reallocated across teams based on usage.”Figure 20.More than three in five software professionals using generative AI use unauthorized tools and solutions.I am using a licensed gen AI tool(s)provided by m

188、y current organization37%I am using a gen AI tool(s)which is not officially authorized and supported by my organization63%Distribution of generative AI tools using workforce by the type of usageSource:Capgemini Research Institute,Generative AI in Software Engineering,Software Professionals Survey,Ap

189、ril 2024,N=1,092 software professionals 49Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringOur research also shows that nearly nine in ten professionals who use generative AI in the public se

190、ctor and in the insurance industry use unauthorized tools and solutions.This would lead to adverse effects without proper governance and human oversight.Source:Capgemini Research Institute,Generative AI in Software Engineering,Software Professionals Survey,April 2024,N=1,092 software professionalsFi

191、gure 21.Public service and insurance have the highest percentage of the workforce using unsupported or unauthorized generative AI tools and platforms.63%Software Professionals are using generative AI tool(s)not officially authorized and supported by their organizationPercentage of workforce using un

192、offical and unsupported generative AI solutions and tools,by industry88%87%81%70%67%Public servicesInsuranceLife sciencesBankingEnergy transition&utilitiesGlobalAerospace&defenceTelecommunicationsConsumer productHigh-techRetailAutomotive63%62%58%58%48%48%47%50Capgemini Research Institute 2024Turboch

193、arging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringThe workforce needs to rely on self-training for generative AI,as only two in five organizations are upskilling their employees.Software professionals are upskilling themselves indep

194、endently on generative AI tools,covering the lack of organizational support.As shown in Figure 22,not more than 40%of software professionals are getting trained by their organizations on relevant generative AI tools.Source:Capgemini Research Institute,Generative AI in Software Engineering,Software P

195、rofessionals Survey,April 2024,N=1,092 software professionals.Figure 22.A third of software professionals are training themselves independently on generative AI toolsMy organization is getting me trained on relevant gen AI tools40%I am getting trained on gen AI tools independently32%I am not getting

196、 any training on gen AI tools at all 28%Distribution of software professionals based on the training they receive51Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringUsing unapproved tools with

197、out proper training and safeguards exposes organizations to risk As seen in Figure 23,legal concerns,functional risks,and data security concerns are the top challenges senior executives associate with implementing generative AI.A recent breach,where employees accidentally shared a piece of source co

198、de in the public domain,illustrates these risks.Figure 23.Various functional and legal risks that underlie generative AI use in an organization.Functional risksTrust&CorrectnessGenerative models have no guarantee at all about correctness,and under“hallucination”sound confident even if factually wron

199、g Inherited riskBuilding on top of a foundational model that is not well understood means that downstream systems inherit unknown risks BiasGenerative models will reflect the biases present in their training data and can be made to be deliberately biased with certain prompts Sustainability Generativ

200、e AI models can require a huge amount of energy both in their initial training and their operational use Legal risksPrivacyCurrent models are often trained without a legal basis for all training data.Using the output with such data will repeat the infringement.User prompts may jeopardize privacy rig

201、hts Data and code leakageCurrent models often memorize their training data,which can leak out either accidentaly or with deliberate prompting in the future Intellectual propertyCurrent foundational models are often trained on data which may be copyrighted or restricted by license.Generative models m

202、ight regurgitate this copyyrighted dataEthics Current models are purely statistical predic-tors and have no inherent model of knowl-edge,ethics or culture.Ethical issues may lead to unintended outcomes and undermine customer trust.Source:Capgemini Research Institute analysis.52Capgemini Research Ins

203、titute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringWith software professionals adopting generative AI tools without organizational approval and oversight,these challenges increase.Two-thirds of our respondents agree

204、 that generative AI can bring unintended consequences unless properly validated through a regression test framework.Even more agree that unsupervised generative AI output can lead to complicated debugging,edge cases,and bugs,which can disrupt project timelines and delivery.The following are key risk

205、s of adopting generative AI without proper governance,guidelines,and training:Security vulnerabilities:Code generated with unauthorized AI tools expose internal systems and applications to malicious actors,which can lead to data leakage,unauthorized access,and cyberattacks.It may also expose the int

206、ernal code to competitors and third parties by making it part of the training data.Reliability concerns:Ensuring reliability and quality of the generated output is a challenge when employees use unauthorized tools without proper validation and system checks.Organizations should double down on data g

207、overnance and prioritize visibility regarding the quality and clarity of the provenance of any data used.Potential legal issues:If AI models are trained on copyright or proprietary code or data,this may cause legal complications and IP issues.A class-action suit has been filed against Microsoft and

208、GitHub,stating that the tool predictively generates code based on what the programmer has already written.The plaintiffs allege that Copilot copies and republishes code from GitHub,disregarding the requirements of the latters open-source license.41When an organization officially approves the use of

209、generative AI,it would generally address such aspects.But individual employees may well overlook these issues without proper training and education on the reliability,security,legal,and ethical aspects of using generative AI.32%Software professionals are getting trained on generative AI tools indepe

210、ndently53Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringHow can organizations harness the full potential of generative AI for software engineering?0454Capgemini Research Institute 2024Turbo

211、charging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringConsidering the outcomes of the research and our experience of helping clients experiment with,scale,and achieve more with generative AI for software engineering,we believe that a

212、calculated and step-by-step approach,as outlined below,has a high chance of harnessing generative AIs full potential.Select and prioritize high benefit use cases.Organizations must prioritize these for quick wins.Use case with high potential for productivity improvement and other benefits such as hi

213、gher quality among others,include:Creating literature and documentation Coding assistance Debugging and testing Identifying security vulnerabilities Code modernization(including code translation,code migration,and code conversion)RFP preparation and evaluation39%of organizations are focusing on code

214、 assistance despite it being high benefit use casesThe first task is to assess the current software engineering lifecycle to highlight key opportunities to generate value for the software organization and workforce.Real experimentation is the bedrock of this implementation exercise,starting with a f

215、ew teams,then extending across multiple domains in the organization.In parallel,a governance process involving key stakeholders from across IT and business teams must decide which use cases to prioritize,based on their risk-weighted rewards(more on this in the next point).The governing body must inc

216、lude a central design authority for designing solutions,controlling their implementation and progressive deployment.55Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering Mitigate risks around s

217、ecurity,IP/copyright issues,and code leakage using a thorough risk management approach.Generative AI use entails risk,which must be addressed at the planning stage.For each use case selected,you should start with a risk assessment(see Figure 24)and involve your organizations legal,IP,cybersecurity,a

218、nd data protection experts early on.Involve and inform other stakeholders(across business,strategy,product,marketing,and sales,as relevant)about potential risks,ascertain an acceptable risk level,and conduct a scenario analysis to simulate outcomes.Mousumi BhattacharyaDirector of IT at Centene,a US-

219、based managed care company“Generative AI for software engineering is still a new field;getting used to the tools is most important so that people are not scared of it.You need to make sure that the workforce is properly trained on the tools as well as on issues of ethics and compliance.People need t

220、o understand that these trainings will make them perform better and make them more needed in the workforce.”56Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringCapgemini Research Institute ana

221、lysis.Figure 24.A risk assessment framework to kickstart generative AI implementation in software engineering Are you planning to use generative AI?Have the risks(see right)been identified,assessed and the intended use approved by the relevant stakeholders?Can the inputs or prompts be used with an a

222、cceptable risk and with the permission of stakeholders?Does it matter if the output of the solution is comprehensively correct?Do you have the expertise and processes to verify if the output is actually correct?Low risk for internal or external uses(which is not the same as”no-risk”,consult with leg

223、al to check)Use with caution(verifying each output according to appropriate governance procedures and implementing the required mitigating actions)Not safe to usePerform a risk assessmentIterative processYesYesYesYesYesNoNoNoNoNo57Capgemini Research Institute 2024Turbocharging software with Gen AI:H

224、ow organizations can realize the full potential of generative AI for software engineeringA software architect from a leading multinational bank says,“As we move from proof-of-concept(POC)to the roll out phase,assessing and documenting the risks becomes important.Selecting the right code for generati

225、ve AI to work on,along with the right checks in place can help mitigate such risks like proprietary code leakage.”From a cybersecurity perspective,most of the existing application security policies still apply to generative AI applications.Some of the important and relevant application security prac

226、tices to follow are:Use threat modelling to identify the security gaps as generative AI is more likely to use modern development approaches(such as DevOps)and architectures(such as service mesh).Preventing the usage of libraries,frameworks and languages with known vulnerabilities from unknown source

227、s by enforcing strict security standards and regular audits.Mandating secure coding standards by frequently scanning the code for known vulnerabilities Ensuring data security,especially the use of sensitive data during the development process.For example,during coding use cases it is integral to pre

228、vent leakage of secrets in public code repositories.Use automated quality checks for secret detection scanning and manual review.To guard against the risks posed by hallucination of generative AI,software professionals must exercise extra caution to recognize these inaccuracies.They must always revi

229、ew code and perform code scanning through built-in tools or third-party tools that detect vulnerabilities or incorrect code.Using prompt engineering,Retrieval-augmented generation(RAG),or fine-tuning can also reduce risks of hallucination besides making the code more contextual and relevant.More on

230、this in the points that follow.While these practices are necessary,they may not be sufficient to guard against all risks.The first line of defense starts with people.Users should be trained enough to use a generative AI system responsibly and have enough expertise to review the output for risks.Tran

231、sform your software organization to ensure optimal usage of generative AI.Traditional approaches focus on incremental improvements in lifecycle productivity and quality.Generative AI allows us to completely transform the traditional approach by asking questions:How to understand the full impact of g

232、enerative AI on the software organization?How to select and plan the right use cases that have highest benefit yet ease of implementation?58Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering H

233、ow to measure(and thus,manage)value realization from the start?How to contextualize generative AI tools with organizations internal or proprietary content?Attempting to answer these questions holds keys to harnessing the most potential of generative AI,multiplying software teams productivity while a

234、lso delivering on quality,security,and other important parameters.Here are a few ways to make it happen:Augment your software teams with a new member a generative AI assistant.Generative AI has applications and benefits across the SDLC,and the accuracy and usefulness of its output is continually inc

235、reasing.Why not embed generative AI assistants in every software team?AI-augmented software teams can automate mundane tasks,expedite debugging,and promote more nuanced collaboration,thereby optimizing the software lifecycle productivity and improving learning curves both the human and AI side(see F

236、igure 25).Figure 25.Concept of a generative AI-augmented software team.Software teamAugmented software teamTeam/Tech leadsLead|Advise|ValidateTeam/Tech leadsLead|Advise|ValidateGen AI SWE assistant Pair with AISoftware engineers/business analystsDev to Ops activitiesSoftware engineers/business analy

237、stsDev to Ops activitiesPrompt|Completion|ValidationSoftware teamSoftware teamAssistance available at any software stepMore productivity and quality in software engineeringMore availability on business require-ments,design and team managementMore availability on business+design activitiesIndustriali

238、zed software engineering platformEnriched with generative AI assistantsSource:Capgemini Research Institute analysis.59Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringEvery professional on an

239、 augmented software team can benefit from the generative AI assistant junior or senior.In our survey,a majority of junior(53%)as well as senior professionals(58%)believe that generative AI tools will augment their day-to-day work within the next two years.For instance,for junior developers,generativ

240、e AI tools offer a faster route to learning new programming languages and techniques that they may not be familiar with and pick up nuances of programming in the organizations context.Whereas senior developers can focus on grooming juniors,sharing their experience and ensuring that juniors are learn

241、ing and retaining their knowledge,reviewing output of junior developers developed independently or with generative AI assistance,solving complex issues,and collaborating with business counterparts to drive higher value.This approach can help software teams retain the foundation principles of Agile a

242、nd DevOps collaboration,adaptability,time to value,product-centricity,and continuous feedback loops while benefiting from new ways of working,such as:Augmented pair programming:The basic unit of an augmented software team is pairs of software engineers working intermittently with each other or indiv

243、idually with the AI assistant,each helping the other.Software professionals review the generated output and enhance its quality.AIhuman collaboration:Team members work individually with the AI assistant to automate repetitive tasks,understand and solve problems,or brainstorm ideas,leveraging an LLMs

244、 speed and knowledge while ensuring human creativity.It can also assist in training junior professionals.As shared by Director of Product Marketing at a multinational technology major,“This is a great onboarding and training tool to help the new hires get up to speed on how to use the companys softw

245、are language.This is one of the areas where we found it to be immensely beneficial.”Cross-functional collaboration:Generative AI assistants help cross-functional teams(such as IT and business)collaborate seamlessly by helping them communicate in each others language.Senior coordination:A senior lead

246、 oversees the augmented team,coordinating efforts and resolving conflicts while ensuring control and validation to promote a smooth workflow and healthy team dynamic.60Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative A

247、I for software engineering Identify requirements for new capabilities and source them.As generative AI performs routine tasks well,it necessitates a rethink of established processes.New capabilities will be needed.As a senior director from a global pharmaceutical major put it,“Software teams would n

248、eed to develop advanced understanding around how generative AI works,understand its strengths and weaknesses and distinguish smartly where generative AI will work and where it wont.Those who crack this code can turn into super developers with generative AI!”For example:Ability to pair with AI:The mo

249、st immediate new capability needed is the ability to interact with generative AI assistants to analyze their output,validate it or iterate on it until an acceptable solution is obtained.More experienced software professionals will find it helpful working hand in hand with a generative AI assistant,m

250、uch like“pair programming”in which two programmers work together on one workstation one programmer writing code,while the other reviewing the code as it is typed in.Leading generative AI assistants/agents and overseeing their work:As generative AI assistants evolve into autonomous agents capable of

251、reliably assisting on specific,routine tasks end-to-end,human teams must oversee multiple AI agents,coordinating their output and resolving/preventing conflicts.For team leads and project managers,it translates to a greater focus on business requirements,design,and team management.Developing new res

252、ources such as prompt libraries and playbooks for software teams and keeping them up to date.A prompt library is a collection of detailed and documented prompt engineering patterns to apply across the full software development lifecycle,further saving software professionals time to come up with usef

253、ul and efficient prompts for generative AI assistants.Playbooks are guideline documents for software teams reference to support and advise on secure ways of working with generative AI assistants.Organizations will also have to contextualize the generative AI assistants and extend their functionality

254、 using their private,sanitized data for use cases under trial/implementation.This would lead to more relevant,robust,and efficient assistance from generative AI.Fixing complex bugs and root-cause analysis of problems:Human software teams will still be needed to fix complex coding problems given thei

255、r inherent advantage and analyze root causes of issues that generative AI cant solve or prevent.This capability will only grow in future as human programmers time is redirected towards complex problems,customer interaction,and debugging.Prepare for generative AI use by delivering technology prerequi

256、sites.Build a repository of platforms and tools for a seamless and augmented software engineering experience.The biggest gap in essential prerequisites is usually access to platforms and tools,including integrated development environments(IDEs),automation and testing tools,and collaboration tools(on

257、ly 27%of organizations claim to have above-average availability of these).Of testing domain 27%Organizations have the platforms and tools to implement and scale generative AI for software engineering61Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize t

258、he full potential of generative AI for software engineeringSource:Capgemini Research Institute analysis.Figure 26.A reference framework of foundational platforms and tools for a seamless software engineering experienceAugmented software teamsSoftware lifecycleDeveloper workspaceSoftware engineersBus

259、iness analystsAgile specialistsCommon foundationsDevSecOps platformQuality engineering and testingOperations,trust,and securityOrchestration APIs and guardrailsHub of knowledge for accelerating software teams e.g.,playbooks,libraries,prototypes,blueprints etc.Foundation LLMs customized with propriet

260、ary knowledgeDeployment and hostingGenerative AI assistants for software engineeringWorkspace toolsCoding assistants,custom assistants,e2e software assistants and agentsAgile teams|IDE|Source code management|Code analysis|Development testingBusiness demandDesignCodingBuildTestReleaseDeployOperateMon

261、itorprofessionals,24%say they have access to these tools,compared to 19%of project and program management professionals,further highlighting gaps within the software engineering function.In addition,only about half of the organizations surveyed have access to the computing infrastructure and support

262、 needed to fully leverage generative AI tools.Here again,significantly fewer(40%)project and program management professionals claim to have this access.Figure 26 shows three key platforms and associated tools that can help provide a seamless and augmented experience to software teams through all pha

263、ses of SDLC and across the organization:Developer workspace:Provides an interface to the generative AI assistance as part of existing or new workspace tools for IDEs,source code management,etc.Common foundations:Host foundational LLMs customized with proprietary data,APIs,and playbooks to accelerate

264、 generative AI deployment.DevSecOps platform:Helps software developers and operations teams perform security testing and evaluation at every while also interfacing with quality engineering and testing.62Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize

265、 the full potential of generative AI for software engineeringFor the software workforce to make optimal use of generative AI,they need these tools seamlessly integrated into their workflow.The Chief Business Digital Officer at a leading automaker said:“Its important that these tools integrate seamle

266、ssly because IT systems are already very complex.We look for tools that can integrate with our development environments.”Amazon Q Developer,AWSs generative AI assistant for software development,which was recently made“generally available,”offers inline code suggestions and chat in popular IDEs,inclu

267、ding JetBrains,IntelliJ IDEA,Visual Studio,and VS Code,and across 15 coding languages.42 This smoothens a developers workflow,making code generation and debugging faster.BT Group used Amazons generative AI coding companion,generating over 100,000 lines of code in the first four months.The solution p

268、rovided 1520 code suggestions per user per day,of which 37%were accepted by software engineers.Privately and safely contextualizing generative AI assistants with an organizations own content Large language models that form the foundations of todays generative AI assistants are generic and do not int

269、ernalize enough information about the specific context in which they are used.This creates code issues and inefficiencies leading to less relevant code and related content creation and,as a result,limits improvements in productivity and quality.This limitation is more evident in use cases requiring

270、access to enterprise data(enterprise content documentation repository,legal data/document repository,business documentation and design requirements,among others)to contextualize the model for specific desired outcomes.Archisman Munshi,co-founder of a senior citizen care company that is tackling fina

271、ncial fraud issues using AI elaborates,“Each company has its own way coding style,standards,patterns,and associated guardrails.Properly training the LLM with our data and processes ensure more robust code suggestions which are better aligned to an organizations conventions and guidelines,thus minimi

272、zing inefficiencies in the code.”Retrieval-augmented generation or fine-tuning are needed to fully adapt the generic models to the specific context or problem domain where they are being applied.Contextualizing generative AI output like this improves the quality of its output significantly by improv

273、ing the accuracy and relevance of responses.Consequently,users productivity is boosted as well since less time is spent looking for organizations internal specifics such as terminologies,conventions,processes etc.There are two major approaches for privately and safely customizing generative AI assis

274、tants for organizations specific context using proprietary or enterprise data and content:1.Retrieval-augmented generation(RAG):This approach combines a question/answer system(like a search engine)that fetches relevant document chunks from a large corpus(say internal knowledge base),and an LLM,which

275、 produces answers for the user using the information from those chunks.2.Fine-tuning:This is the process of taking a foundational LLM and further training it on a smaller,specific dataset to adapt it to a particular task/domain and improve its performance.63Capgemini Research Institute 2024Turbochar

276、ging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering Adopt a measurement protocol for generative AI impact monitoring and use case prioritization.What gets measured,gets managed might be an old saying but rings truer in new paradigms th

277、at generative AI is unfolding.Almost half of organizations in our survey(48%)have no standard metrics to gauge the success of generative AI use in software engineering(see Figure 27).Our survey also reveals that the commonly used metrics,while suitable for regular productivity measures,such as time

278、to deploy or to resolve issues,do not fully capture the benefits of generative AI,especially on non-conventional measures of productivity,such as employee satisfaction.These are better captured by metrics frameworks like DORA and SPACE.43 Although these metrics are yet to gain traction,as they are c

279、ostly and time-consuming to implement.This finding indicates that a set of metrics including KPIs for velocity,quality,security,and developer experience can prove useful.Measuring productivity within software engineering poses inherent complexities due to the multifaceted nature of the SDLC,the dyna

280、mic environment it operates in,and the subjective and intangible aspects of many Figure 27.Most organizations show improvement from use of generative AI when measured using less popular,but more holistic,productivity metrics frameworks,such as SPACE and DORAMetrics predominantly used vs.Metrics impa

281、cted through use of Generative AI for software engineering42%61%0%20%40%60%80%41%51%39%52%38%62%37%69%32%71%21%30%19%72%16%73%SPACE metricsCode churnNumber of user story points completedDORA(DevOps Research and Assessment)MetricsCode commit frequency Pull request resolution timeChange failure rate R

282、elease cycle/deployment time Sprint and release burndown/team velocityKey metrics used in your organization Metrics that have shown a positive impact due to Gen AI usage Source:Capgemini Research Institute,Generative AI in Software Engineering,Software Professionals Survey,April 2024,N=790 responden

283、ts overall.The question on Metrics that have shown an impact with Generative AI has been answered by the subset of respondents who chose the respective metric as a key metric used in their organization.64Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realiz

284、e the full potential of generative AI for software engineeringof its components.A comprehensive measurement approach must encompass both qualitative and quantitative factors,tailored to the specific context of the project.Similarly,assessing software quality presents challenges as it involves variou

285、s dimensions such as functionality,performance,reliability,usability,maintainability,security,and scalability,each requiring its own set of metrics and criteria.We recommend that organizations adopt a measurement protocol involving a measurement approach,metrics,team,and a well-defined process,to cr

286、eate actionable and reliable results:1.Measurement approach:It includes steps for measuring progress,such as:planning,setting baselines,and running an experimentation.Once the right metrics have been identified(more on this in the next point),tools like SonarQube,CAST,Jira,and developer surveys are

287、used for collecting and analyzing data.Team stability,duration,technology,legal considerations,and cybersecurity are key prerequisites and success factors for a consistent measurement approach.2.Metrics:Coding velocity serves as a pivotal metric to measure team productivity,focusing on coding and un

288、it testing activities,typically measured by implemented story points.Moreover,coding velocity per developer capability or seniority level discerns productivity variances,comparing time taken to complete a set number of story points with and without generative AI assistance.This approach offers insig

289、hts into how generative AI influences productivity across different developer skill levels and the intricacies of software development tasks.Test coverage,code efficiency,code security,code smells,code duplication among others,serve as essential KPIs for relevant use cases.3.Team:In the single team

290、measurement approach,one team sequentially executes a backlog of user stories of consistent size or complexity,comparing performance with and without generative AI assistance.On the other hand,a multiple teams approach involves parallel execution of the same backlog by at least two teams with differ

291、ent tool setups(e.g.,with and without generative AI tools),allowing for simultaneous assessment of generative AIs effects across different team dynamics.The seniority or capabilities of a team are important for normalization,thus its mandatory to know what kind of team mix is working on the defined

292、backlog.4.Process:Once all the ingredients have been defined,a process is needed to ensure high quality results and reduce side effects due to estimation inaccuracy and human factors:Define the team(s)organization and the experimentation scope and timeline.Define the measurement approach.Validate th

293、e prerequisites and success factors.Conduct a baseline for the metrics,without generative AI assistance.Execute the experimentation sprints with generative AI assistance.Collect and normalize the metrics and the feedback.Consolidate and report the measured results.65Capgemini Research Institute 2024

294、Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineering Put people at the heart of this transformation by creating a learning culture at your organization.Provide upskilling and cross-skilling opportunities.As we saw earlier,a siz

295、eable share of the workforce(41%)is rushing ahead to equip themselves with skills to use generative AI tools for software engineering.40%of these“independent learners”are doing so through paid courses highlighting the importance of the training for them.These are key indicators of how organizations

296、are lagging in providing upskilling opportunities to their workforce.Figure 28 shows a blueprint for a training program to build generative AI skills in the software workforce.Indeed,51%of senior executives believe that leveraging Figure 28.Blueprint of a training program to impart generative AI ski

297、lls to software teams.Introduction to generative AI for software engineeringLegal,security,ethical concernsIntroduction to various toolsPrompt engineeringConversational software engineeringUse cases across software engineering lifecycleCertificationsGenerative AI for software engineering tool-based

298、pathwaysIntroduction to generative AI LLMsAdvanced prompt engineeringLLM use cases across software engineering lifecycleCertificationsGenerative AI for software engineering LLM-based pathwaysBeginnerPractitionerMasterTraining levelsSource:Capgemini Research Institute analysis.66Capgemini Research In

299、stitute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringgenerative AI in software engineering will require significant investment in upskilling and cross-skilling the software workforce.Yet only 39%of organizations have

300、 a generative AI upskilling program for software engineering.Moreover,current upskilling programs are ill-equipped to provide training in areas of evaluation of generative AI code/output(56%),security(43%),and IP and legal issues(33%).Source:Capgemini Research Institute analysis.Figure 29.A mobiliza

301、tion plan to develop generative AI delivery capability.TrainingHands on complete developmentCommunityThought leadershipDevelop multi-level training plans Focus on creating mass delivery capabilityDevelop a predefined set of practice use casesEngage shadow teams for parallel developmentAll internal a

302、pps development leveraging generative AISetup communities to share experiences and learningLeverage internal knowledge management platforms,and newslettersConduct hackathonsEngage with partners for beta testing of new or upgraded toolsCreate proofs-of-value and proofs-of-conceptOur survey data revea

303、ls that workforces at organizations that provide training and access to generative AI are more optimistic about their productivity(84%vs.74%),job satisfaction(43%vs.32%),and future employability and pay(35%vs.29%)than those at organizations without such training and access.They also report being hap

304、pier(34%vs.23%)and less negative about the adoption of generative AI for software engineering(6%vs.22%).Having a training and upskilling program is just the first step.They can prove to be insufficient and rendered irrelevant soon if not backed by systems and processes that create a culture of learn

305、ing and thought leadership within the organization adopting generative AI for software.Figure 29 shows a blueprint of a mobilization plan to develop generative AI delivery capability.67Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potentia

306、l of generative AI for software engineering Address employees work displacement concerns 44%of senior executives cite risks with displacing and transitioning of workers as the biggest challenge for generative AI implementation.It is important to make the workforce understand the role of AI as a tool

307、 that augments and assists and does not replace.Archisman Munshi agrees,“It is imperative to educate people to view AI as an assistant that will make it easier and faster to complete their day-to-day tasks.Also,people wont get replaced,it is the tasks that will get replaced.Instead of spending time

308、on routine and mundane tasks like documentation,people will now get to spend more time on value-adding and challenging work.”“It is imperative to educate people to view AI as an assistant that will make it easier and faster to complete their day-to-day tasks.”Archisman Munshi,Co-founder of a senior

309、citizen care company that is tackling financial fraud issues using AI68Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringGenerative AI offers a new way to augment software engineering by boost

310、ing new software development,software quality,and software teams productivity.While it is still early days in the trial and actual use of this technology,benefits are already being seen.However,as of today,more software professionals are using generative AI tools using unauthorized ways than via aut

311、horized access.If left unchecked,this unauthorized usage can expose organizations Conclusionto various kinds of risks and damages.Leading organizations are able to manage this risk and yet derive the maximum potential generative AI has to offer today by transforming their software organization,deliv

312、ering technology prerequisites,measuring the realized value,and taking their people along on this transformative journey.69Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringResearch methodolog

313、yWe surveyed 1,098 senior executives(director and above)and 1,092 software professionals(architects,developers,testers,and project managers,among others).We also conducted 20 in-depth interviews with leaders from clients,partners,and startups,along with several professionals.Nearly all of the softwa

314、re professionals and senior executives represent a similar set of organization.Senior executives,by country of organizations headquartersSenior executives,by sectorAustralia5%Finland5%France12%Germany12%India9%Italy4%NL5%Norway5%Singapore5%Spain4%Sweden5%US17%UK12%Aerospace&Defence 14%Automotive13%B

315、anking14%High-Tech12%Consumer Products7%Retail8%Life sciences&Healthcare8%Insurance5%Energy transition&Utilities5%Telecommunications7%Public services7%Senior executives,by functionStrategy/general management8%R&D/product development/engineering18%Software product engineering9%Product and services de

316、velopment/portfolio strategy22%IT11%Technology21%Innovation11%70Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringSoftware professionals,by country of organizations headquartersAustralia5%Finl

317、and5%France13%Germany12%India8%Italy4%NL4%Norway5%Singapore4%Spain5%Sweden5%US18%UK12%Software professionals,by sectorAerospace&Defence 14%Automotive13%Banking14%High-Tech12%Consumer Products7%Retail7%Life sciences&Healthcare8%Insurance6%Energy transition&Utilities5%Telecommunications7%Public servic

318、es7%Software professionals,by software domainArchitecture-Softwawre architect,system design11%Development-frontend,backend,UI/UX design,integration,data management30%Testing29%Maintenance and operations16%Project/program management14%71Capgemini Research Institute 2024Turbocharging software with Gen

319、 AI:How organizations can realize the full potential of generative AI for software engineeringAppendixList of use cases analyzed in this research:Business DemandPhases of SDLCUse casesDesignCodingBusiness requirement/demand analysis and writingRFP preparation and proposal evaluationUser story genera

320、tionArchitecture writing assistance and modellingUX/UI design assistanceCoding assistance(including code generation,completion,etc.)Debugging and error predictionUnit test generationCode explanationLegacy code modernization(migration,conversion,etc.)Identification of software security vulnerabilitie

321、sApplication integrationTest case generation/testing code generationUser acceptance testingCode review and quality assurancePlatform provisioning and configuration(Infrastructure as code script writing support)Incident analysis and resolutionLog analysisProject management and collaboration(e.g.,back

322、log planning,effort estimations,team effectiveness analysis,communication)Agile ceremoniesDocumentationContinue to the next page.72Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringPhases of S

323、DLCUse casesBuildTestDeployMonitorApplication integrationTest case generation/testing code generationUser acceptance testingCode review and quality assurancePlatform provisioning and configuration(Infrastructure as code script writing support)Incident analysis and resolutionLog analysisProject manag

324、ement and collaboration(e.g.,backlog planning,effort estimations,team effectiveness analysis,communication)Agile ceremoniesDocumentation73Capgemini Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringAs a

325、 worldwide leader,generative AI for Software Engineering is strategic and core for our businessto deliver the bestvalueto our clientsCapgemini is a global leader with over 100K developers specializing in custom software engineering.We identify and evaluate the potential best generative AI use cases

326、for software engineering in order to create custom,innovative solutions.We rigorously assess and test through rigorous pilot programs with industrialized and standardized protocols to measure impact.Once proven,we seamlessly integrate them into our delivery processes to ensure we maximize value for

327、our clients.Evaluate Market Tools:We select the most suitable tools for each phase of the Software Development Life Cycle(SDLC).Legal and Compliance Framework:We develop frameworks to understand and meet clients specific legal and compliance requirements,ensuring secure and safe generative AI usage.

328、Impact Assessment:We analyse the impact on team organization,structure,skills,and workflows.Solution,Asset and Method Development:We build solutions,assets and methods to address software engineering needs,ways of working and enhance existing market tools throughout the SDLC.Measurement Protocols:We

329、 implement an auditable measurement protocol to evaluate the impact of generative AI on software engineering across the SDLC.Global Partnerships:We leverage our global partnerships with leading technology and tool vendors.Value Generation:We utilize generative AI in our projects to deliver enhanced

330、value to our clientsLeader in designing&managing complex IT transformations,engaging CIO/CxOsCapgeminis strategic partnerships on generative AIGlobal Leader in Software Engineering,with 100,000+expertsExpected outcomesProductivity GainQuality and SecurityTime to MarketDeveloper Experience74Capgemini

331、 Research Institute 2024Turbocharging software with Gen AI:How organizations can realize the full potential of generative AI for software engineeringMaximizing Software Engineering Transformation with generative AIIn addition to utilizing generative AI for software engineering in client projects,we

332、have a comprehensive strategy to guarantee our clients maximize the benefits of their generative AI driven software engineering transformation.Solution demonstrationsTransformation at scalePhase 2 Joint Software HousesMeet our expertsPractice on real casesDiscuss potential Define the prerequisites t

333、o deploy GenAI at scale,and manage HR implications Scale&deploy while measuring and materializing the expected outcomes in a business caseLong-term,continuous transformation partnershipSoftware Houses powered by GenAI and related transformationPhase 1 Assessment/Target/RoadmapAssess current SDLCUnderstand overall business and IT contextDefine the ambitionDetail the transformation journeyPilots/pro