《德勤：2024年第三季度企業生成式人工智能應用現狀報告（第3份）（英文版）（31頁）.pdf》由會員分享，可在線閱讀，更多相關《德勤：2024年第三季度企業生成式人工智能應用現狀報告（第3份）（英文版）（31頁）.pdf（31頁珍藏版）》請在三個皮匠報告上搜索。

1、Deloittes State of Generative AI in the Enterprise Quarter three reportAugust 2024Now decides next:Moving from potential to 2ForewordIntroductionNow:Key findings1 Building on initial success2 Striving to scale3 Modernizing data foundations4 Mitigating risks and preparing for regulation5 Maintaining

2、momentum by measuring valueNext:Looking aheadAuthorship&AcknowledgmentsAbout the Deloitte AI Institute About the Deloitte Center for Integrated Research About the Deloitte Center for Technology,Media&TelecommunicationsMethodologyTable of contentsIn the rapidly evolving landscape of artificial intell

3、igence(AI),the connection between technology and value has become increasingly apparent.What is known about major technology innovations in the past holds true with Generative AI(GenAI).Technology application on its own is not enough.Results and business outcomes matter.The real measure of success f

4、or GenAI will be how it enables enterprise strategies and drives tangible value.As organizations are scaling,and learning from,their GenAI pilots,I have heard the discourse around GenAI shift from unbridled excitement to a more nuanced and critical evaluation of its real impact on business outcomes.

5、I am also beginning to see organizations think more about tailored GenAI toolsevolving from large language models(LLMs)to small language models(SLMs)for more targeted needs.They are also exploring how the rise of AI agents can redefine interactions within their digital environments,offering new aven

6、ues for automation and personalization.Amid this maturation,regulatory considerations are coming to the fore.Our past survey results indicated a strong market appetite for smart GenAI regulation and oversight.Businesses and governments alike are navigating a dynamic landscape and are struggling to k

7、eep pace with the rate of technology innovation.The challenge is to unlock the benefits of GenAI while facing regulatory uncertainty,orchestrating governance and building trust.No small task.The complex discussions around creating value and managing risk makes it clear to me that we need to keep hum

8、ans at the center of all this decision-making.It is the human stakeholders who impact how applications are conceived and developed,how they are adopted and used,and how they are managed for trust and security.In this,employee upskilling and change management remain indispensable elements of value-dr

9、iving GenAI programs.With a focus on business outcomes and human-centered change,I feel the future with GenAI grows brighter by the day,even as the journey ahead will continue to surprise and challenge us.Learn more about the series and sign up for updates at http:/ Rowan,Applied AI SGO LeaderIntrod

10、uctionForeword3 3IntroductionMoving from potential to performanceThe clock is ticking for organizations to create significant and sustained value through their Generative AI initiatives.Promising pilots have led to more investments,escalating expectations and new challenges.During this pivotal phase

11、,C-suites and boards are beginning to look for returns on investment.There is a chance that their interest in Generative AI could wane if initiatives dont pay off as much,or as soon,as expected.Will organizations demonstrate the patience and perseverance needed to unlock the transformational potenti

12、al of Generative AI?To get there,value-led use cases with strong return on investment(ROI)and a clear path to scale will be essential.Theyll need to address challenges across the board:people,process,data and technology.Change management and organizational transformation will need to be given as muc

13、h consideration as technology.In this quarters survey,we focused on two critical areas to scalingdata and governance,and risk and complianceand how organizations are measuring and communicating value.Are data-related issues hindering efforts?How are organizations ensuring the right oversight of Gene

14、rative AI-powered applications?Is regulatory uncertainty holding them back?Are they developing a comprehensive set of financial and nonfinancial measures to form a complete picture of benefits achieved?These questions must be explored in-depth as organizations journey from Generative AI promise to p

15、erformance.4 4Building on initial success Improved efficiency and productivity and cost reduction are still the top benefits sought by organizations.Those are also cited by 42%of respondents as their most important benefits achieved to date.However,58%reported they realized a more diverse range of m

16、ost important benefits,such as increased innovation,improved products and services,or enhanced customer relationships.Respondents said that embedding Generative AI deeply into critical business functions and processes is the top way to drive the most value from their Generative AI initiatives.Strivi

17、ng to scale Two of three surveyed organizations said they are increasing their investments in Generative AI because they have seen strong early value to date.However,many are still challenged to successfully scale that valuenearly 70%of respondents said their organization has moved 30%or fewer of th

18、eir Generative AI experiments into production.IntroductionMoving from potential to performance(contd)All statistics noted in this report and its graphics are derived from Deloittes third quarterly survey,conducted May June 2024;The State of Generative AI in the Enterprise:Now decides next,a report s

19、eries.N(Total leader survey responses)=2,770.Percentages in this report and its charts may not add up to 100,due to rounding.Generative AI is an area of artificial intelligence and refers to AI that in response to a query can create text,images,video and other assets.Generative AI systems can intera

20、ct with humans and are often built using large language models(LLMs).Also referred to as“GenAI.”5Modernizing data foundations Three-quarters of respondents said their organizations have increased investment around data life cycle management to enable their Generative AI strategy.Top actions include

21、enhancing data security(54%)and improving data quality(48%).Data issues are limiting options55%of organizations reported avoiding certain Generative AI use cases because of data-related issues.Top data-related concerns include using sensitive data in models and managing data privacy and security.Mit

22、igating risks and preparing for regulation Organizations feel far less ready for the challenges Generative AI brings to risk management and governanceonly 23%rated their organization as highly prepared.In fact,three of the top four things holding organizations back from developing and deploying Gene

23、rative AI tools and applications are risk,regulation(such as the European Unions AI Act,in effect August 1),and governance issues.To deal with regulatory uncertainty,about half of organizations reported they are preparing regulatory forecasts or assessments.Maintaining momentum by measuring More tha

24、n 40%of respondents said their companies are struggling to define and measure the exact impacts of their Generative AI initiatives.Less than half said they are using specific KPIs to measure Generative AI performance,and many standard measures of success arent currently being applied.IntroductionMov

25、ing from potential to performance(contd)The wave three survey covered in this report was fielded to 2,770 director-to C-suite-level respondents across six industries and 14 countries between May and June 2024.Industries included:Consumer;Energy,Resources&Industrials;Financial Services;Life Sciences&

26、Health Care;Technology,Media&Telecom;and Government&Public Services.The survey data was augmented by additional insights from 25 interviews with C-suite executives and AI and data science leaders at large organizations across a range of industries.This quarterly report is part of an ongoing series b

27、y the Deloitte AI InstituteTM to help leaders in business,technology and the public sector track the rapid pace of Generative AI change and adoption.The series is based on Deloittes State of AI in the Enterprise reports,which have been released annually the past five years.Learn more at the State of

28、 Generative AI in the Enterprise:Wave three survey results6 6 Now:Key findings7 7Organizations say they are seeing value from their early Generative AI forays and those successes are driving more investment.Two-thirds of the organizations we surveyed(67%)said they are increasing investments in Gener

29、ative AI because they have seen strong value to date.A head of AI strategy and governance in the banking industry has seen this first-hand:“Before GenAI,most senior leaders only had a vague understanding of what AI was or what it can do.Now,they have AI at their fingertips,and it has opened their ey

30、es to the possibilities.We have applied for additional resources.”As in our prior quarterly surveys,improved efficiency and productivity and cost reduction continue to be the most common benefits sought from Generative AI initiatives.Those benefits were cited by 42%of wave three respondents as their

31、 single,most important benefit achieved to date(figure 1).However,for most wave three respondents(the other 58%),the top benefit achieved through the new technology is something other than efficiency,productivity or cost reduction.This includes increased innovation(12%),improved products and service

32、s(10%),and enhanced customer relationships(9%).The diversity of possible sources of value from Generative AI initiatives is exciting to many leaders and shows the potential and versatility of this new technology.This distribution could mean a couple of different things.Organizations may be seeking e

33、fficiency,productivity and cost reduction,but arent seeing it materialize yet;they may be getting unexpected value from less tangible areas;or they may be prioritizing these other types of value.There is no one-size-fits-all approach to employing Generative AI,and there is a wide range of benefits t

34、hat could be gained.It is important for organizations to be clear about what kind of value they are seeking before embarking on any Generative AI initiativesstart with value first.1Building on initial successNow:Key findingsTop benefit achieved through Generative AI initiativesFigure 1Q:What is the

35、most important benefit your organization has achieved to date through your Generative AI initiatives?(May/June 2024)N(Total)=2,7704%Better detection of fraud and risk management34%Improved efficiency and productivityEncouraged innovationIncreased revenueIncreased speed and/or ease of developing new

36、systems/softwareShifted workers from lower-to higher-value tasksDeveloped new products and servicesImproved existing products and servicesReduced costs9%Enhanced relationships with clients/customers7%6%4%6%12%10%9%67%of organizations we surveyed said they are increasing investments on Generative AI

37、given strong value seen to date.8 8Our executive interviews provided examples of Generative AI use cases that are already delivering real-world value across a wide range of industries.Although they are working toward things like automated decision-making,accelerated research and development,and grea

38、ter innovation and market differentiation,most projects further along in the scaling process are still focused on improving productivity(figure 2).Now:Key findingsFigure 2Generative AI use cases delivering real-world value by industryA customer service tool that handles messages,using both chat and

39、voice,and provides cross-sell opportunities based on the interaction.A system to provide customer support and handle simple support tickets.The system can automatically pull data for human agents to use for more complex tasks.Support tools deployed for retail and technical field staff,and systems fo

40、r troubleshooting and preventive maintenance,all to reduce costs.An internal medical claims appeal review tool that provides increased response quality and a decreased time to respond.Customer segmentation tools leveraged to create more precise and customized segments across geographies.Continuous i

41、mprovement processes enhanced by directly leveraging customer feedback to inform product development road maps.BankingTransportationTelecomInsuranceConsumerTechnologyFinancePharmaceuticalsProject management tools that quickly create summary materials for key stakeholders.Internal tool that provides

42、instant enterprise information(such as standard operating procedures)for thousands of staff.9What do organizations think will most help drive greater value for their Generative AI initiatives?While many different factors contribute to Generative AI value creation,the action cited most often by the l

43、eaders we surveyed is embedding the technology deeply into business functions and processes(figure 3).“An LLM is like an engine,”said a VP at a banks AI center of excellence.“No one just wants the engine of a car or a plane;they want a car or a plane.So,there are all these things you need to do to m

44、ake it part of business processes,so the business can use it.”The value from any Generative AI initiative wont be fully realized if it sits apart.As with other technologies,it will only reach its potential when it is embedded in everyday tasks.Many organizations are already employing enterprise tool

45、s enhanced with this emerging technology resource to try and make this happen.Although many have seen promising results from early projects and are increasing investment in Generative AI,it is important that organizations show sustained and significant value as quickly as possible.“CEOs and executiv

46、e leadership teams are getting much more excited and interested in whats possible and are looking for use cases to demonstrate the value and benefit,”said the global head of AI,machine learning,analytics and data at a pharmaceutical company.“There is a lot of willingness to test,experiment and scale

47、.However,the potential danger is that people might get disappointed and lose attention if its not paying off fast enough.”C-suite and board members are still intrigued,but there are some potential signs of enthusiasm beginning to wane as the“new technology shine”wears off.Survey respondents said tha

48、t interest in Generative AI remains“high”or“very high”among most senior executives(63%)and boards(53%);however,those numbers have declined since the Q1 2024 survey,dropping 11 percentage points and 8 percentage points respectively.Time is of the essence as organizations look to scale their early ach

49、ievements.Now:Key findingsFigure 3Q:Which behavior/action do you think will drive the most value for the Generative AI initiatives in your organization?(May/June 2024)N(Total)=2,770Behaviors driving the most value for Generative AI initiatives22%Deeply embedding GenAI into functions/processes13%Effe

50、ctively managing risks11%Deploying the latest technology10%Developing creative and differentiated applications10%Tailoring/customizing models with proprietary data9%Hiring the best talent8%Providing enough budget8%Completely measuring performance7%Providing access to as much of the workforce as poss

51、ible10Striving to scaleNow:Key findingsSelecting and quickly scaling the Generative AI projects with the most potential to create value is the goal.However,many Generative AI efforts are still at the pilot or proof-of-concept stage,with a large majority of respondents(68%)saying their organization h

52、as moved 30%or fewer of their Generative AI experiments fully into production(figure 4).This isnt necessarily surprisingdespite rapid and impressive advances in Generative AIs capabilities,its applications are still relatively new and organizations are figuring out what it can(and cant)do well.Many

53、organizations are learning through experience that large-scale Generative AI deployment can be a difficult and multifaceted challenge.As with a lot of digital transformation efforts,projects can fail or struggle for a variety of reasons.“Most of our applications are still in the minimum-viable-produ

54、ct or proof-of-concept phase,”said a senior specialist for AI compliance in the automotive industry.2Figure 4Q:In your estimation,what percentage of your Generative AI experiments have been deployed to date into your organization(moved into production)?(May/June 2024)N(Total)=2,770A large majority o

55、f organizations have deployed less than a third of their GenAI experiments into production“Scaling across an organization where you have thousands of employees has several basic requirements,and theyre quite challenging.”-Senior specialist for AI compliance in the automotive industry10%20%30%40%50%6

56、0%70%80%26%24%14%7%4%3%1%19%0%1%Successfully scaling may mean different things to different organizationsbased on their goals,what approach they are taking with Generative AI,and to what extent scaling is actually necessary.They could be expanding from one market to multiple markets,from a small gro

57、up within a function to the entire function,or from a portion of a process to multiple,integrated processes.It also depends on what Generative AI-powered tools and applications are being used:scaling a code generator across an IT department is going to be different than scaling a customized LLM for

58、the finance function,or a new enterprise customer relationship management application with Generative AI features.Organizations GenAI experiments moved into production11Effective model management and operationsProvisioning the right AI infrastructureModern data foundationModular architecture and com

59、mon platformsData&technologyDespite these differences,some fundamentals are consistent.“Foremost,you need a strategy,”the senior specialist for AI compliance continued.“Strategy means you cant start by purchasing separate solutions.if you really want to scale,first you need to base your strategy on

60、platforms.”This platform-centric approach could include establishing centers of excellence,technology platforms to enable multiple use cases,and centralized teams of experts.In our Q2 report we advocated for centralized resources that can accelerate deployment of similar use cases and enable organiz

61、ations to make the most of scarce Generative AI expertise.More broadly,organizations should invest in the foundations of Generative AI and concurrently assess and advance their strategy,processes,people,data and technology(figure 5).Many of the fundamentals may look similar to prior digital transfor

62、mation efforts,but due to the unique nature of Generative AI,things like robust governance,transparency for building trust,transforming talent,and mature data life cycle management take on increased importance.Now:Key findingsFigure 5Essential elements for scaling Generative AI initiatives from pilo

63、t to productionAcquiring (external)and developing(internal)talentTransparency to build trust in secure AITalentTransformed roles,activities and cultureIntegrated risk managementAgile operating model and delivery methodsStrategyProcessRobust governanceStrong ecosystem collaborationAmbitious strategy&

64、value management focusClear,high-impact use case portfolio12Now:Key findingsHow do organizations feel like they are doing across these areasare they prepared to scale?We asked how highly prepared respondents thought their organizations were across some of the essential scaling elements(figure 6).Tec

65、hnology infrastructure(45%)and data management(41%)fared the best,followed by strategy(37%),risk and governance(23%),and talent(20%).This indicates that there are still some fundamental challenges holding organizations back from successfully scaling their Generative AI initiatives.A senior director

66、and head of a Generative AI accelerator in the pharmaceutical industry identified a number of pressing issues:“The heritage of our processes and approaches,that is whats really holding us back right now.Number two is that the performance of the LLMs still needs to be improved Data readiness;data is

67、going to be problem forever.Deep Generative AI understanding as well.Theres not enough people who understand and can drive transformation.”To help start a conversation on how to overcome some of these barriers,in this quarters survey we focused on two areas critical to scalingexploring how organizat

68、ions are approaching data and governance,and risk and compliance.With respect to data,more organizations leaders reported they are initially prepared.For risk and governance,they know they are not.Both need attention.Q:For each area,rate your organizations level of preparedness with respect to broad

69、ly adopting generative AI tools/applications?(May/June 2024)N(Total)=2,770Figure 6Do organizations think they are ready?Percentage of organizations that are highly prepared for GenAI across the following areas41%Data managementTalent20%Risk&governance23%Strategy37%45%Technology infrastructure13Moder

70、nizing data foundations Now:Key findings3Compared with the other aspects of Generative AI readiness,survey respondents judged that their organizations are fairly mature with respect to data life cycle management(as a reminder,survey respondents are from more AI-savvy organizations).This could be bec

71、ause they had a good foundation to start with or that,according to our survey,75%of organizations have increased their technology investments around data life cycle management due to Generative AI.This increased focus was evident in our executive interviews.“Theres a whole series of questions GenAI

72、is triggering about data strategy,that in the past were far less important,”said the chief technology officer at a manufacturing company.“I think were probably spending as much time on data strategy and management as on pure GenAI questions,because data is the foundation for GenAI work.”However,even

73、 those executives who consider themselves highly prepared will likely need to do more as they progress in their journeys.Some we interviewed said that as they moved from proof of concept to scale,unforeseen data issues were exposedhighlighting a need to be agile.These issues could be because of the

74、Generative AI-specific demands to data architecture and management.More robust governancequality,privacy,security,transparencyis needed overall,especially around using data that doesnt already exist inside the organization(e.g.,public domain,synthetic and licensed third-party data).Documenting data

75、sources and labeling has an increased importance.With more people potentially leveraging data,data access frameworks and literacy require more attention.It may change approaches toward cloud or on-premises data services.For more advanced LLM users,working with synthetic data may eventually come into

76、 play.75%of organizations have increased their technology investments around data life cycle management due to Generative AI.1414One of these challenges was highlighted by a former vice president of data and intelligence for a media and entertainment company:“The biggest scaling challenge was really

77、 the amount of data that we had access to and the lack of proper data management maturity.There was no formal data catalog.There was no formal metadata and labeling of data points across the enterprise.We could go only as fast as we could label the data.”Data-related issues could be hindering organi

78、zations in their quests for getting the levels of value that they are seeking.Data-related issues have caused 55%of the organizations we surveyed to avoid certain Generative AI use cases.That could be because of data-quality issues,intellectual property concerns,not having the right data,or worries

79、about using certain kinds of data(e.g.,public domain,synthetic or licensed third-party data).The concerns that organizations were worried about the most in our survey included using sensitive data in models(58%had at least a high level of concern),data privacy issues(58%),and data security issues(57

80、%)(figure 7).Organizations were much more worried about using sensitive data(e.g.,customer or client data)than they were using their own proprietary data(e.g.,sales,operational,financial).Now:Key findingsQ:For the following,how much concern does your organization have with respect to its data manage

81、ment for Generative AI implementations?(May/June 2024)N(Total)=2,770Figure 7Levels of concern around data management(high+very high)Using sensitive data in modelsManaging data privacy-related issuesManaging data security-related issuesComplying with data-related regulationsUsing our own proprietary

82、data in models57%49%38%58%58%15Consistent with those concerns,the top actions organizations are taking to improve their data-related capabilities are enhancing data security(54%),improving data quality practices(48%),and updating data governance frameworks and/or developing new data policies(45%)(fi

83、gure 8).The value from Generative AI initiatives will increasingly come from organizations leveraging their differentiated data in new ways(whether for fine-tuning LLMs,building an LLM from scratch or utilizing enterprise solutions).1 For Generative AI to deliver the kind of impact executives expect

84、,companies will likely need to increase their comfort with using their proprietary data,which may be subject to existing and emerging regulations.Q:What specific actions has your organization taken to improve its data-related capabilities to support its Generative AI initiatives?(May/June 2024)N(Tot

85、al)=2,770Figure 8Improving data-related capabilitiesNow:Key findings“Data quality is key.Understanding what data is good data.Where is that data held?How is it secured?How is it permissable?All those things are key to making Generative AI scalable.”-Chief operations officer&chief of strategy for a f

86、inancial services firm54%Enhanced data securityImproved data quality practicesUpdated governance frameworks/Developed new data policiesCollaborated with cloud service provider or IT integrator to improve capabilitiesUpgraded IT infrastructure48%45%43%37%Integrated data silosHired new talent to fill

87、data-related skill gaps34%27%Moved to a more flexible,open data architecture24%16According to our survey respondents,three of the top four barriers to successful development and deployment of Generative AI tools and applications are:Currently,these are considered even more significant than other cri

88、tical barriers such as implementation challenges,a lack of an adoption strategy,and difficulty identifying use cases.Likely driving these concerns are new and emerging risks specific to the new tools and capabilitieslike model bias,hallucinations,novel privacy concerns,trust and protecting new attac

89、k surfaces.This environment may be why organizations feel far less ready for the challenges Generative AI brings to risk management and governancesince only 23%rated their organization as highly prepared.These issues will be increasingly important as activities shift from small-scale pilots to large

90、-scale deployments and Generative AI becomes more deeply embedded into the fabric of organizations.Highlighting the importance,respondents selected effectively managing risks as the second-most reported way to drive the most value for Generative AI initiatives.Mitigating risks and preparing for regu

91、lationNow:Key findings436%worries about regulatory compliance30%difficulty managing risks29%lack of a governance model17Actions to manage riskQ:What is your organization currently doing to actively manage the risks around your Generative AI implementations?(May/June 2024)N(Total)=2,770Figure 9Establ

92、ishing a governance framework for the use of GenAI tools/applicationsMonitoring regulatory requirements and ensuring complianceUsing a formal group or board to advise on GenAI-related risksKeeping a formal inventory of all GenAI implementationsUsing outside vendors to conduct independent audits and

93、testingConducting internal audits and testing of GenAI tools/applicationsTraining practitioners how to recognize and mitigate potential risksEnsuring a human validates all GenAI-created contentSingle executive responsible for managing GenAI-related risks51%49%43%37%35%33%30%23%19%Now:Key findingsThe

94、 chief operations officer and chief of strategy in a financial services company summed up the challenge:“How do you democratize Generative AI across your business while having all of the right controls in place?We have an AI board,we have an ethics framework,we have an accountability model.We want t

95、o know whos using it for what,and that its being used in the right way.”To help build trust and ensure the responsible use of Generative AI-powered tools and applications,organizations are generally working to establish new guardrails,educate their workforces,conduct assessments,and build oversight

96、capabilities.Specific actions surveyed organizations are currently taking include establishing a governance framework for using Generative AI tools and applications(51%),monitoring regulatory requirements and ensuring compliance(49%),and conducting internal audits/testing on Generative AI tools and

97、applications(43%)(figure 9).Despite their importance for effective scaling,each of these actions is only being taken by less than roughly half of the organizations we surveyed.18Implementing new processes and controls is rarely easy and will likely require active change management to build support w

98、ithin the organization.“Before launching anything,we have strict AI governance,”said the chief analytics officer at a professional services firm.“In the past we had a bit of a siloed approach,but today,at a minimum,everything has to go through privacy and compliance because we have a methodical way

99、of managing risk.This is new and challenging to some.”On top of risk and governance issues,Q3 surveyed organizations were exceedingly uncertain about the regulatory environment that may exist in the future(depending on the countries they operate in).In our first quarterly report,78%of leaders agreed

100、 that more governmental regulation of AI was needed.However,there is a difference between theory and practice.Organizations are struggling with regulatory uncertainty,and worries about interpretation and enforcement may be preventing them from pursuing certain use cases in specific geographies.The u

101、ncertainty around AI regulation may make it feel like there could be many varied outcomes,but our research suggests most countries are following a similar path concerning AI policies.2 Governments are working to balance protection,innovation and economic benefit,so future actions will likely be in l

102、ine with the regulatory traditions of each country and region.Now:Key findings78%of leaders surveyed in Q1 agreed that more governmental regulation of AI was needed.1919Some organizations reported taking action to prepare for potential regulatory changes.Top areas include preparing regulatory foreca

103、sts or assessments(50%),monitoring by the general counsel(48%),and working with external partners(46%)(figure 10).However,some organizations arent doing anything to prepare;14%said they arent making any specific plans.Now:Key findingsInsights from our executive interviewsHow some real-world organiza

104、tions are dealing with compliance,risk management and governance issuesAn increasing number of organizations are making risk a central factor when selecting Generative AI use cases and investments.However,many are walking a tightropetrying to minimize risk without being too risk averse,which could l

105、ead to missed opportunities and open the door to competitors.Here are some risk-related actions revealed through our in-depth executive interviews:How organizations are preparing for regulatory changesQ:How is your organization preparing for potential regulatory changes with respect to Generative AI

106、?(May/June 2024)N(Total)=2,770Figure 10Avoiding specific tools and use casesLimiting data exposureBuilding frameworksManaging regulatory uncertaintyAvoid use cases that could require additional regulatory scrutinyShut off access to specific Generative AI tools for staffFor organizations that rely he

107、avily on owned intellectual property,be extremely cautious when exposing data to Generative AI modelsPut in place guidelines to prevent staff from entering organizational data into public LLMsInvest in custom solutions throughout the firms technology stack to enable more controlBuild walled gardens

108、in private clouds with safeguards to prevent data leakage into the public cloudDevelop strong frameworks with compliance,risk and privacy teams to actively manage risksWork with partners to develop ecosystem solutions for regulatory complianceDetermine how to concurrently address multiple regulation

109、sEmploying technologyCorporate strategy prepares regulatory forecasts and assessmentsFormal monitoring from general counsel teamConsultation with external partnersBusiness lines prepare regulatory forecasts and assessmentsNot making any specific plans at this time50%48%46%40%14%2020Maintaining momen

110、tum by measuring valueNow:Key findingsMany organizations are still figuring out how to best measure and communicate the value of their Generative AI initiatives.According to this quarters survey results,41%of organizations have struggled to define and measure the exact impacts of their Generative AI

111、 efforts.With the high level of experimentation and everything moving so quickly,it may be hard to pause and assess progress in a comprehensive way.(And,we note,the bar on measuring and communicating Generative AIs impact could soon rise.)“When it comes to Generative AI,for now,we are doing qualitat

112、ive assessments,”said the director of AI business development and strategy at a technology company.“However,once you scale past proof of concept,you must have a more quantitative assessment that becomes part of a whole portfolio prioritization process within the organization.”A more subjective appro

113、ach to early Generative AI investment is reasonable when there are few fully scaled examples to rely on as proof points.However,there is growing need to demonstrate value to ensure continued support and funding.Already some enterprises reported they are employing formal approaches to measure and com

114、municate Generative AI value creation,including using specific KPIs for evaluating Generative AI performance(48%)and building a framework for evaluating Generative AI investments(38%)(figure 11).It is worth noting that although a majority(54%)of organizations are seeking efficiency and productivity

115、improvements,only 38%reported they are tracking changes in employee productivity.5Q:What actions has your organization taken to measure and communicate value creation from your Generative AI initiatives?(May/June 2024)N(Total)=2,770Figure 11Used specific KPIs for evaluating GenAI performance48%Built

116、 a framework for evaluating GenAi investmentsTracked changes in employee productivityTracked return on investmentTracked nonfinancial benefitsAppointed someone to track value creation from GenAI initiativesProduced regular reports for the CFOActions taken to measure&communicate valueNone of these38%

117、38%35%34%29%16%6%41%of organizations have struggled to define and measure the exact impacts of their GenAI efforts.21Additionally,only 16%of organizations reported they produce regular reports for the CFO about the value being created with Generative AI.As Generative AI becomes an integral part of h

118、ow business gets done,we expect increasing focus on traditional financial metrics as organizations start to demand more tangible and measurable results from their Generative AI investments.In our executive interviews we heard that cost will increasingly become a key factor in decision-making about G

119、enerative AI.Looking ahead,a comprehensive set of financial and nonfinancial measures will be needed to present a complete picture of the value created from investments in Generative AI initiatives.3 In the future,we may see new metrics emerge that reflect its unique characteristics and capabilities

120、.For example,there could be a metric that quantifies the performance of human workers and Generative AI systems(together vs.separately)on creative and innovation-related tasks.Figuring out how to effectively measure and communicate the technologys value will be critical for setting expectations and

121、maintaining interest,support and investment from the C-suite and boardroom.Now:Key findings“Just like in any ideation phase,theres lots of excitement.But streamlining will come when cost becomes a constraint.Then teams that have stronger use cases will be able to spend and create the ROI.”-Director

122、of data science and AI in the tech industry 2222How are different industries approaching Generative AI?Weighted global averageConsumerEnergy,Resources&IndustrialsFinancial ServicesLife Sciences&Health CareExpertise:Percentage that assess their organizations GenAI expertise as“high”or“very high”Now:K

123、ey findingsFigure 12(May/June 2024)N(Total)=2,770Technology,Media&TelecomGovernment&Public ServicesIncreasing investment:Percentage that agree that their organization is increasing investment in GenAI initiatives because they have seen strong value to dateScaling progress:Percentage that estimate on

124、ly 30%or fewer of their GenAI experiments have been moved to productionRisk and governance preparedness:Percentage that say their organization is“highly”or“very highly”prepared with respect to risk&governanceData challenges:Percentage that say data-related challenges are slowing GenAI effortsActions

125、 to measure value:Percentage tracking ROIBetter than averageAverageWorse than average36%40%39%39%33%17%56%63%70%67%60%62%89%72%60%68%56%71%79%72%68%17%25%23%36%25%26%43%40%44%40%38%51%34%34%42%35%35%28%35%37%16%23 Next:Looking ahead2424Use Generative AI where appropriate to drive efficiency,producti

126、vity and cost reduction through large-scale deploymentbut dont stop there.Consider actively reinvesting the resulting cost savings(and freed-up capacity)to pursue Generative AIs many other potential benefitsincluding increased innovation,improved products and services,enhanced customer relationships

127、,and revenue growth.Many organizations are already seeing tangible value from Generative AI in these other areas,and such benefits will only become more important in the future.Imagine how Generative AI could combine with your organizations other technologies and strategic initiatives to transform e

128、very aspect of your business,not just for improving productivity(doing the same things better),but for innovation(doing new things).Ultimately,the biggest value could likely come from using Generative AI to fundamentally reinvent your business processes.Build toward transformation with enduring valu

129、e.Next:Looking ahead25Many organizations are learning that they cant even get started with Generative AI until they address their data deficiencies.Activities such as LLM tuning and training require high-quality data that is free of issues related to privacy,confidentiality and intellectual property

130、.In addition,many organizations likely havent paid as much attention to external data as to existing internal data.As such,data life cycle management should be at the top of every organizations Generative AI priority list.Focus on improving your data foundations(e.g.,quality,security,privacy,extract

131、ion,labeling).Also,bolster strategic relationships with members of your data ecosystem(e.g.,B2B partners,data end users,third-party data providers),just like you have with your key technology vendors.Make data an accelerator,not a barrier.Using publicly available large language models(LLMs)and nonco

132、nfidential data for efficiency and productivity improvements are likely to become less differentiating over time.Value will increasingly be driven by more innovative applications of Generative AI and strong enabling processeslike technology governance,data life cycle management,workforce development

133、,and process integration expertise.Additionally,improved organizational flexibility and stronger change management capabilities could also accelerate scaling and drive value.Those capabilities will aid in the quick integration of new models for new uses cases as industries move beyond LLMs to custom

134、 domain and industry-specific models and small language models(SLMs).Focus on fundamentals and adaptability.Next:Looking ahead26As Generative AI technologies and use cases mature,organizations will be less inclined to invest based solely on lofty visions,big promises and/or wishful thinking(or fear

135、of missing out).Establishing more rigorous mechanisms for measuring and communicating the value from Generative AI initiatives can help organizations secure and maintain the funding required for effective large-scale deployment.In the proof-of-concept stage,organizations can often get by with qualit

136、ative metrics;and thus far,Generative AIs results and performance against those metrics have been promising enough to invest more.However,once you get past the initial stage and try to scale,you also need quantitative metrics to measure and communicate value in a more tangible way.And,you need to pr

137、epare for oversight and cost pressures to increase over time.Measure performance more rigorously.Leaders grasp how essential governance,risk and compliance are for responsible Generative AI adoption.However,there still seems to be a“knowing”versus“doing”gap for most organizations.To help ensure your

138、 organization isnt held back by these issues,its critical to do three key things.First,boards and C-suites should stay regularly engaged in comprehensive conversations about Generative AI.Second,cross-functional teams should lead the identification and mitigation of risks.Finally,a single executive

139、should be charged with and responsible for managing Generative AI-related risks.This third piece is something very few organizations currently have.This executive should be prepared to manage the unforeseen risks that emerge as experiments scale.This executive should also carefully consider pursuing

140、 Generative AI applications using more sensitive dataand not altogether avoiding those use cases.Finally,with regulatory development still in early stages,this executive should ensure that regulatory monitoring and assessments are completed frequently.Democratize responsibly and with accountability.

141、Next:Looking ahead27Brenna Sniderman Executive Director Deloitte Center for Integrated Research Deloitte Services LP Authorship and AcknowledgmentsJim Rowan Applied AI SGO Leader Deloitte Consulting LLP David Jarvis Senior Research Leader Deloitte Center for Technology,Media&Telecommunications Deloi

142、tte Services LP Acknowledgments The authors would like to thank our project sponsors and the many talented professionals who brought this research to life:Nitin Mittal,Kevin Westcott,Lynne Sterrett and Jeff Loucks for their leadership,as well as the additional Deloitte subject matter specialists who

143、 contributed to the development of the survey and report:Ed Bowen,Bjoern Bringmann,Lou DiLorenzo,Rohan Gupta,Kellie Nuttal,Baris Sarer,Laura Shact,Ed Van Buren,Ajay Tripathi and Ashish Verma.We would also like to thank the many hands and feet that brought this report and campaign to life,including:A

144、hmed Alibage,Siri Anderson,Hali Austin,Saurabh Bansode,Natasha Buckley,Vanessa Carney,Dystnct Media,Tracy Fulham,Jordan Garrick,Gerson Lehrman Group(GLG),Lou Ghaddar,Jeanie Havens,Divvya Hocchalter,Karen Hogger,Susie Husted,Wendy Jenkins,Lisa Iliff,David Jarvis,Justin Joyner,Diana Kearns-Manolatos,L

145、ena La,David Levin,Michael Lim,Joe Mariani,Cullen Marriott,Rajesh Medisetti,Judy Freeman Mills,Melissa Neumann,Inal Olmez,Jamie Palmeroni,Jonathan Pryce,Emily Rosenberg,Negina Rood,Meredith Schoen,Kelcey Strong,Kate Schmidt,10 EQS,Sandeep Vellanki,Ivana Vucenovic,Talia Wertico,Marianne Wilkinson and

146、 Sourabh Yaduvanshi.Costi Perricos Global Office of Generative AI Leader Deloitte UK cperricosdeloitte.co.ukBeena Ammanath Executive Director Global Deloitte AI Institute Deloitte LLP Business leadershipResearch leadership28About the Deloitte AI Institute The Deloitte AI Institute helps organization

147、s connect all the different dimensions of the robust,highly dynamic and rapidly evolving AI ecosystem.The AI Institute leads conversations on applied AI innovation across industries,using cutting-edge insights to promote human-machine collaboration in the Age of With.The Deloitte AI Institute aims t

148、o promote dialogue about and development of artificial intelligence,stimulate innovation,and examine challenges to AI implementation and ways to address them.The AI Institute collaborates with an ecosystem composed of academic research groups,startups,entrepreneurs,innovators,mature AI product leade

149、rs and AI visionaries to explore key areas of artificial intelligence including risks,policies,ethics,future of work and talent,and applied AI use cases.Combined with Deloittes deep knowledge and experience in artificial intelligence applications,the institute helps make sense of this complex ecosys

150、tem and,as a result,delivers impactful perspectives to help organizations succeed by making informed AI decisions.About the Deloitte Center for Integrated ResearchThe Deloitte Center for Integrated Research(CIR)offers rigorously researched and data-driven perspectives on critical issues affecting bu

151、sinesses today.We sit at the center of Deloittes industry and functional expertise,combining the leading insights from across our firm to help leaders confidently compete in todays ever-changing marketplace.About the Deloitte Center for Technology,Media&TelecommunicationsThe Deloitte Center for Tech

152、nology,Media&Telecommunications(TMT Center)is a world-class research organization that serves Deloittes TMT practice and our clients.Our team of professional researchers produce practical foresight,fresh insights,and trustworthy data to help clients see clearly,act decisively and compete with confid

153、ence.We create original research using a combination of rigorous methodologies and deep TMT industry knowledge.Learn moreLearn moreLearn more29To obtain a global view of how Generative AI is being adopted by organizations on the leading edge of AI,Deloitte surveyed 2,770 leaders between May and June

154、 2024.Respondents were senior leaders in their organization and included board and C-suite members,and those at the president,vice president and director levels.The survey sample was split equally between IT and line of business leaders.Fourteen countries were represented:Australia(100 respondents),

155、Brazil(115 respondents),Canada(175 respondents),France(130 respondents),Germany(150 respondents),India(200 respondents),Italy(75 respondents),Japan(100 respondents),Mexico(100 respondents),the Netherlands(50 respondents),Singapore(75 respondents),Spain(100 respondents),the United Kingdom(200 respond

156、ents),and the United States(1,200 respondents).All participating organizations have one or more working implementations of AI being used daily.Plus,they have pilots in place to explore Generative AI or have one or more working implementations of Generative AI being used daily.Respondents were requir

157、ed to meet one of the following criteria with respect to their organizations AI and data science strategy,investments,implementation approach and value measurement:influence decision-making,are part of a team that makes decisions,are the final decision-maker,or manage or oversee AI technology implem

158、entations.All statistics noted in this report and its graphics are derived from Deloittes third quarterly survey,conducted May June 2024;The State of Generative AI in the Enterprise:Now decides next,a report series.N(Total leader survey responses)=2,770Methodology1.Chris Arkenberg,Baris Sarer,Gillia

159、n Crossan and Rohn Gupta,“Taking control:Generative AI trains on private,enterprise data,”Deloitte Insights,November 29,2023,https:/ August 1,2024.2.Joe Mariani,William D.Eggers and Pankaj Kishnani,“The AI regulations that arent being talked about,”Deloitte Insights,November 10,2023,https:/ August 1

160、,2024.3.Andy Bayiates,“Safeguard tech budgets by showing the full picture of techs value,”Deloitte Insights,March 11,2024,https:/ August 1,2024.Endnotes30About DeloitteDeloitte refers to one or more of Deloitte Touche Tohmatsu Limited(DTTL),its global network of member firms,and their related entiti

161、es(collectively,the“Deloitte organization”).DTTL(also referred to as“Deloitte Global”)and each of its member firms and related entities are legally separate and independent entities,which cannot obligate or bind each other in respect of third parties.DTTL and each DTTL member firm and related entity

162、 is liable only for its own acts and omissions,and not those of each other.DTTL does not provide services to clients.Please see to learn more.Deloitte provides industry-leading audit and assurance,tax and related services,consulting,financial advisory,and risk advisory services to nearly 90%of the F

163、ortune Global 500 and thousands of private companies.Our people deliver measurable and lasting results that help reinforce public trust in capital markets,enable clients to transform and thrive,and lead the way toward a stronger economy,a more equitable society,and a sustainable world.Building on it

164、s 175-plus year history,Deloitte spans more than 150 countries and territories.Learn how Deloittes approximately 457,000 people worldwide make an impact that matters at .This publication contains general information only and Deloitte is not,by means of this publication,rendering accounting,business,

165、financial,investment,legal,tax,or other professional advice or services.This publication is not a substitute for such professional advice or services,nor should it be used as a basis for any decision or action that may affect your business.Before making any decision or taking any action that may affect your business,you should consult a qualified professional advisor.Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.Copyright 2024 Deloitte Development LLC.All rights