《UiPath：2025年AI和自動化趨勢報告（英文版）（13頁）.pdf》由會員分享，可在線閱讀，更多相關《UiPath：2025年AI和自動化趨勢報告（英文版）（13頁）.pdf（13頁珍藏版）》請在三個皮匠報告上搜索。

1、UIPATH.COM|AI AND AUTOMATION TRENDS 20252 2Welcome to UiPath AI and Automation Trends 2025INTRODUCTIONWelcome to our report on the forces shaping AI and automation through 2025 and beyond.Our goal is to provide you with the information and“future-sight”you need to understand the environment and take

2、 advantage of opportunities arising in it.These trends reflect our analysis of third-party research,melded with unique insights drawn from our relationships with our 10,000+UiPath customers and millions of members of the global UiPath Community,along with thousands of technology and go-to-market par

3、tners.Weve also incorporated perspectives from UiPath AI scientists,product developers,software engineers,and automation strategists,as well as those of our sales,marketing,and customer support teams.Its always a challenge to distill whats happening out there into a short list of trends.But this yea

4、r has proved even harder than usual.Thats because there are so many profound forces at work.Chief among them:the rapid emergence of agentic AI,which equips software agents with the intelligence to not only understand what actions need to be taken to reach a goal,but also to take those actions themse

5、lves.A great leap forward in AI science,agentic AI has significant implications for enterprise automation and is interwoven into many of our trends.Thats not to say that generative AI has fallen by the wayside.Its true that many organizations are still struggling to define and scale their GenAI init

6、iatives.But theyre still capturing gains,both from the“outside-in”AI built into enterprise software and new LLM-powered approaches to taming and gaining advantage from their internal data.Finally,as AIs impact on our public and private lives expands and intensifies,enterprises and technology provide

7、rs will need to stay on top of a slew of new legislation regulating everything from managing a virtual workforce to copyright ownership to the use of data in model training and beyond.Clearly,theres a lot to cover in this years Trends report.So lets dive in.UIPATH.COM|AI AND AUTOMATION TRENDS 20253R

8、egulation escalation:the world acts to rein in AIs power.Lawmakers and the courts press on in their quest to regulate the virtual world.7From RAGs to riches:new tools tame the data deluge.This year,context is king;knowledge(graphs)are power;and LLMs are living larger.6Built-in AI lifts enterprises f

9、rom the trough of disillusionment.Companies still struggle to capture AI value by themselvesbut tech companies make up the gap.5Job sharing with the machine:the great work reallocation begins.Who can best do the taskpeople or machines?Enterprises rethink jobs from the ground up.4Agents get to work o

10、n long-tail automation opportunities.Going live in 25:valuable use cases across the enterprise that require agents AND robots.3Strike up the orchestration:the agentic ecosystem takes shape.Tech providers create an environment where agents,robots,and people can work together in harmony.2AI moves from

11、 thought to action as the age of agentic AI dawns.Agents gain the capacity to understand,plan,and act on their own.And that changes everything.1The 2025 UiPath AI+Automation TrendsOVERVIEWUIPATH.COM|AI AND AUTOMATION TRENDS 20254AI moves from thought to action as the age of agentic AI dawns.GenAI ha

12、s laid the groundwork for a great leap forward in applied AI:one that will profoundly transform organizations,the people who work in those organizations,and the very nature of work itself.That leap forward is agentic AI,a type of artificial intelligence that enables software agents to plan,make deci

13、sions,and adapt autonomously and on the fly.Built on a foundation of GenAI and large language models(LLMs)but given new powers to plan and act through an ensemble of large action models(LAMs)and other advanced AIthese goal-and action-oriented agents can act autonomously to take on a multitude of com

14、plex tasks,nuanced decisions,and end-to-end processes.This new cadre of virtual workers doesnt need to rely on business rules or follow tightly predefined processes to get the job done.Instead,agents will be able to leverage their ability to respond to plain language prompts and event triggers,reaso

15、n through complex processes,take a series of actions to reach a goal,and learn and improve.Theyll be able to plan for and direct the resources and tools required to take these actions,including,most notably,RPA robots(see sidebar:“The future is agentic and robotic”).The advent of agents changes ever

16、ything to come.Imagine,if you will,the opportunities to redesign business processes for greater speed,efficiency,and accuracy.To reallocate work between humans and agents so people are freer to create,collaborate.To redefine the basis of competition and reconceive the operating model for the agentic

17、 age.Imagine working in an organization where agents are making 15%of all decisions autonomouslywhich Gartner predicts will happen by 2028.1 Will agentic become endemic?Signs point to yes.First movers success in a wide range of use cases and industrieswhich have demonstrated significant improvements

18、 in speed,reaction time,personalization,efficiency,and innovationare stirring up interest among both technology analysts and corporate leadership.Gartner has named agentic AI a top 25 technology trend for 20252and IDC notes that global expenditures on agentic AI have already jumped from$0 in 2023 to

19、 almost$400 million in 2024.But thats just the beginning;they estimate the category will grow 10-fold by 2028,to over$4 billion.Thats a 68%compound annual growth rate(CAGR).3 2025 will set the foundation for that growth,with early adoption of agentic use cases;broader usage of the agents that are al

20、ready built into leading technologies;and investment to build out an orchestrated agentic ecosystem across the enterprise.TREND 171715 52 2%of executives say AI agents will drive higher levels of automation in their workflowssay they plan to use agents in 20254The future is agentic and robotic.Will

21、agentic automation replace robotic process automation(RPA)?Not at all.In fact,IDC predicts RPA spending will more than double from 2024 to 2028 to reach$8.2 billion.5Thats because agentic AI opens up more of the organization to automationand agents will turn to secure,efficient,and reliable RPA robo

22、ts to execute a lot of the new work.YOUR 25 TO-DOS:Learn more about agentic AI and agentic automation.Launch at least one process using agentic automation.Along those lines,keep an eye on early moversor become one yourself.Its time to call your agent.UIPATH.COM|AI AND AUTOMATION TRENDS 20255agent ac

23、tions,model input and output,and the logic and data underlying agent decisionsin fact,everything the agent doesneeds to be transparent,governable,predictable,and secure.Clearly,the agentic ecosystem comprises a complex set of capabilities.But make no mistake:its rapidly coming together in 2025and co

24、ming to an enterprise near you.Strike up the orchestration:the agentic ecosystem takes shape.A human workplace without defined roles,systems,and processes would be chaotic,underperforming,and unproductive.And the same holds true for a virtual workplace.Without enabling infrastructure,orchestration,a

25、nd controls,agentic AI wont scale effectively.To capture agentic AIs full potential,enterprises will require vibrant agentic infrastructures.The technology must be capable of enabling a multitude of agents to work effectively across fragmented technology landscapes and highly complex operations.It m

26、ust enable collaboration between agents,robots,and peoplewhile providing control,visibility,and active governance.While many pieces of these capabilities already exist,the complete ecosystem for enterprise agentic automation hasnt yet fully coalesced.But its starting to emerge.In 2025,well see a goo

27、d deal of activity in this area,as a select group of AI technology companiesUiPath among theminnovate and invest to bring enterprise-grade agentic automation platforms to market.Orchestration is one of the most important pieces of the agentic ecosystemcritical for coordinating agents tasks,managing

28、workflows,and optimizing operations across the complex patchwork of enterprise technologies,systems,and apps.This capability must be able to support multiple agents working separately or in collaboration,knitting together decisions and actions into coherent sequences.It needs to facilitate handoffs

29、among AI agents and RPA robots,giving agents ready access to the robotic“arms and legs”they need to get the work done.Capabilities for ensuring the right data gets to the right agent at the right time are also critical,as are context grounding,trigger alerting,and memory and learning systems.In conc

30、ert with agentic orchestration capabilities,agentic automation platforms will also need to provide a place to rapidly build,configure,and launch a wide variety of agents for different processes,levels of authority,and workplace personas.Moreover,they should include a robust prompt library and pre-co

31、nfigured agentsthings that can significantly reduce agent development time and boost performance.Last but certainly not least,the technology must include ways to continually monitor agent activities and assess performance,as well as security,governance,and humanintheloop capabilities.Agent-built wor

32、kflows,TREND 23x3x3636%agentic automation market growth,2024-2025(from$300MM to$900MM)CAGR,2024-20286YOUR 25 TO-DOS:Watch for advances in agentic AI technologytheyll be coming fast.Develop your plan for establishing and scaling your agentic AI ecosystemand take steps to move it forward.“Without orch

33、estration,there is no agentic AI.”-Daniel Dines,CEO and Co-Founder|UiPathAutomation companies get to work so agents can,too.UIPATH.COM|AI AND AUTOMATION TRENDS 20256Agents get to work on long-tail automation opportunities.Agentic automation is an equal opportunity opportunityapplicable across differ

34、ent industries,departments,and activities.And it opens up a long tail of new workflows to end-to-end automation.So,where will agentic automation gain its earliest footholds?And how can that information help you plan your agentic agenda?Some areas to consider:Customer service:Agents can be a huge pro

35、ductivity boost in call centers.For example,McKinsey reports that agents helped a large call center resolve 14%more issues each hour.7 Hyper-personalized sales and marketing:Equipped with real-time data,context,and the right demand and behavioral predictive models,agents can either support human rep

36、s in providing one-to-one serviceor use their own conversational abilities to provide it themselves.Business operations:Agents ability to continuously monitor and assess vast amounts of streaming dataand then decide,plan,and execute in real timeequip them to autonomously manage supply chains,invento

37、ry levels,demand forecasting,and logistics planning.Patient care:Complex data environments,continual monitoring of information flows,probabilistic,non-rules-based decisions made in a changing environmentpatient care is a great place for agents to flex their AI-powered muscles.Software development:AI

38、 is already deeply entrenched in coders activities.For example,its creating more than 25%of the code for Googles products.8 And thats not all:agents can design system architectures,develop and execute tests,auto-debug,and much more.Scientific/pharma research:By autonomously designing and running exp

39、eriments,analyzing results,and formulating hypotheses,agents can expand laboratories capacity and speed up innovation.These are just a start;agentic automation might be your answer anywhere youve got complex,costly,labor-intensive,data-saturated,and time-sensitive processes.Look for use cases that s

40、hare these qualitiestheyre where your agents can do their best work and have the most impact.TREND 3YOUR 25 TO-DOS:In short order,create your short list of best use cases.Know who and what youll need to implement them.Implement a test case or two.“The value that agents can unlock comes from their po

41、tential to automate a long tail of complex use casesthat have historically been difficult to address in a cost-or time-efficient manner.”9-McKinsey&Co.,“Why Agents Are the Next Frontier of Generative AI,”2024Where are UiPath customers adopting agentic?Heres what were seeing at the leading edge:Intel

42、ligent document processing(IDP)and communications mining:Agentic automation is helping them establish end-to-end automation.Banking and financial services:For investment advisors,agents collect,analyze,and develop reports on financial data.In loans,theyre doing things like reviewing documentation,fo

43、rmulating decisions,and auto-generating and processing all the papers.Agents are also monitoring for fraud and compliance.Customer support:Some AI agents gather data and offer real-time recommendations and treatments.Others enable end-to-end,hyper-personalized agent-to-customer interactions.Insuranc

44、e and healthcare claims management:Agents are dramatically speeding up claims processing,from assessing claims submission to resolution.Manufacturing:AI agents monitor the flow of Internet of Things(IoT)data 24/7 in supporting equipment maintenance,product quality,and supply chain optimization.The p

45、ower of agentic automation comes alive in 25.UIPATH.COM|AI AND AUTOMATION TRENDS 20257Job sharing with the machine:the great work reallocation begins.A study from OpenAI estimates that AI could take on half the work of almost 20%all workers.10 McKinsey estimates that,by 2030,30%of all work hours wil

46、l be performed by machines,not people.11 And these studies were completed BEFORE agentic AI had fully emerged.Given the possibilities arising in the agentic age,the new imperative for enterprise workforce management is“redesign and reassign.”Starting in 2025 but extending through the end of the deca

47、de and beyond,enterprises will need to zero in on the huge challenges of reinventing operating models,rescoping jobs,retraining people,and reallocating tasks and processes between virtual and human workers.The C-suite is on the front lines of this change,challenged to usher their companies into a di

48、mly litbut fast-approachingfuture state.Theyll be assisted by a burgeoning rank of consultants and operations designers specialized in conceiving of new AI operating models,managing massive change,and creating and implementing cross-enterprise agentic systems.Human resources will have to retrain and

49、 upskill tens of thousands of employees in using new AI tools and partnering effectively with agents.Theyll need to find new workers with the right combination of technological skills and core capabilities in critical thinking,problem solving,and creativity.Theyll have to rethink hiring plans and re

50、do evaluation and reward systems.IT doesnt get off easy,either.Theyre the ones wholl be asked to build out an AI and automation-infused workplace ecosystem that encourages collaboration among virtual workers,human workers,and AI tools.And the current automation center of excellence(CoE)will emerge a

51、s the new center of agentic automationskilled at building and managing agents as they go about the swift completion of their appointed rounds.Unions,governments,academia,policy makersthis is a global phenomenon that will touch everyone as it plays out.Many thought leaders are predicting significant

52、disruption in labor supply and demand patterns.McKinsey thinks that might be good for workers in STEM,healthcare,legal,and creative areaswhich tend to pay more.But the flipside is that well have an oversupply of office workers,production workers,and customer service reps.TREND 4We begin to rework ho

53、w work gets done.In all,millions of workers around the world will undergo,as McKinsey terms it,“occupational transitions”e.g.,12 million in Europe,and 12 million in the United States.12 13 For perspective,that equals about 5.5%of the European workforce in 2Q 2024,14 and about 7.4%of U.S.workers as o

54、f September 2024.15 And for those that continue working,the way they work will be transformed.The takeaway for the coming year?A huge wave of change is building under us.You need to start now to imagine the futureand position your organization to ride into it successfully.YOUR 25 TO-DOS:Use process

55、mining and task mining to identify jobs with the highest portion of AI-doable tasks.Convene HR,employees,business line owners,IT,and Operations to begin the process of:Mapping out your companys“workforce of the future”Identifying the right agentic technology and tools Upskilling and retraining worke

56、rs Getting ready to reallocate“AI wont take your job.Its somebody using AI that will.”-Richard E.Baldwin,economist and IMD Business School professor,speaking at the 2023 World Economic Forums Growth SummitUIPATH.COM|AI AND AUTOMATION TRENDS 20258Built-in AI lifts enterprises from the trough of disil

57、lusionment.When it comes to capturing AIs potential,the C-suite has become a bit discouraged,and not without reason.Theyre spending many millions of dollars on AI,but only about half of all AI prototypes make it into production.Seventy percent of respondents in a recent survey say theyve launched 30

58、%or fewer of their planned initiatives.16 So,its small wonder that two-thirds are ambivalent or dissatisfied with their own organizations ability to capture AIs gains.17 But that doesnt mean enterprises arent getting a lot of value from AI.Its just coming from a different place:enterprise technology

59、 companies,whose use of AI within their products has simply exploded over the past year.And its not slowing down anytime soon.In fact,Gartner predicts that by 2026,more than 80%of enterprise software vendors will have embedded AI within their productsup from just 1%in 2023.18One of the most ubiquito

60、us AI use cases?Copilots.Theyre flying high in virtually every major enterprise software provider,including Google,GitHub,SAP,Salesforce,Microsoft,and others.(Ours is UiPath Autopilot.)And theyre producing value.About 70%of early users have reported Microsoft Copilot made them more productive and im

61、proved their work quality.19 GitHubs copilot has delivered a 26%increase in task completion.20 Closer to home,UiPath Autopilot for Developers has cut automation development time by 75%.21 And for one of our customers,UiPath Autopilot for Testers has eliminated 50%of their manual testing.22But theres

62、 still more value to be hadcentered around getting people to use their copilots.(Just an example,it seems that 30 to 40 percent of programmerspresumably,tech-friendliesstill arent using GitHubs copilot.23)So we expect a lot of enterprise activity around training,tracking,and incentivizing end users

63、to get on board with their copilots.But theres a lot more AI inside tech companies offeringsand its delivering significant productivity,accuracy,and cost savings to their customers.As an example,take a look at the next page to see how much AI is doing in one of our products,UiPath Test Suite.And tha

64、ts just UiPath.Multiply this type of innovation across all the major enterprise software companies,and the AI impact should be more than enough to lift enterprises from disillusionment to elation.In 2025,enterprises can expect a lot more AI value supplied by their enterprise technology suppliersand

65、should definitely be putting programs in place to make the most of it.TREND 5Tech providers“AI inside”lets users reap gains without pain.YOUR 25 TO-DOS:Ensure your enterprise fully leverages the AI inside your enterprise technology:Understand whats there,and what it makes possible.Train people on ho

66、w to apply built-in tools like copilots and habituate usage.“GenAl-centric technology is moving from hype to a critical enabler for most tech providers.”-Gartner,“Emerging Tech Impact Radar:Artificial Intelligence,”January 2024Test AutomationTest ManagementTest DesignSearch documentationGenerate low

67、-code test automationFix validation errors in test automationUIPATH.AI|AI AND AUTOMATION TRENDS 20259Generate tests for SAP transactionsPerform fuzzy verificationsGenerate coded API test automationSearch project in natural languageGenerate tests for requirementsGenerate synthetic test dataRefactor c

68、oded test automationImport manual test casesQuality-check requirementsGenerate coded test automationGenerate expressionsGenerate test insights reportsExample:What AI does in UiPath Test Suite AI and automation take on many once-manual taskssignificantly improving testers productivity while expanding

69、 test coverage.Source:UiPathUIPATH.COM|AI AND AUTOMATION TRENDS 202510From RAGs to riches:new tools tame the data deluge.Its no secret:data overload is dragging us down.According to a Gartner study,almost half of digital workers say they struggle to find the data they need to effectively perform the

70、ir jobs.The average employee wastes almost 3.5 hours per week just dealing with information burden.And 38%of workers say they must work extra hours to get on top of it.24 Its not just people;disorganized and scattered data also creates difficulties in building and scaling up AI models.In a recent su

71、rvey of AI executives,36%said they had difficulties in obtaining good training data for their models,and 39%said ensuring data security and privacy was a major issue.25Thankfully,new techniques and tools are emerging that leverage GenAI,large language models(LLMs),and other recent breakthroughs in A

72、I science to provide new and better solutions to the data dilemma.Chief among them:knowledge graphs,retrieval augmented generation(RAG),GraphRAG,and internal/private LLMs.Knowledge graphs are graphical representations of real-world entities like events,concepts,and documents.They connect related bit

73、s of information scattered throughout different data repositories.And they can drive big gains.For example,by improving a major e-commerce platforms search engine recommendations,they boosted clickthrough rates by 35%.26 A$1 billion market in 2023,theyre projected to reach$6+billion by 2030(18.1%CAG

74、R).27Retrieval augmented generation(RAG)improves GenAI models performance by giving them access to real-world data while they generate responses,elevating GenAI“from a neat parlor trick to a business advantage.”28 For example,RAG has helped a global consulting firm cut the time consultants spend sea

75、rching for information by 40%,saving$5 million annually.29 RAGs are projected to chalk up a 44.7%CAGR,2024 to 2030,hitting$9 billion by the end of the decade.30 GraphRAG brings knowledge graphs into the RAG process,“super-refining”the data flowing into AI systems.The result:better accuracy,fewer hal

76、lucinations,computation efficiency,and more context-aware responses.One study has found that GraphRAG-based systems can reduce required tokens by anywhere from 26%to 97%.31Internal or private LLMs are AI models securely trained on an organizations proprietary data,and often focus on a particular pro

77、cess or decision stream.Done right,they deliver far more relevant and accurate insights,turning enterprise data into a significant advantage within a companys firewalls.They can be tricky and expensive to build from the ground up,so many companies have been starting with a foundational LLM and furth

78、er refining it with proprietary data.Global spending on LLMs will exceed$22 billion in 2030,growing almost 50%per year from 2024 on.32 These four approaches are fueling a step-change improvement in how enterprises can access,organize,and leverage their data.Now,add in plain language interfaces that

79、let people“ask the machine”to find and synthesize dataand its clear were on the cusp of transforming our relationship with data for the better.TREND 6 7575%Improvement in data relevance by using a specialized LLM versus a general LLM*UiPath DocPath LLM versus ChatGPT-4o,tested by UiPath R&D on a div

80、erse set of enterprise documentsNew ways emerge to change the data gameand people and models get the win.YOUR 25 TO-DOS:As new technologies and tools change whats possible,make“transform data management with AI”a priority project this year.UIPATH.COM|AI AND AUTOMATION TRENDS 202511“Harnessing AI for

81、 good and realizing its myriad benefits requires mitigating its substantial risks.”-Joe Biden,46th President of the United States,“Executive Order on the Safe,Secure,and Trustworthy Development and Use of Artificial Intelligence,”October 2023Regulation escalation:the world acts to rein in AIs power.

82、While AI continues to unlock new technological frontiers,the window for unfettered experimentation is starting to close.Around the globe,regulators are making it clear that its time for the industry to mature and embrace a culture of accountability.And the courts are getting into the fray,too,with a

83、 number of important AI-related cases around intellectual property,image,and data ownership,among other things.Regulation activity has been building up a head of steam for a decadebut the stunningly fast and broad adoption of GenAI has turned the heat on high.In the U.S.alone,nearly 500 bills have b

84、een proposed across all 50 states in 2024,a significant jump from 130 in 2023.Congress,too,is doubling down on AI oversight,proposing twice the number of bills compared to 2023.The U.S.Federal Trade Commission(FTC)in particular isnt messing aroundits already cracked down on data misuse and algorithm

85、ic biases in a number of cases,notably against Meta for violating a 2012 order related to facial recognition technology.Globally,the quick uptick in regulatory activity is similar,with the EUs landmark AI Act leading the charge.This sweeping legislation,which took effect in August 2024,uses a risk-b

86、ased approach to AI applications.It outright forbids some applications such as government scoring and imposes strict requirements on high-risk systems like those used in employment screening and healthcare.The AI Act also tackles some of the thorny issues around copyright and content ownership,forci

87、ng GenAI models to disclose their training data and comply with copyright laws.And speaking of copyrights,both the U.S.and the EU have recently reaffirmed that 100%-AI-generated output doesnt qualify for a copyrighta human must have been involved.33 But by how much?Still unclear.All this regulatory

88、ruckus is creating significant uncertainty for organizations around the world.Its not that business leaders dont want legislationto the contrary,a Deloitte survey of executives found that 78%sought more AI regulation.34 But they need to know what those regulations will be before theyre comfortable m

89、oving forward.In fact,“regulatory uncertainty”was the largest single reason when companies were asked what was holding their GenAI initiatives back(36%).35 Beyond legislation,the judiciary branch is also getting into the fray.Recent high-profile lawsuits,like Getty Images case against Stability AI f

90、or alleged copyright infringement,underscore the growing legal risks associated with AI development.The upshot is that organizations leveraging GenAI must prepare for heightened scrutiny regarding the origin and usage rights of their training data.Over the course of 2025,organizations will really st

91、art to bear down to make themselves regulation-ready.They have a significant amount of work ahead of them.Consider:right now only about half have GenAI governance frameworks in place.Less than half conduct formal legislative monitoring.Only one-third keep a formal inventory of their GenAI implementa

92、tions.36 As government and judicial activity around the world quickly transforms“anything goes AI”into a highly regulated industry,thats not going to be enough.TREND 7Legislators,leaders,and lawyers build out AIs legal framework.YOUR 25 TO-DOS:Keep a close eye on the courts and the legislatorsand be

93、 ready for the requirements new legislation could impose.Implement robust data governance and security measures.Prioritize transparency and explainability in AI algorithms.Establish clear accountability structures for AI-related decisions.UIPATH.COM|AI AND AUTOMATION TRENDS 202512BIBLIOGRAPHY1.Natio

94、nal CIO Review,“LIVE from Gartner IT Symposium:The Rise of Agentic AI,”October 2024.2.Gartner*,“Gartner Top 10 Strategic Technology Trends for 2025,”October 2024.3.IDC,“Worldwide Intelligent Process Automation Software Forecast,2024-2028,”August 2024.4.Capgemini,“Harnessing the Value Of Generative A

95、I 2nd Edition:Top Use Cases Across Sectors,”May 2024.5.IDC,“Worldwide Intelligent Process Automation Software Forecast,2024-2028,”August 2024.6.IDC,“Worldwide Intelligent Process Automation Software Forecast,2024-2028,”August 2024.7.McKinsey&Company,“The Promise and the Reality of Gen AI Agents in t

96、he Enterprise,”May 2024.8.Fortune,“Over 25%of Googles Code Is Now Written by AIand CEO Sundar Pichai Says Its Just the Start,”October 2024.9.McKinsey&Company,“Why Agents Are the Next Frontier of Generative AI,”July 2024.10.Eloundou,Tyna,et al.,“GPTs are GPTs:An Early Look at the Labor Market Impact

97、Potential of Large Language Models,”August 2023.11.McKinsey&Company,“The economic potential of generative AI:The next productivity frontier,”June 2023.12.McKinsey&Company,“A New Future of Work:The Race to Deploy AI and Raise Skills in Europe and Beyond,”May 2024.13.McKinsey&Company,“Generative AI an

98、d the Future of Work in America,”July 2023.14.Eurostat,“EU Labour Market-Quarterly Statistics,”September 2024.15.U.S.Bureau of Labor Statistics,“The Employment Situation October 2024,”November 2024.16.Deloitte,“Deloittes State of Generative AI in the Enterprise Quarter Three Report,”August 2024.17.B

99、CG,“AI Radar:From Potential to Profit with GenAI,”January 2024.18.Gartner*,“Emerging Tech Impact Radar:Artificial Intelligence,”January 2024.19.Microsoft,“What Can Copilots Earliest Users Teach Us About Generative AI at Work?”November 2023.20.Cui,Zheyuan(Kevin),et al.,“The Effects of Generative AI o

100、n High Skilled Work:Evidence from Three Field Experiments with Software Developers,”September 2024.21.UiPath Automation Center of Excellence22.UiPath Customer23.Cui,Zheyuan(Kevin),et al.,“The Effects of Generative AI on High Skilled Work:Evidence from Three Field Experiments with Software Developers

101、,”September 2024.24.Gartner*,“Gartner Top 10 Strategic Technology Trends for 2025,”October 2024.25.LXT,“The Path to AI Maturity 2024-An Executive Survey,”March 2024.26.Enterprise Knowledge,“Top Graph Use Cases and Enterprise Applications(With Real World Examples),”February 2023.27.Maximize Market Re

102、search,“Knowledge Graph Market:Global Industry Analysis and Forecast(2024-2030),”September 2024.28.IDC,“GenAI Awareness,Readiness,and Commitment(ARC)Survey,”August 2023.29.IngestAI,“RAG in Enterprise Search:How Retrieval-Augmented Generation is Transforming Search Strategies,”March 2024.30.Grandview

103、 Research,“Retrieval Augmented Generation Market Size,Share,&Trends Analysis Report 2024 2030,”October 2024.31.Edge,Darren,et al.,“From Local to Global:A Graph RAG Approach to Query-Focused Summarization,”April 2024.32.Grandview Research,“Large Language Model Powered Tools Market Size,Share,&Trends

104、Analysis Report 2024 2030,”October 2024.33.Deloitte AI Institute,“The Legal Implications of Generative AI,”October 2020.34.Deloitte,“Deloittes State of Generative AI in the Enterprise Quarter One Report,”January 2024.35.Deloitte,“Deloittes State of Generative AI in the Enterprise Quarter Three Report,”August 2024.36.Deloitte,“Deloittes State of Generative AI in the Enterprise Quarter Three Report,”August 2024.*GARTNER is a registered trademark and service mark of Gartner,Inc.and/or its affiliates in the U.S.and internationally.All rights reserved.