《IDC:欧洲、中东和非洲(EMEA)中的生成式人工智能:机遇、风险和未来(英文版)(53页).pdf》由会员分享,可在线阅读,更多相关《IDC:欧洲、中东和非洲(EMEA)中的生成式人工智能:机遇、风险和未来(英文版)(53页).pdf(53页珍藏版)》请在三个皮匠报告上搜索。
1、Generative AI in EMEA:Opportunities,Risks,and Futures Neil Ward-Dutton,Jack Vernon,Elena Semenovskaia,Melih Murat,Ewa Zborowska,Adriana Allocato With contributions from across the IDC EMEA analyst teamAt recent IDC EMEA events and in countless customer inquiries,in just the pastthree months generati
2、ve AI has become one of the most common topics ofconversation.In the first three months after its introduction in November 2022,ChatGPT reached100 million users,while services such as DALL-E,Midjourney,and Stable Diffusionhave become hugely popular.These services have lit a huge fire under an area o
3、fresearch and development that had already been very active(although withoutmuch public attention)for at least five years.Now,everybody wants to talk about generative AI.Whats possible?Whatscoming next?What should we worry about?Where(if anywhere)is the moneygoing to be made?This ebook provides an o
4、verview of generative AI in EMEA,including relatedopportunities,risks and challenges,and a first look at what it might mean fordifferent ICT market segments and communities.Large parts of this ebook arebased on an internal survey of IDCs EMEA analyst team.This is a very fast-changing area,and we kno
5、w that some of the detail referenced inthis ebook will become out of date in the coming months.Nevertheless,we hope itwill provide at least an introductory perspective.About This eBookIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Five Key Facts About Generative AIIDC eBook Generati
6、ve AI in EMEA:Opportunities,Risks,and Futures SummaryThree main use case clustersThere are three main clusters of use cases in enterprises:focused oncustomer and employee experiences;software development and delivery;and knowledge management.A wave of excitementThe current wave of industry attention
7、 is focused on consumer-friendlyimage generation and chatbot services(DALL-E,Stable Diffusion,Midjourney,ChatGPT,Bard,etc.)that are based on large language models(LLMs).The great remixerGenerative AI enables computers to create synthetic data or content basedon previously created content.No true int
8、elligenceDespite the hype,the current wave of generative AI innovation does not putus on a path to artificial general intelligence(AGI).These systems also comewith challenges and risks that all adopters must address.Years of quiet advancesBeyond the current wave of excitement around LLMs,industry-sp
9、ecializedsystems(e.g.,in advanced manufacturing and pharmaceuticals)as well asother generative technologies(like GANs)have been in use for years.76%10%62%38%41%24%The ability to generate text with LLMs is capturing most industryattention with generative AI,but this is just one of six generative AIco
10、ntent domains text,software code,images,videos,sound(including voices),and designs and structures.Generative AI systems are already demonstrating some capability ineach of these domains,particularly in terms of automating isolatedtasks(e.g.,generating plausible text in response to a question).In som
11、e domains,early progress is being made at a higher level.Systems are emerging that can generate processes or plans inresponse to prompts and then decompose those into individualactions that the systems are able to execute by themselves,autonomously.%of IDC EMEA analyst contributors saying generative
12、 AI will have a bigimpact in this domain(each respondent could select up to three)Generative AI currently demonstrating clear capability Emerging generative AI capability-Abstraction+PlansProcessesTasksSound/voiceVideoImagesCodeTextDesigns/structuresAcross the IDC EMEA analyst team contributing to t
13、his ebook,systems that can generate text and software code wereoverwhelmingly cited as having the greatest potential impact onbusiness.IDC EMEA team perspectiveGenerative AI Is More than Just Text But Text and Code Generation Are Likely to Have the Most ImpactIDC eBook Generative AI in EMEA:Opportun
14、ities,Risks,and Futures Our initial research suggests there are three main clusters of potential use cases for generative AI in EMEA enterprises.Beyond these broad industry-neutral clusters of use cases,IDCs EMEA analyst team highlighted important industry-specific use cases,from clinical knowledge
15、management,patient engagement,and drug discovery in healthcare,tofinancial advisory agents in financial services,citizen engagement in government,advanced materials design inmanufacturing,and personalization and content generation in retail.IDC EMEA team perspectiveEuropean Organizations Are Explori
16、ng Three Broad Clusters of Enterprise Generative AI Use CasesIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Softwaredevelopmentand deliverylife cycleCustomer andEmployeeExperience KnowledgemanagementOf large EMEA organizationspolled in March 2023 Digital assistants,copilots,and virt
17、ual agentsMarketing content creationConversation summarizationTranslationCommercial document creation(e.g.,RFIs,RFPs)Report generationContent translationDocument summarizationCode creation and pair programmingTest creationBusiness process and task automationConversational query/configurationInvestin
18、g ingenerative AI in202321%Exploringpotential usecases55%Source:IDC EMEA,Future Enterprise Resilience 2023,Wave 2,March 2023 Organizations with 500+employees,n=220Employee productivityCostsRevenueTime to marketSecurity and risk issuesProfitabilityOperational efficiencyCustomer acquisition/retentionE
19、mployee retentionWorkforce stabilityNegative impactPositive impactEmployee productivity tops the list of positive impacts becausegenerative AI could help professionals and administrators save timeon varied tasks and processes.Similarly,the optimization of operations both in terms of reducedcosts and
20、 higher profitability will drive investments.However,security and risk concerns need to be addressed to helporganizations make effective and safe investments.The EMEA analyst team expects that in general more junior roles willbe most impacted by the opportunities and challenges of generative AIuse.T
21、his will pose change management challenges for many and willpotentially create competency gaps,due to established approaches tostaff development and promotion(which revolve around junior stafflearning routine tasks,before being promoted into senior positions).As it is adopted by enterprises in EMEA,
22、generative AI could have a significant impact inmultiple areas in terms of how enterprises operate and the outcomes they achieve.Business outcomeLikelihood of strong impactSignificant Impacts of Generative AI on EnterprisesIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures What capabili
23、ties are required to buildand deliver value based on generative AItechnologies?This value chain mapillustrates how things are shaping up.Trainingdata setsAI hardware(e.g.,NVIDIA)InfrastructureModel creationUse cases and applicationsData prepservicesModelcreationtools(e.g.,NVIDIANeMo)AI executionplat
24、forms(from AWS,Google,Microsoft,etc.)Domain-specificmodelsSpecialistbusinessappsConsultingandintegrationservicesEndcustomersGenericbusinessappsFoundation models(BERT,GPT-3,LaMDA,LLaMA,PaLM,etc.)Consumer apps(from Apple,MSFT,Google,etc.)IDC eBook The Emerging Generative AI Value ChainThe number of ge
25、m icons shown in each segment of the value chain(a maximum of three)indicates the extentto which the IDC EMEA analyst team expects,based on current market conditions,each capability to be able tocapture value(and make money).Our current view is that most of the value is likely to be captured bysuppl
26、iers that can offer model creation tools and execution platforms,and suppliers that can create domain-specific generative AI models and business applications built on them.In contrast,horizontal offerings builtwith general-purpose models are much more vulnerable to commoditization.IDC EMEA team pers
27、pectiveGenerative AI in EMEA:Opportunities,Risks,and Futures IDCs initial research suggests there are three main aspects of generative AI that create challenges and risks:the inherent capabilities(and limitations)ofsystems,the processes by which they are created,and the ways they are potentially use
28、d.CapabilityProcessUsageHallucinationIP ownershipBiasCarbon andenergyintensityBland/genericcontentExploitationRogue/unplannedbehaviorFraud,hacking,theft,etc.Mis/DisinformationPlagiarismThe stars next to the individual challenges above reflect the collective views of the IDC EMEA analystteam regardin
29、g generative AI challenges and risks that are most likely to impact market development(a maximum of three).Beyond these,team members also highlighted further potential investmentinhibitors,including the availability of skills,system quality and reliability,perceptions of the potentialfor job losses,
30、the potential impact of regulation,explainability,and compliance challenges.IDC eBook Challenges and Risks Stem From Capability,Process,UsageIDC EMEA team perspectiveGenerative AI in EMEA:Opportunities,Risks,and Futures The EUs proposed AI Act,which is now close to being finalized,has atthe last min
31、ute been amended to impose specific requirements on thedevelopers of LLM-based systems.The amendments require LLMproviders to disclose any copyrighted material used in model training.In March 2023 the Italian data protection authority ruled that ChatGPTcollects data in a way that is incompatible wit
32、h data protection law.OpenAI,the developer of ChatGPT,was given until the end of April2023 to respond to the regulator regarding GDPR compliance;as aninterim measure,OpenAI blocked access to ChatGPT in Italy.The blockhas now been lifted.Particularly within the EU,the implementation and use of genera
33、tive AIsystems will be shaped and restricted by regulation although,as is thecase with AI more broadly,regulators will be challenged to keep pace withtechnology developments.The details will evolve over time,but regulation will overall continue tofocus on shaping how the rights of citizens are balan
34、ced against eachother and against the interests of corporations.Key regulation focus areas to watchIntellectualpropertyPrivacy andsecurityEqualitySustainabilityA strong reflection of the risks and challenges associated with generative AI isthe fast-evolving landscape of policy and regulation,particu
35、larly within the EU.Examples of the pace of change include:The Potential Impact of Policy and RegulationIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures How is the generative AIdevelopment timeline shapingup?IDCs EMEA analyst teamexpects generative AI to becomea major transformation f
36、orce in1224 months,mainstream inthree to five years,and normalor boring in 10 years.In short:the shared perspective is thatgenerative AI is just at thebeginning of making an impact,and its impact will be felt formany years to come.With concerns over risks and challenges,and emerging pressure from a
37、fast-changing policy and regulationlandscape,the development of the generative AI market in EMEA is still significantly uncertain at this point.We seethree potential scenarios in play.Scenario ASteady growth driven by a blend of broaduse cases and specialized industrysystems fueled by aligned AIdeve
38、lopments,in sync with regulation,governance,and clear positive businessoutcomes for adopters.Scenario BInitial surge in growth driven by broaduse cases,halted by concerns oversystem risks and/or questionablebusiness outcomes;return to growthfueled by specialized/controlled industrysystems.Scenario C
39、High initial surge in growth driven byunmanaged experimentation,killed byweak business outcomes stemming frompoorly managed implementations.TimeEnterprise investmentFuture Possibilities for Generative AI in EMEAIDC eBook IDC EMEA team perspectiveGenerative AI in EMEA:Opportunities,Risks,and Futures
40、increases thechances of marketdisruptionRated 3.69/5 risks outweighpotential valueRated 2.46/5 introduces manyopportunities to deliverbusiness valueRated 3.62/5Generative AI is overratedRated 2.33/5The IDC EMEA Team View:Opportunity and Disruptive Potential Outweigh RisksIDC eBook Taking everything
41、into account,the IDCEMEA analyst team is emphatic aboutthe potential for significant generative AIimpacts and opportunities,and is onlymoderately concerned about thepotential market development barriersposed by risks and hype.IDC EMEA team perspectiveGenerative AI in EMEA:Opportunities,Risks,and Fut
42、ures IDC eBook Realize that generative AI has limitations and that systems need to be handled and implemented with care.Look forways to put humans in the loop whenever consequential decisions will be made based on AI systems.Adopt an experimental lab approach that tests the boundaries of what is pos
43、sible and builds expertise andknowledge,while also clearly measuring impacts and assessing risks and shortcomings.Search for specific use cases,taking into account the breadth of potential generative AI capabilities(beyond text andchat)and the risk profiles of specific uses(and data sources).Systems
44、 leveraging generative AI are systems that automate activities and automation is always business change.Some may fear significant workforce disruption;others may see opportunities to empower workers and reduce skillshortages.Understand the security,confidentiality,privacy,and quality implications of
45、 generative AI technologies and planimplementations accordingly.Dont get carried away with the technology;start with understanding how it can add value.Define clear goals andobjectives for any investment.Based on the analysis in this ebook,we recommend that EMEA organizations considering adopting ge
46、nerative AI focus on these six recommendations.Focus on value andfunctionalityUnderstandlimitationsStart small andexperimentFind appropriateuse casesUnderstand the impacton work and jobsTake care tomanage risksIDC eBook Advice for EnterprisesGenerative AI in EMEA:Opportunities,Risks,and Futures IDC
47、eBook Some use cases will justify increased customer spending,but generative AI features can add value even if vendorsdont charge extra for them,by increasing stickiness of customer relationships.Develop early partnerships with customers that can help you learn and then demonstrate to others how tod
48、eliver value in the key use cases.Particularly in Europe,customers and prospects will have critical concerns and will operate under regulatoryconstraints.Take time and care to understand their issues,and invest to maximize the transparency of your methods.It is tempting to focus on horizontal use ca
49、ses with the largest theoretical addressable market,but it is likely that theuse cases generating the most value will be industry,market,company,and/or line-of-business specific.Outside the largest EMEA markets,ecosystems of technology and implementation partners that can help customersdeliver value
50、 with generative AI(or AI generally)are underdeveloped.Help to educate regulators and policymakers.The path of generative AI development is uncertain,but ignoring it is not a sensible option.Balance this against thetemptation to hype the technology.Focus on how you can deliver real value to customer
51、s.Based on the analysis in this ebook,we recommend that technology suppliers considering implementing generative AI as part of their product/service portfoliofocus on these six recommendations.Get started,but dont addto the hypeConsider value beyondproduct pricingShow value through realstoriesUnders
52、tand enterpriseconcerns and constraintsThink beyond generichorizontal use casesBuild local ecosystemsIDC eBook Advice for Technology SuppliersGenerative AI in EMEA:Opportunities,Risks,and Futures IDC eBook Digging Deeper:Specific EMEAMarketPerspectivesGenerative AI in EMEA:Opportunities,Risks,and Fu
53、tures IDC eBook SoftwareDeliveryGenerative AI in EMEA:Opportunities,Risks,and Futures Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures IntroductionIn tandem with developments and interest in the use of generative AI for text convers
54、ations,the potential for LLMs togenerate software code has also been quickly seized on(code is,after all,just text).Microsofts GitHub launched CoPilot in June 2021,and the service now helps developers write code in Python,JavaScript,Typescript,Ruby,and more.Generative AI can also potentially generat
55、e test scripts for existing code.Tabnine launched a unit test generationservice for Python,Java,and JavaScript in February 2023.Vendors offering low-code development environments are also actively exploring this space.There are risks,though.Earlier in 2023,for example,engineers from Samsung inadvert
56、ently transferred highlyconfidential,commercially sensitive code into the public domain when using ChatGPT to get insights into the code.Key use cases and opportunitiesBeyond code generation and unit test generation,LLM-based generative AI systems could automatically createcomments for existing code
57、,summarize code,support developer training,and provide automated pair-programming.Generation of test data alongside auto-generated unit test cases is also a likely use case,as is code refactoring,application architecture,and code translation(from one programming language to another).Major transforma
58、tional impact:three tofive yearsMainstream:three to five yearsCommodity:10+yearsMarket Shaping FactorsIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures The potential of commercial IP to leak into thepublic domain if employees use public generativeAI systems is likely to make enterprise
59、s cautious.Customers active in the EU in particular will also becautious of regulatory and reputational risksassociated with inappropriate development of oruse of systems.Risks and challengesBecause LLMs hallucinate,accuracy of generatedartifacts cant be guaranteed.This means that inthe short term a
60、t least code and test generationwill be most useful for prototyping and to helptrain junior staff and new developers rather thanto create production code autonomously.Asmodels become more capable,code refactoringand design copilots may become feasible.Enterprise impactSuppliers of software delivery
61、tools and platforms including suppliers of low-code tools will beexpected to offer increasingly intelligent copilotcapabilities to customers.They will be pushed tooffer private instances of tools and models toindividual clients,to minimize risk of IP leakage.Systems integrators will increasingly be
62、expected touse these tools to help accelerate software projectdelivery,reduce onboarding for new staff(or fornew projects),and improve technology knowledgemanagement.Supplier impactSecurityIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Digging Deeper in EMEA:Key use cases and opport
63、unitiesPotential use cases for generative AI center on enhanced threat detection(analyzing large volumes of data toidentify suspicious behavior and detect anomalies that may be indicative of a cyberattack);rapid incidentresponse(analyzing real-time data and providing summary reports and remediation
64、recommendations);anddetecting and prioritizing the elimination of vulnerabilities in software.OverviewIDC eBook Major transformational impact:12 monthsMainstream:24 monthsCommodity:three to five yearsGenerative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures IntroductionIn early 202
65、0(before the current wave of interest in generative AI really began),a Japanese company wasdefrauded of$35 million when criminals cloned the voice of the companys CEO using AI tools and used fakevoice calls and messages to convince a manager to wire the money to a fake lawyer as part of a fake merge
66、r.The potential impact of generative AI is a hot conversation topic in the security community,and could haveboth positive and negative impacts.Generative AI can be used by threat actors to create programs used in malware,ransomware,and phishingattacks.On the other hand,security technology vendors ar
67、e exploring how generative AI can augment orenhance defensive capabilities,particularly around real-time threat detection/response and incidentmanagement.Microsoft has launched Security CoPilot,which processes alerts and translates them into a natural languageexplanation of the security incidents wi
68、th guidance on how to contain them.IDC eBook Market Shaping FactorsThe major risk factors shaping the security and compliance spaceare not factors that directly impact the creation of securityofferings(which will be trained on data about threats,rather thanon private organizational data),but factors
69、 that affect the overallsecurity environment.The use of generative AI systems by unwaryor untrained users may increase the risk of data leakage,whichmay in turn lead to that data being used to create new threats.Risks and challengesSecurity technology vendors such as Microsoft,Google,Orca,SentinelOn
70、e,and Veracode are already embracing generative AI to help improve the capability oftheir tools.Generative AI could make security technology users more productive,help automate tedious repetitive aspects of security tasks,and enable morediverse user groups to get more value out of security technolog
71、y investments.Improvements are likely to particularly benefit security analysts,giving them moretime to focus on more complex tasks.At the same time,the security community is working to understand the business risk implications of GenAI.Reported incidences of sensitive information leakagevia ChatGPT
72、 include software engineers reviewing confidential source code,and employees copying and pasting confidential data to generate a presentation.Inresponse,companies are implementing policy guardrails on internal use by employees.Privacy compliance could become an issue if other national regulatorsfoll
73、ow Italys lead in compelling generative AI vendors to alter their offerings.Enterprise impactSecurity technology vendors will be expected to introduce new features orofferings to help organizations detect and respond to new threatsassociated with malicious use of generative AI,as well as introducing
74、 newfeatures or offerings to help organizations control and manage proprietaryor personal data that might be used to train generative AI systems.A newopportunity is evolving around access control for generative AI systems,toenforce access permissions for confidential data or content that might beret
75、rieved by systems in response to prompts.Supplier impactGenerative AI in EMEA:Opportunities,Risks,and Futures IDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures CloudDigging Deeper in EMEA:OverviewIDC eBook Major transformational impact:now to 6 monthsMainstream:1218 monthsCommodity:24+
76、monthsGenerative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures IntroductionIts no surprise that the providers of hyperscale public cloud platforms have so far been at the forefrontof generative AI research and development.Until now,effective generative AI R&D has required hugeamou
77、nts of compute and storage capacity for model training,and,where public services have beenlaunched,enough infrastructure capacity to support millions of users.Microsoft now owns a major stake in pioneer OpenAI;Google has developed LaMDA and PaLM(and hasa competitor to ChatGPT,Bard);AWS is offering i
78、nfrastructure services for third-party developers(Bedrock and Titan)as well as new hardware options.Key use cases and opportunitiesUse cases for cloud providers start with the generic:primarily,providing tools and execution venues fortraining and inferencing LLMs.Opportunities will also quickly emer
79、ge relating to providing customerswith private instances of LLMs,together with tools to help customers fine-tune those instances andcreate and manage custom models.Additionally,like other big tech companies,cloud providers are pursuing investments and allianceswith AI start-ups(Google acquired Anthr
80、opic,AWS allied with StabilityAI).These partnerships usuallyresult in multiyear preferred cloud partner deals.It is likely that the public cloud platform vendors willdevelop these strategies further,offering increasingly sophisticated and diverse marketplaces for LLMsand LLM-based services.IDC eBook
81、 Market Shaping FactorsBecause todays providers of hyperscale publiccloud platforms are providing most of the digitalinfrastructure powering the development of thismarket,and are hosting todays consumer-facingservices,they are the most likely to come undersignificant regulatory scrutiny.The new EU A
82、I Act,along with the GDPR and the forthcoming EU DataAct,all have significant implications for cloudproviders.Additionally,quick growth of generativeAI may bring risk to cloud operations:over-utilization of cloud platforms may result in slowerperformance,increased cost,or even a highercarbon footpri
83、nt.Risks and challengesPublic cloud providers leadership positions,together with the current economics of deliveringLLM-powered services,mean that many enterpriseswill use public cloud providers services at scale,even for sensitive use cases(where they will still beused for experimentation and proto
84、typing).By usingpublic cloud providers services,enterprises willbenefit by having access to advanced models andinfrastructure,even when they dont have theresources(computing,skills,money)to develop orimplement their own platforms.Moreover,enterprises may end up using public cloud providersindirectly
85、 via AI companies running their services inthe public cloud.Generative AI potential may furtherdrive cloud adoption,but it will require a very strongbusiness case and cost sustainability from cloudproviders,as weak cost control mechanisms willdiscourage companies from using public clouddeployments.E
86、nterprise impactClearly,public cloud providers stand to benefitfrom the explosion in interest in generative AI.The tools and infrastructure they provide arelikely to pull new compute-intensive workloads(and associated large data volumes)to theirplatforms.Of course,both global hyperscalecloud provide
87、rs and sovereign cloud providerswill be expected to offer cutting-edgeinfrastructure and services that minimize thecosts of model training and inferencing,as wellas ensuring that customers can experiment anddeploy safely in compliance with relevantregulations.Development of generative AI willfurther
88、 drive the need for chip and hardwareinvestments from cloud providers;those thatcan invest will have an advantage.Supplier impactGenerative AI in EMEA:Opportunities,Risks,and Futures IDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures ContactCenter andUCC Digging Deeper in EMEA:OverviewI
89、DC eBook Major transformational impact:24 monthsMainstream:three to five yearsCommodity:three to five yearsGenerative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures IntroductionContact centers are already investing in systems that leverage large language models,trainingthem with th
90、eir own data to drive improved outcomes for chatbots and virtual agents.Vendors such as Genesys are already using generative AI to power multiple features in theirofferings.Microsoft and Google,both leaders in unified communications technology,have announcedlarge investments in generative AI.Microso
91、ft CoPilot is designed to drive productivity withMicrosoft 365 suite:for example,it can summarize emails and meeting notes,prepareagendas,organize discussion points from meetings,and set action items.Zoom is embracinggenerative AI to power the next generation of its Zoom IQ smart assistant,offering
92、transcriptsummarization,auto-generation of whiteboard content,and more.Key use cases and opportunitiesGenerative AI is driving innovations across both unified communications(UC)andcollaboration.For UC,it is mainly used to create and summarize emails and meeting notes,and for digital assistants.In co
93、ntact centers,the use is more widespread,driven by the needfor continued cost efficiency improvements ranging from improving bot response todriving agent productivity and improving call quality management.Within contact centers,cost pressures mean thatenterprises will increasingly see agent taskstra
94、nsferred to chatbots and virtual agents,butbecause brand engagement is not static,the needfor human agents will remain.Improved customerexperiences will lead to more brand engagement,and there will always be a need for humanintervention to address issues that requirecomplex problem solving and empat
95、hy.Enterpriseconcerns regarding data sharing risks may pushbuyers toward insisting any AI models are privatewith no data shared.Enterprise impactMarket Shaping FactorsIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures The potential of commercial IP to leak into thepublic domain if emplo
96、yees use public generativeAI systems is likely to make enterprises cautious.Customers active in the EU in particular will alsobe cautious of regulatory and reputational risksassociated with inappropriate development of oruse of systems.Risks and challengesNot all use cases will create new revenue st
97、reams,though contact center deployments are more likelyto support premium features than horizontal UCofferings.Concerns over the potential forgenerative AI to be used to create fake voices andfaces may mean suppliers come under pressure toimplement identity verification features that canhelp users c
98、learly distinguish between real peopleand synthetic bots.In contact centers in particular,suppliers will need to ensure that generative AI canbe deployed with humans in the loop to verify(and/or correct)auto-generated content.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Fut
99、ures EnterpriseApplications Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures Major transformational impact:12 monthsMainstream:three to five yearsCommodity:10+yearsIntroductionLarge enterprise software application vendors have been
100、quick to explore the opportunities presented by the newwave of interest in generative AI.For example:Salesforce announced EinsteinGPT in March 2023,and announced a$250 million generative AI fund from itsSalesforce Ventures VC arm.ServiceNows new Vancouver Now Platform release will include a number o
101、f new generative AI powered features.Microsofts CoPilot,available as part of Microsoft 365,introduces a raft of generative AI powered features to improveproductivity within Outlook,Teams,PowerPoint,Excel,and Word.Although SAP has not yet made any formal product announcements of its own,it has triale
102、d a partnership withOpenAI and inked a partnership with IBM to integrate Watson language AI technologies into SAP Start.Key use cases and opportunitiesPotential use cases for generative AI center on customer engagement improvements(via enhanced chatbots andvirtual agents)and employee experience impr
103、ovements(again,via the use of intelligent agents and copilots toassist with internal service and knowledge management).In procurement,the ability to automatically generate RFI documents,and summarize responses,is anotherpotentially attractive use case.Market Shaping FactorsIDC eBook Generative AI in
104、 EMEA:Opportunities,Risks,and Futures Customers active in the EU in particular will becautious of privacy risks associated with modelstrained on individuals data,IP risks associatedwith the transfer of data to public AI systems,andregulatory and reputational risks associated withinappropriate develo
105、pment of or use of systems.Risks and challengesGenerative AI has the potential to makeapplication users more productive,help automatetedious repetitive aspects of knowledge work,andenable more diverse user groups to get morevalue out of enterprise application investments.Enterprise concerns may push
106、 buyers towardinsisting any AI models are private with no datashared.Enterprise impactNot all use cases will create new revenue streams.Suppliers will need to be careful to align withtechnology partners to implement features in waysthat minimize risks and costs,and align with servicepartners to ease
107、 customers paths to value.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Automationand Analytics Digging Deeper in EMEA:IDC eBook Generative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures Major transformational impact:24 monthsMainstream:three to fiv
108、e yearsCommodity:three to five yearsIntroductionAlmost every business software vendor offering low-code tools is exploring howgenerative AI can help further accelerate the creation of systems(or aspects of systems),and so enhance the low-code value proposition.Business automation and analytics platf
109、orm vendors are at the forefront of the low-codemovement,and a range of vendors(from Appian and Pegasystems,through SnapLogic,to Qlik and Informatica)are already announcing integration of generative AI capabilitiesinto their tools and platforms.Key use cases and opportunitiesPotential use cases for
110、generative AI in low-code environments center on naturallanguage interfaces for prototyping and development,generation of test cases,andgeneration of test data.In the analytics space,investments are focused on enablingconversational analytics.OverviewBecause LLMs hallucinate,accuracy of generatedart
111、ifacts cant be guaranteed,which means that inthe short term at least,code and test generation willbe most useful for prototyping and to help trainjunior staff and new developers rather than tocreate production code autonomously.As modelsbecome more capable,code refactoring and designcopilots may bec
112、ome feasible.Specifically relatingto analytics,successful conversational interfaces willenable more business users to engage more withdata and analytics.Here again,junior roles aremore likely to be affected than senior roles.Enterprise impactMarket Shaping FactorsIDC eBook Generative AI in EMEA:Oppo
113、rtunities,Risks,and Futures The potential of commercial IP to leak into thepublic domain if employees use public generativeAI systems to help them build solutions is likely tomake enterprises cautious.Customers active inthe EU in particular will also be cautious ofregulatory and reputational risks a
114、ssociated withinappropriate development of or use of systems.Risks and challengesSuppliers of automation and analytics platformsthat offer low-code tools will be expected to offerincreasingly intelligent copilot capabilities tocustomers.Where code is being generated,suppliers will be pushed to offer
115、 private instances oftools and models to individual clients,to minimizerisks of IP leakage.System integrators willincreasingly be expected to use these tools to helpaccelerate software project delivery,reduceonboarding for new staff(or for new projects),andimprove technology knowledge management.Sup
116、plier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures IT Services Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures Major transformational impact:24 monthsMainstream:three to five yearsCommodity:three to five ye
117、arsIntroductionAs with any innovative new technology with the potential to impact business at scale,IT serviceproviders have been quick to begin to explore the potential advisory,consulting,and systemsintegration opportunities associated with generative AI.For example,Accentures new Tech Vision 2023
118、 initiative extensively references generative AI,Capgemini is publishing extensive guidance on the topic,and Globant has implemented agenerative AI assistant in its low-code application development offering.Key use cases and opportunitiesPotential opportunities for generative AI for IT services span
119、 advisory and consulting services(helping organizations build strategy and capability relating to generative AI);systemsintegration(designing and implementing use cases for generative AI powered systems andservices for clients);project acceleration(using low-code tools alongside generative AI copilo
120、tsto prototype,build,and test software faster);and support service automation(through theincreased use of virtual agents).Market Shaping FactorsIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures IT service providers are not immune from risksand challenges surrounding generative AI.Altho
121、ugh many players will be able to draw onmore technical talent than enterprises,providersmust pay particular attention to the reputationrisk associated with delivering generative AIsystems that may be faulty or inappropriatelydeveloped.They must also take seriously the IPrisk associated with incautio
122、us use of public AI-based systems.More subtly,providers must alsobe careful not to overpromise or mislead clients orprospects.It is easy to seduce clients with grandpromises,but providers must be prepared to backthese promises up with real action.And given thatgenerative AI could reduce billable hou
123、rs onsystem integration projects in particular,there arepowerful forces that may inhibit providersadoption of these technologies for client projects.Risks and challengesWith the exception of highly innovative/disruptiveenterprises with large budgets and highlyspecialized and skilled technology teams
124、,organizations will look to their IT service partners tohelp them navigate the risks and complexitiesassociated with generative AI.Enterprises willnevertheless need to look carefully at their existingpartners capabilities and pedigrees to see if theyhave the skills and processes that will enable the
125、mto deliver value quickly and safely.Enterprise impactSystem integrators will be increasingly pressured toadopt generative AI tools in their project work,withthe potential risk that they begin to cannibalize theirown integration and development servicesbusinesses(in the same way that adoption of low
126、-code tools does).Providers will have to figure outhow they can take advantage of this opportunitywith minimal disruption to their businesses.Juniorroles are likely to be most impacted,but theoversight that more senior roles can provide willbecome perhaps even more important.Because ofquality challe
127、nges,in the short term the impact islikely to be largest on prototyping work and projectswhere a very fast turnaround is required.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Finance andInsurance Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerat
128、ive AI in EMEA:Opportunities,Risks,and Futures Major transformational impact:24 monthsMainstream:three to five yearsCommodity:10+yearsProhibit use for employees(examples include Bank of America,Citi,Goldman Sachs)Use for chatbots(examples include Morgan Stanley to assist financial advisors,HedgeFund
129、 Citadel for software development and information analysis)Use for sentiment analysis and news classification based on language models(e.g.,Bloomberg)IntroductionTo date there has been limited interest in generative AI in the finance and insuranceindustries,although interest is now growing.There are
130、 three main responses in industryso far:Conversational finance and digital financial advisorsFinancial analysisSynthetic data and content generationKey use cases and opportunitiesTo date there are three clusters of potential use cases in finance and insurance:Finance and insurance businesses are lik
131、ely toexplore and implement generative AI in three areas:risk management(exploring new AI models,pushing larger data volumes of diverse datasources);compliance(generating synthetic data totest compliance scenarios);and lines of business(customer experience enhancements withimproved conversational AI
132、).Adoption,however,islikely to be slow,as banks and insurance providersare typically risk-averse.The first adopters are likelyto be neobanks and challengers and U.S.tier 1firms,but all will stay clear of use cases that affectconsumers.Enterprise impactMarket Shaping FactorsIDC eBook Generative AI in
133、 EMEA:Opportunities,Risks,and Futures Financial institutions have so far been earlyadopters of AI,but the reality is that existing AIinitiatives have a poor track record.One reason isthe zero room for error culture.Unless asolution is 100%proven,institutions will notproductize it.This is exacerbated
134、 by regulatoryscrutiny on resilience,reliability,privacy,andtransparency.Given the buzz around generativeAI,these paradigms will be more in focus thanever.Finance and insurance organizations mustbe able to explain how material decisions aremade,so in many use cases for AI systems,transparency and ex
135、plainability are crucial.Risks and challengesOne of the biggest challenges for suppliers will be tomitigate the black box problem meaning alayman essentially needs to be able to follow howdecisions are made.This needs to be demonstratedto regulators.It is crucial for suppliers to putbusiness users i
136、nto a position to develop,test,andfine-tune decision-making models on a user level,given the complexity of credit,risk,and frauddecisions.Simplicity is key.FSI organizations willlook for tangible use cases for generative AI,wherepeers are generating first value.Prioritize specificuse cases for produ
137、ction,rather than the potentialof the technology.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Manufacturing Digging Deeper in EMEA:IntroductionThere is a lot of interest in generative AI,with 27%of manufacturing respondentsin IDCs FERS 2023 Survey saying they have a
138、lready invested in GenAI technologiesand 38%saying they are doing some basic investigations.Most of the exploration of generative AI use cases in manufacturing has focusedon supply chain applications and on non-core manufacturing processes(such aslogistics,purchasing,maintenance,and quality manageme
139、nt).This will change asthe generative AI powered technology will be embedded into the enterprise ITapplications and digital tools.Information quality is key.Over time,managers can lose grip of the practicalaspects of the information they evaluate(where it comes from,how it has beencreated,are there
140、inherent biases in it,etc.),with an increased risk of garbagein/out.In addition,users and developers must not overlook security.Organizations must make sure the technology includes safeguards against biasand jailbreaking,or tricking AI chatbots into disregarding filters intended toprevent the genera
141、tion of dangerous or hateful content.OverviewIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Workflow enablement:Generative AI could be used to review material/productrecords and provide summaries for the logistics planner or warehouse operatorat the point of use.Product information
142、augmentation:Augmentation of supply-related information,such as origin of the product,its journey through the supply chain,certifications,and third-party audits that have been conducted.Knowledge management:Generation of process guidelines and training materialfor internal and external use,such as s
143、upplier management guidelines,includingdelivery condition updates or sustainability frameworks.Sustainability reporting:Helping to create sustainability reports by providingperformance data analysis and insights.A new category of report could becreated(and its production automated)if the outputs of
144、generative AI poweredtechnology and intelligent automation technology are combined.Key use cases and opportunitiesIn manufacturing,there are four main areas where generative AI can add value:OverviewIDC eBook Major transformational impact:24 monthsMainstream:three to five yearsCommodity:7+yearsGener
145、ative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures IDC eBook Market Shaping FactorsLimited data availability and data quality(cited in IDCs FERS Survey2023,manufacturing)due to siloed data sources limits the ability tocontextualize complex tasks within heterogenous production and
146、 supplychain environments.Manufacturers tend to take a conservativeapproach to black box technology a limiting factor for wideradoption of AI as well as generative AI.Generative AI powered solutionswork with labeled data sets for training AI systems.Figuring out how toconduct labeling at a large sca
147、le remains a significant challenge.Risks and challengesGenerative AI technology can analyze data from various sources(purchase orders,invoices,and shipment tracking information)to identify patterns and potentialareas for improvement.It can help develop predictive models that identify changes in dema
148、nd or supply,find the best production and shipping scenario,or findpotential disruptions in the supply chain.It can also help create sustainability reports by providing data analysis and insights into performance.A new category ofreport could be created(and its production automated)if the outputs of
149、 generative AI powered technology and intelligent automation technology are combined.Generative AI can also reduce bottlenecks that might impact process workflows,especially if there is a human element(e.g.,having to wait until somebody has theright information to proceed to the next step of the wor
150、kflow).To obtain the full benefits of generative AI or any other new technology,for that matter manufacturing organizations must test,develop,deploy,and create focus groups that study the application across processes.Tools such as ChatGPT democratize AIin daily operations,but using them for business
151、 purposes requires being 100%sure about their accuracy,reliability,and scalability.Enterprise impactTech vendors providing solutions to manufacturing and supply chain organizationshave already begun implementing generative AI capabilities in solutions such as ERP(Microsoft Dynamic 365 Copilot),CRM(S
152、alesforce),and PLM solutions(SiemensTeamcenter).Tech suppliers will initially expect manufacturers to start usinggenerative AI powered technology perhaps as just one of the information sourcesconsulted when taking a final decision.Beyond this,however,tech suppliers need toshow manufacturers how to u
153、se generative AI as part of complex tasks todemocratize AI in daily operations.Tech suppliers are also ready to support focusgroups that study the generative AI application across processes available inmanufacturing organizations.Supplier impactGenerative AI in EMEA:Opportunities,Risks,and Futures I
154、DC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Government Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures Major transformational impact:three tofive yearsMainstream:three to five yearsCommodity:10+yearsIntroductionG
155、enerative AI has had no major impact yet on government organizations in EMEA.Someorganizations,however,have begun to explore ChatGPT to draft public procurementRFIs and RFPs a process that has historically been very manual and governmentagencies in the UAE plan to leverage generative AI technologies
156、 for public servicedelivery.Generative AI may have a future impact in terms of improving the functionality of publicservice website chatbots,which are currently typically rudimentary.Key use cases and opportunitiesIn the public procurement space,writing RFIs and RFPs requires organizations tocompose
157、 large documents that include an introduction to the business issues,technical specifications,contractual conditions,etc.By training generative AI algorithms on historical procurement records,time spent ondrafting such documents could be significantly reduced;employee oversight wouldbe crucial due t
158、o strict regulatory requirements.Market Shaping FactorsIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Government data is very fragmented and not ofconsistent quality.That increases capability risksbecause generative AI cant always be trained oncomprehensive and good quality data.Civ
159、ilservants are also likely to resist changes broughtabout through use of generative AI.Culture andbureaucratic definition of processes and tasks willneed to be transformed.Risks and challengesOperational tasks such as preparation of tendersand citizen experience functions such ascommunication are be
160、ing considered for generativeAI,but governments will not allocate significantbudgets to generative AI in 2023.Organizations willstart to run pilots in 2024.Enterprise impactThere is an opportunity for ERP and CRMtechnology providers to start to embed generativeAI in their respective procurement and
161、marketingofferings.Professional services firms should scaletheir capabilities to guide governments in theidentification of generative AI use cases and helpwith the organizational transformation that theywill require.Data platform companies andprofessional services firms should target emerginggenerat
162、ive AI use cases as an opportunity toprovide data governance,aggregation,andpipelining solutions to government institutions.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Retail Digging Deeper in EMEA:OverviewIDC eBook Generative AI timelineGenerative AI in EMEA:Oppor
163、tunities,Risks,and Futures Major transformational impact:24 monthsMainstream:three to five yearsCommodity:10+yearsIntroductionRetailers in EMEA want to understand how generative AI can improve efficiency(throughimproved collaboration and communication across functions),produce personalizedmarketing
164、and sales content(to improve the customer experience),produce SEO-orientedcontent and product descriptions,and improve marketing automation.Organizations such as Mattel,JD.com,Carrefour,and Fathy Rashad are actively exploringthese opportunities.Key use cases and opportunitiesPotential use cases for
165、generative AI in retail in EMEA are likely to be focused on thegeneration of SEO-oriented content and product descriptions,as well as drivingmarketing automation and content personalization for customer experienceimprovement.Market Shaping FactorsIDC eBook Generative AI in EMEA:Opportunities,Risks,a
166、nd Futures In retail the main risks that could hamper adoption(or that suppliers will have to counter to enableadoption)relate to data accuracy,proprietary datasharing,customer data privacy and security,andcopyright ambiguity of product images andmarketing campaigns.Risks and challengesGenerative AI
167、 is most likely to impact marketing andcustomer experience functions,as well aschallenging organizations in data privacy,security,and compliance.Enterprise impactTechnology suppliers will need to partner withretailers and brands across the end-to-end valuechain,from product innovation to personalize
168、dengagement and post-purchase customer service.Overall,in the short term,retailers will approachgenerative AI through easy pre-built solutions(i.e.,content marketing and blog writing);in the longerterm,they will want to develop more data-integrated and strategic solutions(i.e.,productideation and pr
169、ototyping).This will require theintegration of existing enterprise solutions withgenerative AI models.Supplier impactIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Healthcare Digging Deeper in EMEA:IntroductionWhile there are many concerns about the challenges of adopting generative
170、AI in healthcare,the expected benefits related to improving clinicalexperience and enhancing patient engagement,along with greater efficiencyof care,are increasingly promising.Generative AI is showing early success in a number of use cases,includingstreamlining medical notetaking.However,the role of
171、 medical professionalswill continue to be essential in most use cases,to tackle data inaccuracies,potential bias,and any weaknesses in AI governance.OverviewIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures Advancing precision medicine:analyzing clinicians notes,medical imaging,and oth
172、erunstructured data to produce recommendations for diagnosis and treatmentsInforming and educating patients on their health status:sharing medical advice andinformation based on data from multiple sources(wearables,EHR,etc.)Improving the clinician experience:automating administrative tasks such as w
173、ritingreferral letters,clinical coding,and summarizing clinical consultationsKey use cases and opportunitiesGenerative AI could be a game changer in healthcare because its a data-rich and text-heavyindustry with increasing demand for automation.We expect generative AI use cases to revolve around:In
174、life sciences,generative AI could be used to identify patterns in clinical trials throughpatient-generated data and comments,or among non-responders to therapy.It could alsoanalyze data sets such as those from clinical trials and genomic data,proposing new drugs,and identifying new purposes for exis
175、ting therapies.OverviewIDC eBook Major transformational impact:24 monthsMainstream:three to five yearsCommodity:10+yearsGenerative AI timelineGenerative AI in EMEA:Opportunities,Risks,and Futures Industry complexity and outdated ITinfrastructure:Care provision is still a human-centric business and c
176、hange managementwill be critical to encourage clinicians/healthprofessionals to adopt generative AI solutions.Data accuracy and bias:Data sets are notalways high quality,and this presentssignificant challenges for safe model training.Health data privacy and security:Regulatorswill struggle to keep u
177、p with advances ingenerative AI in the next few years,slowingadoption until solutions can be demonstratedto deliver an adequate level of safety forpatients.Healthcare organizations will face three types ofchallenge in adopting generative AI:Generative AI has great potential to positivelyaffect the c
178、linician experience at all levels.As anautomation technology,it could dramaticallyspeed up administrative processes and letclinicians focus more on their patients and theirhealth needs.In administrative functions,generative AI will be more impactful for juniorand midlevel management,with a redefinit
179、ion ofroles and possibly some streamlining oforganizations.For clinical roles,the impact interms of employment numbers will be lesspronounced,but it will impact the way care isdelivered and clinical decisions taken.Vendors providing generative AI will beresponsible for highlighting the risks andchal
180、lenges of this technology deployment.Theyneed to provide information and tools tohealthcare organizations that they can use toimplement generative AI safely,accurately,andethically.In particular,the support and thecollaboration of healthcare professionals isessential to find the right balance betwee
181、nprocess automation and human skills in each usecase.IDC eBook Market Shaping FactorsRisks and challengesEnterprise impactSupplier impactGenerative AI in EMEA:Opportunities,Risks,and Futures This ebook has been developed with a wide range of inputs,and we are very grateful for contributions from acr
182、oss the IDC EMEA analyst team.In no particular order,the ebook is based on contributions from:Bo LykkegaardChristopher SilberbergOru MohiuddinMelih MuratGeorge MironescuUzair MujtabaTolga YalcinNitesh RathiGiovanni CervellatiLapo FiorettiAdriana AllocatoSilvia PiaiElena SemenovskaiaTom ZinkShabnam S
183、haikhHarish DunakheJohn OBrienArif Sultan ShiekhEric SamuelFederico MayrDuncan BrownJan BurianMassimiliano ClapsAshok PatelAyse KaptanogluEwa ZborowskaOrnella UrsoRalf HelkenbergList of IDC EMEA Research ContributorsIDC eBook Generative AI in EMEA:Opportunities,Risks,and Futures IDC 2023.All rights reserved.No reproduction without permission.Further Information European Intelligent Process Automation Systems-click hereWorldwide Artificial Intelligence Spending Guide-click hereGenerative AI in EMEA:Opportunities,Challenges and Risks:an IDC EMEA Webcast