《易觀梅森：2024電信中的AI：運營商和供應商的戰略指南（英文版）（29頁）.pdf》由會員分享，可在線閱讀，更多相關《易觀梅森：2024電信中的AI：運營商和供應商的戰略指南（英文版）（29頁）.pdf（29頁珍藏版）》請在三個皮匠報告上搜索。

1、AI in telecoms:a strategic guidefor operators and vendors October 2024AI in telecoms:a strategic guide for operators and vendorsContents2IntroductionDrivers for AI investment in telecoms and use casesA blueprint for AI adoption in telecomsThe AI ecosystem for telecomsAbout Analysys Masons AI-related

2、 research and insightsAI in telecoms:a strategic guide for operators and vendorsIntroduction:the telecoms industrys exploration of AI is hampered by a lack of certainty about which options to implementAI applications:delivery layerAI platforms:development layerAI infrastructure:compute layerServicea

3、ssuranceChatbotsNetwork planning/designContentgenerationServicebundle creationSecurityNetwork and IT operationsNetworkautomationDatamanagementAI applicationdevelopmentFoundationmodelsMLOpsModeldevelopmentAIgovernance Cloud software(CaaS/IaaS)Cloud-basedcomputeCloudnetworkingSource:Analysys MasonColl

4、ectively,the telecoms industry is continuing its active exploration of ways to use AI technologies to transform networks,operations and customer experience.Individually,there is concern and uncertainty about which AI opportunities to pursue.Industry stakeholders need clarity around which activities

5、will prove instrumental to success in the telco AI market:this includes the key use cases themselves,the nuances associated with these use cases and the most valuable capabilities and ecosystem partnerships to develop.This document provides an overview of our rich variety of AI research and reports

6、to offer a framework that sheds light on difficult questions regarding AI adoption and the critical success factors for operators.Components of the AI value chain AI in telecoms:a strategic guide for operators and vendors4Challenge:telecoms operators are unsure about which AI use cases to prioritise

7、,how to implement them and which vendors to engage to achieve these goalsTelecoms operators recognise the importance of investing in AI but are finding it difficult to prioritise AI use cases and to test and implement them.Several challenges must be addressed for operators to succeed with AI impleme

8、ntation.Many operators understand that AI presents broad opportunities to accelerate and achieve business objectives but are conscious of an awkward conflict:move at speed to avoid being left behind;move carefully to avoid investing in the wrong technology.It is not obvious which AI use cases will r

9、eap the greatest benefits,or the extent of return on investment(ROI),making the initial investment proposition more complex,especially when budgets are limited.Our research highlights additional specific challenges in deploying these technologies.Successful deployment requires access to extensive,hi

10、gh-quality data in an environment with the right technology maturity(notably the infrastructure and skillsets needed to deploy and support AI-related operations).As the industrys adoption of AI technologies evolves,so does the value chain.The AI value chain has seen providers changing roles to captu

11、re greater market opportunities.This more complex(and shifting)value chain makes it harder still for operators to achieve clarity on which partners to engage with.Vendors,on the other hand,lack certainty about where the market gaps are and which opportunities to pursue as a result.CSPs AI challenges

12、1Access to high-quality dataImmature technologyInability to scale AI use case deploymentsOrganisational changes needed to adopt AILack of budgetLack of the right skillsetsDifficulty in justifying the business caseLimited access to AI infrastructure for model developmentLack of executive supportManua

13、l AI/ML operationsSource:Analysys Mason21%18%15%13%12%11%4%4%1%1%1 An Analysys Mason survey of 84 senior operator executives worldwide in September and October 2022.Question:“What are the main challenges that you are facing/or expect to face in achieving telco AI/analytics goals?”AI in telecoms:a st

14、rategic guide for operators and vendorsAI requires a range of new infrastructure assets,including data centres and edge locations fitted with AI cloud infrastructure,with access to data and AI platforms and AI applications.Such assets come at a high cost but are vital to execute AI projects effectiv

15、ely.Operators must carefully weigh up the magnitude and affordability of these AI investments against the capabilities that they support.More-flexible,lower-cost options can offer access to assets,but are more limited in scope.All investments will need to be justifiable in terms of capabilities and

16、outcomes.Operators must be prepared to assess the upfront investment needed for AI and be ready to justify the investmentWith several AI use cases to pursue,and limited budget to fund AI projects,operators must identify and prioritise those use cases that align most closely with their key business o

17、bjectives.Clarity on business priorities is needed before exploring AI use cases to ensure alignment between potential outcomes and priorities.Operators must seek out information that will help them to identify relevant use cases and the solution providers that can implement themOperators must be cr

18、ystal clear about their business objectives for deploying AI before they decide where to prioritise their effortsWith the growing importance of AI tools,operators must be prepared to make wider changes to structures and processes in preparation for rapid and successful implementation of AI projects.

19、Organisational changes will be necessary to ensure common governance and implementation practices.A valuable change is the establishment of AI centres of excellence(CoEs)such as those created by AT&T and Telenor.Operators must also invest in upskilling staff and recruiting talent to facilitate AI pr

20、ojects.Operators need a clear sense of the organisational structures necessary for a successful AI strategy and should develop a plan to achieve themRecommendations for operators:be clear on your aims,modify organisational structures to accommodate AI,rethink your partnerships and focus on RoI AI in

21、 telecoms:a strategic guide for operators and vendorsWith the changes occurring in the AI value chain for operators,AI solution providers must be proactive;they must understand current trends in both the industry and the AI technology market,as well as the implications of these trends on their busin

22、ess.They should track developments as the ecosystem responds to these market trends,and analyse the market to identify partnerships across the value chain that they need to make,to remain well positioned in the AI value chain.The AI ecosystem for telecoms is in a state of flux and AI solution vendor

23、s must be active in forming strong and flexible partnerships across the ecosystemOperators will need guidance from vendors to identify the AI use cases that are most relevant.Vendors should seek to understand operators key priorities in order to explore possible options and identify the solutions th

24、at will help achieve these priorities.Vendors should set up workshops and demonstrations to illustrate how their solutions can help operators meet their business objectives.An emphasis on the outcomes that their solutions have achieved for other customers will be valuable in building operators knowl

25、edge and confidence.AI solution vendors targeting the telecoms sector should engage with operators to elucidate the AI opportunities that can help them to achieve business outcomesAI solution providers should help operators to make a critical assessment of their current AI implementation landscape i

26、ncluding the organisation,technologies,toolsets and infrastructure.These assessments can then be used to identify how relevant solutions can fill gaps to help accelerate operators AI implementation projects.and support operators in developing and implementing blueprints for the effective deployment

27、of AI solutionsRecommendations for vendors:assist operators in understanding the changing AI landscape,help shape their plans,and be open to new and flexible partnershipsAI in telecoms:a strategic guide for operators and vendorsOperators need to maintain a laser focus on those AI uses cases that are

28、 critical to their broader business objectives.This document summarises the most powerful AI investment drivers and offers insight on the use cases that leading operators have deployed and results they have achieved.The best returns on AI investments will come when operators have a clear understandi

29、ng of implementation best practices.These include the technology and organisational changes that operators should consider as they invest in AI.This document provides an overview of these changes and the key components of an AI implementation blueprint for operators.Operators will need to rethink th

30、eir ecosystem of partners,and engage with multiple vendors to optimise their AI blueprint.The successful implementation of AI requires an understanding of the evolving AI value chain and the role that various vendors will play in delivering AI solutions.Clarity on the AI value chain also creates opp

31、ortunities for vendors to discover how to tap into AIs potential to increase revenue.This document presents an overview of the AI value chain and the opportunities around some of its key components.We also discuss the need for vendor partnerships across the AI ecosystem.7Insights and guidance from A

32、nalysys Masons AI research and consulting teams help industry stakeholders to maximise the impact of their AI investmentsSource:Analysys MasonAnalysys Masons strategic guide for operators and vendorsDiscover operators drivers for investing in AI and the use cases that they are developing to address

33、these driversDriversUnderstand the key elements that operators need to invest in to derive maximum RoIBlueprintReappraise the AI value chain and the ecosystem players involvedEcosystemAI in telecoms:a strategic guide for operators and vendorsContents8IntroductionDrivers for AI investment in telecoms

34、 and use casesA blueprint for AI adoption in telecomsThe AI ecosystem for telecomsAbout Analysys Masons AI-related research and insightsAI in telecoms:a strategic guide for operators and vendorsEach generational evolution in telecoms has increased the complexity of networks,services and customer eng

35、agement models.In recent years,industry revenue has remained broadly flat,and this puts pressure on operators to manage costs to maintain margins.Operators have set up business and operational transformation strategies to improve market performance.Analysys Masons perspective Accelerating the adopti

36、on of telco AI to deliver autonomous networks indicates that operators top priorities for investing in AI technologies are improving service quality,increasing revenue,improving customer service and reducing opex.These insights are based on the ongoing discussions we have with operators to enrich ou

37、r research around AI.These business priorities are not new,but what has changed is AIs potential to accelerate progress towards these goals,and to broaden the means by which they are achieved.For example,operators are exploring opportunities to create and deliver new services based on generative AI(

38、GenAI)-including offering AI infrastructure services to enterprise customers-to capture new revenue.At the same time,they are exploring GenAI to gain further operational efficiency through improved employee productivity,with additional positive impact on customer experience.Key drivers of operators

39、investment in AI9Operators investment in AI is motivated principally by the need to grow revenue through enhanced quality of service and also to minimise opex costsSource:Analysys Mason1 An Analysys Mason survey of 84 senior operator executives worldwide in September and October 2022.Question:“What

40、are your organisations business objectives for AI and automation?”(Rank top-three objectives,1=most important)AI in telecoms:a strategic guide for operators and vendorsOperators have been using non-generative AI(non-GenAI)technologies such as descriptive,predictive and prescriptive AI for several ye

41、ars.Tools for network failure prediction and detection,churn prediction and customer sentiment analysis are typically based on these technologies.Over the last 2 years,operators have shifted attention to GenAI.This is a democratic form of AI that anyone can access through tools such as ChatGPT.Opera

42、tors are adopting GenAI to benefit from its potential to support superior productivity and customer experience.The hyperbole around GenAI exceeds the reality,but operators take-up and usage are increasing(as seen in Deutsche Telekoms GenAI activities).10There is currently intense focus on GenAI,but

43、operators technology investments also cover non-GenAI technologiesComparison ofnon-generative AI and generative AI Analyses and interprets existing data to make predictions,decisions or classifications Is typically developed using relatively small AI models that are task-specific Can support predict

44、ion,anomaly detection and recommendation systems Accounts for a larger share of operators AI activities(70%of Deutsche Telekoms total AI activities)1Non-GenAIGenAI Creates new content(for example,text,images,music and code)that resembles its training data Is typically developed using multi-purpose f

45、oundation models(FMs)such as large language models(LLMs)Can support text transcription,summarising and code generation Accounts for a modest proportion of AI activity(30%of Deutsche Telekoms AI activities)1 but is a hot topicSource:Analysys Mason1 Presentation from Ahmed Hafez,Vice President,Technol

46、ogy Strategy at Deutsche Telekom,at the 2024 FutureNet World Event held in April 2024.AI in telecoms:a strategic guide for operators and vendors11Operators have deployed AI for use cases that improve revenue generation and enhance the customer experienceCustomerservicesMarketing Customer service Sal

47、es IoTSmart industriesContact centresChurn prediction(Axiata Group);customer sentiment analysis(Telkomsel);customer value management(Ooredoo Oman)Customer issue prediction and prevention(Etisalat and Pakistan Telecoms Company)Lead management and customer engagement(Vodafone)IoT service or device fau

48、lt prediction and management(Bharti Airtel in partnership with IBM)Smart manufacturing(China Mobile);self-driving vehicles(SK Telecom)Text generation for promotional campaigns(Charter Communications)andcustomer loyalty programmes(Mobily)GenAI-based chatbot(TELUS);call summarisation at call centres(T

49、elstra,Vodafone and Windstream)Creation of communications service products or bundles(Telkomsel)Selected AI use cases that are already being implemented by operators 1/2Revenue-generatingservicesAI use case categoriesExamples of AI use case sub-categoriesNon-GenAI use casesGenAI use cases Source:Ana

50、lysys MasonAI in telecoms:a strategic guide for operators and vendors12AI investments are improving operational efficiency and reducing costs in both the network and non-network areas of operators businessesAI use case categoriesExamples of AI use case sub-categoriesNon-GenAI use casesGenAI use case

51、s NetworksNetwork design and planningNetwork deploymentNetwork optimisationNetwork operationsCyber securityIT operationsSmart capex investment(Vodafone and Orange);capacity planning (Bell Canada andGlobe Telecom)Network equipment,tower and customerpremises equipment quality audits(BT)Spectrum,radio

52、parameter and neighbour optimisation(AT&T,Safaricom and Globe Telecom)Fault management,performance monitoring,troubleshooting and root cause analysis(Orange,AT&T and Globe Telecom)Cyber threat detection and prevention(Deutsche Telekom and SK Telecom)Smart capex planning using natural languageto iden

53、tify areas that have networkconnectivity gaps(Telkomsel)Knowledge search to augment network optimisation tasks(AT&T)Knowledge search for field technicians Selected AI use cases that are already being implemented by operators 2/2Internal ITSoftware developmentAlarm monitoring,noise reduction androot

54、cause analysis(BT and Verizon)IT operations copilotGenAI-based copilot for software development(AT&T);translating code between languagesSource:Analysys MasonAI in telecoms:a strategic guide for operators and vendorsOperators are also investing in AI toolsets to deliver new services to consumer and e

55、nterprise customers.Our Communications service provider AI/analytics activity tracker shows that deal activity around these use cases is gathering pace.1 NTT Docomo offers AI applications for productivity gains,predictive maintenance and the application of the metaverse across industries.Telefnica o

56、ffers AI applications for industries such as health and manufacturing through its Telefnica Tech division.Verizon broadens its enterprise offerings by delivering services that combines partners AI platforms with its 5G MEC platform.Over the last 18 months,operators have been sharpening their focus o

57、n revenue-generating AI use casesOperator announcements of AI activity13Operators in Western Europe and the AsiaPacific region are the most active,but we see a good representation of operators in the Middle East and North America.These operators are building on their internal IT and AI expertise wit

58、hin IT services subsidiaries,to create and deliver these services.Current IT services engagements with enterprise customers are a key channel for these services.Public cloud providers(PCPs)and other AI technology players such as data and AI platform providers are also key channels and development pa

59、rtners.Operator announcements on revenue-generating AI use casesSource:Analysys Mason0501001502001H20192H20191H20202H20201H20212H20211H20222H20221H20232H20231H2024Cumulative dealsCustomerInternal ITNetworkRevenue generating services-0204060North AmericaMiddle East and North AfricaWestern EuropeEmerg

60、ing AsiaPacificDeveloped AsiaPacificNumber of deals1H 20192H 20191H 20202H 20201H 20212H 20211H 20222H 20221H 20232H 20231H 20241 Please note that the data on operator AI activities is based on publicly available data.AI in telecoms:a strategic guide for operators and vendorsOperators AI investments

61、 are already yielding quantifiable results.Many of these relate principally to efficiency but may also improve customer satisfaction and revenue.Other examples are included in our Communications service provider AI/analytics activity tracker,which we publish twice a year.Verizon implemented predicti

62、ve AI algorithms to forecast service-affecting network failures that would affect its FiOS offering.As a consequence,its FiOS revenue increased by 2.3%in 3Q 2019.1 Orange Group is applying machine-learning technologies to its FTTH data to prioritise network problems that must be resolved and to iden

63、tify root causes of network failures and associated remedies.By applying AI and automating resolution workflows,Orange has avoided 280 000 field trips and saved EUR20 million per annum.Globe Telecom in the Philippines complemented its energy savings programme with an AI-based solution,leading to a 3

64、6%reduction in energy usage.China Unicom built an LLM that helped transform the operation and management of home broadband services,and to accelerate drive-test analysis.The operator saw a 60%reduction in onsite handling time and an 80%reduction in data query time.2 14Operators AI investments have i

65、ncreased productivity,reduced costs and improved the customer experienceValue already derived from telco AI implementationReduced time for employees to complete daily activities:China UnicomReduced time to perform network activities:TelefnicaCapex optimisation:Orange and Liberty GlobalImproved spect

66、ral efficiency:Hutchison 3 IndonesiaEnergy savings:Globe Telecom,Three Ireland,Liberty GlobalReduced calls to contact centres:TelefnicaReduced field visits:OrangeFaster response to reported network and customer problems:Globe Telecom and SK TelecomReduced churn rate:SK TelecomReduced customer waitin

67、g time for issue resolution:China UnicomProductivity/efficiency gainsImproved customerexperienceCost reduction1 For more information,see Analysys Masons Verizon:AI and analytics implementation to support FiOS.2 TM Forum Inform(2024),China Unicom uses large language models to transform experience.AI

68、in telecoms:a strategic guide for operators and vendorsContents15IntroductionDrivers for AI investment in telecoms and use casesA blueprint for AI adoption in telecomsThe AI ecosystem for telecomsAbout Analysys Masons AI-related research and insightsAI in telecoms:a strategic guide for operators and

69、 vendorsOperators need a clear blueprint that addresses AI implementation challengesAI applicationsDelivery layer:create and manage AI applications that embedAI models to realise use casesAI platformsDevelopment layer:Deploy data,AI and development platformsto facilitate the creation and lifecycle m

70、anagement of AI modelsAI infrastructureCompute layer:Invest in robust physical locations to hostAI cloud infrastructure required to provide computationand run environments for AI workloads Organisational changeKey components of operators AI blueprintOrganisationAI in telecoms:a strategic guide for o

71、perators and vendorsTo deliver the greatest impact,AI tools need the right context.Trained users must operate within an optimised system of processes,structures and applications,which themselves depend on clean and consistent data.Many operators will need to reorganise themselves to adopt and embed

72、the practices that support accelerated and successful implementation of AI deployments.Leading operators such as AT&T and Telenor are creating centres of excellence(CoEs),and recruiting new C-level executives to represent and direct data/AI activities at board level.Operators are training staff to d

73、evelop internal AI expertise,as detailed in our report Building the operator business case for GenAI adoption.The creation of CoEs reflects an understanding that AI will have a huge impact on the sector.Operators investing in CoEs recognise that this initiative requires careful co-ordination(for exa

74、mple around AI development toolsets,and how data is collected,stored and managed to ease access)to ensure a unified and interoperable approach to AI tools.Without CoEs,operators risk wasting time and money:the adoption of different practices and tools will slow the pace of development,and will be an

75、 inefficient financial investment.Orange and Telefnica are using CoEs to reskill staff,who are learning to create AI solutions and exploring how to operate and manage solutions.Orange has set up a skills challenge with data/AI as a core area of development for employees.Staff members access online t

76、raining from Microsoft AI School,Coursera and others.Organisational factors that can accelerate operators AI adoption17Organisational transformation:operators are reassessing processes,structures and skills to be AI-ready,including setting up CoEsSenior executive vision and supportChange managementN

77、ew governance structureSkills/trainingEmployee buy-inSource:Analysys MasonOrganisationAI in telecoms:a strategic guide for operators and vendorsPhysical locations and AI cloud infrastructure:operators investment in internal and external infrastructure provides the engine to develop AIAI projects typ

78、ically have very specific latency and compute requirements.Operators that are planning AI processing functions across multiple locations will need to build or lease cloud infrastructure for AI.Many operators are planning or already building locations such as centralised data centres and cell site ed

79、ge locations(though many are at an early stage).As technology develops,operator infrastructure will need to facilitate on-device inferencing for smartphones and IoT devices,which will become increasingly important locations for operators.To support their own AI deployment locations,some operators ar

80、e investing in access to cloud infrastructure that can offer the storage and compute power for the AI activities executed at each location.Cloud infrastructure provides the computational power and connectivity that will fuel all AI activities and support the heavy lifting of model training and infer

81、encing.It consists of three main segments:chipsets,cloud hardware(that is,compute,networking and storage equipment)and cloud software.While the cloud architecture for AI use broadly mirrors that for general IT and network services,there are additional requirements for GenAI that operators must under

82、stand to maximise AI.Analysys Masons Cloud and AI Infrastructure research programme assesses these requirements and the market dynamics,opportunities and challenges.For example,our report Building data centre networks for AI workloads:key requirements and considerations for CSPs discusses the impact

83、 of AI workloads on data-centre network performance.Physical locations and cloud infrastructure enablers for AI 18Operators investing in AI cloud infrastructureOperatorAI cloud infrastructure investment SoftBankSoftBank is collaborating with NVIDIA to create AI cloud infrastructure for GenAI and 5G/

84、6G applications.Iliad GroupIliad is investing in an NVIDIA DGX SuperPOD:an AI data centre infrastructure platform to train LLMs.FastwebFastweb has purchased 31 NVIDIA DGX H100 systems,which will be installed to train GenAI models.YTL CommunicationsYTL is working with NVIDIA to build an AI data centr

85、e in Malaysia using NVIDIAs GPUs.YTL plans to use the data centre to customise and deploy a Malay language LLM.Physical location Data centre(public,private,hybrid)Edge(interconnect,metro,industrial)Devices(smartphones,network infrastructure,IoT devices)AI cloud infrastructure Cloud software platform

86、s Hardware(computer networking and storage)ChipsetsOrganisationAI in telecoms:a strategic guide for operators and vendorsThe AI models at the heart of AI tools need to be developed,trained and tested before being deployed in operational applications and end-user services targeted at consumer and ent

87、erprise customers.AI and data platforms provide a safe and consistent environment to carry out monitoring and retraining of the AI models to ensure the accuracy of results.These platforms enable developers to build applications based on these models or embed them within existing applications.The int

88、roduction of GenAI technologies has raised the profile of AI models to being key assets that are critical to the development of AI use cases.Foundation models(FMs)are very large,with billions(even trillions)of parameters,with the capability to address several AI use cases.They are therefore critical

89、 to the development of AI systems.Operators requirements for these platforms are constantly evolving.This creates new opportunities but also further challenges(especially intensified competition)for incumbent providers of these platforms.Analysys Masons research programme AI and Data Platforms track

90、s the evolution of AI technologies and the impact on operators and the vendors offering these platforms.For example,our report Data architecture:how vendors can help CSPs to build a foundation for data management sets out best practices that operators should observe when using AI for networks.Main A

91、I-related technology platforms19AI and data platforms:operators are setting up AI and data platformsfor the creation and lifecycle management of AI-based solutionsAI model development platformsThese systems support the creation of new foundation models from scratch to drive GenAI applications.Data p

92、latformsThese platforms manage the lifecycle of data(including governance)required to train and customise AI models.AI operations platformsThese include MLOps and FMOps(for GenAI developments)that automate and manage the AI lifecycle.AI models These are pretrained models(e.g.foundation models)produc

93、tised and sold to support AI application development.AI application development platformsThese platforms include the application development toolsets required to create intelligent applications based on AI models.Source:Analysys MasonOrganisationAI in telecoms:a strategic guide for operators and ven

94、dorsAI applications are consumable assets created from AI models developed,using tools within the platforms.Operators can create these applications on their own using internal resources or purchase them from vendor partners.Operators may deploy these applications as:AI application platforms.Applicat

95、ions with platforms(middleware required for lifecycle management)tightly integrated to them and sold together by the same vendor.AI point applications.These applications address specific use cases,and are typically developed to be sold using the software-as-a-service(SaaS)model.This approach allows

96、the application to use platform services such as data and AI resources from third parties such as public cloud providers(PCPs).Analysys Masons research covers both application types and our AI/analytics and Data Platforms:worldwide forecast 20232028(produced annually)presents our industry view on ho

97、w operators investments will shift between these applications.These AI applications occur across several domains related to the operators networks and business operations.Given operators recent engagements with satellite communications providers and the complexity usually associated with terrestrial

98、 and space communications,we expect to see operators use AI in this domain.Key telco AI application domains20AI applications:AI applications with embedded AI models inject intelligence-driven workflows into operators processes and operationsAI application platformsApplications and data and AI platfo

99、rm services(embedded middleware)Cloud infrastructure or on-premises deployment environmentAI point applicationsApplicationsData and AI platform services(disaggregated)middlewareData and AI platform services(disaggregated)middlewareCloud infrastructure or on-premises deployment environmentKey:Provide

100、d by application vendorProvided bycloud providerProvided by cloud-agnostic playerSource:Analysys MasonOrganisationAI in telecoms:a strategic guide for operators and vendorsContents21IntroductionDrivers for AI investment in telecoms and use casesA blueprint for AI adoption in telecomsThe AI value cha

101、in for telecomsAbout Analysys Masons AI-related research and insightsAI in telecoms:a strategic guide for operators and vendorsThe value chain starts with the physical infrastructure,which is the base camp that provides the physical locations for AI development and operations.The AI cloud infrastruc

102、ture and platforms are the engine room and mining tools that support AI development.22The AI value chain for telecoms is broad and complex but a mining analogy is a useful way to understand the processes and relationshipsThe most valuable layers of the value chain are AI applications and services.Th

103、ese layers are where the AI models are embedded within applications,network infrastructure,and telecoms and vendor services for specific use cases.Source:Analysys MasonAI in telecoms:a strategic guide for operators and vendorsThe AI value chain in telecoms has many layers,and the ecosystem is comple

104、x and expanding.Operators and vendors will need to understand and navigate it carefully,even at this early stage when technology is relatively immature.AI functionality was previously available as integrated assets within AI platforms for specific use cases.With the introduction of FM/LLMs with GenA

105、I,and the strategic role that they play in accelerating AI adoption,these models are now being productised and sold as separate components.Start-ups such as Anthropic,Cohere and Mistral AI(which develop and sell FM/LLMs)are now active players in the evolving value chain.The vendor ecosystem is also

106、changing as players add new capabilities to their products to disrupt the market and win market share.Strategic investments by Databricks,Snowflake and other data platform providers have transformed platforms into an ecosystem of solutions.Operators have an opportunity to capture new opportunities i

107、n the AI value chain,extending their connectivity and data-centre capabilities to support enterprise customers AI needs.1 SK Telecom targets enterprise AI opportunities by marketing itself as an AI company.It plans to use recent engagements with Anthropic to offer operators GenAI solutions.Deutsche

108、Telekom and Telefnica are taking different paths but aim to generate new revenue from AI and industry skills.Telco AI value chain,including operators and vendors across layers of the chain23The vendor ecosystem around the AI value chain for telecoms is changing and expanding,and operators are invest

109、ing to capture a slice of the market1 Analysys Masons Communications service provider AI/analytics activity tracker includes new service launches from operators that involves the use of AI or delivery of an AI-related service.VendorsOperatorsEcosystem componentAccenture,Amdocs,Deloitte,McKinseyServi

110、cesDeutsche Telekom,NTT DOCOMO,Orange,SK TelecomAmdocs,Ericsson,Netcracker,Nokia,Salesforce AI applicationsDeutsche Telekom,Korea Telecom(KT),SK TelecomAWS,Azure,GCP,Mistral AI,Cohere,OpenAI,Snowflake,Databricks PlatformsKT,SK TelecomArista,AWS,Azure,Cisco,GCP,Red Hat,VMwareAI cloud infrastructureDa

111、ta centre providers including EquinixPhysical infrastructureSK Telecom.SoftBankSource:Analysys MasonAI in telecoms:a strategic guide for operators and vendorsTelco AIs momentum will be sustained by the rising complexity of telecoms networks and operations,and operators continued focus on improving c

112、ost efficiencies and customer experience.Analysys Masons Data,AI platforms and applications and Development platforms:worldwide forecast 2023-2028 provides forecasts for operators predicted expenditure on AI platforms and tools to support AI deployments.It forecasts that spending on data-and AI-rela

113、ted tools will grow from USD10.1 billion in 2023 to USD13.6 billion by 2028.Non-GenAI tooling will continue to be the largest constituent of AI investment,but spending on GenAI will see the fastest growth.Spending on GenAI will be driven by the need to improve business performance,especially as oper

114、ators remain under pressure to reduce costs and improve customer experience.Although it is early days for GenAI,operators are keen to harness its potential to transform employee productivity and create operational efficiencies.AI applications will be the primary focus for investment:operators are in

115、creasing the variety of AI applications they invest in while also scaling up existing deployments following the progression from proof of concept to production.The telco AI market will continue to grow and those vendors that have the solutions to accelerate telco AI need to understand how to capture

116、 revenue and increase their market share.Operator data,AI and development product spending worldwide24Operators desire to improve business performance will push expenditure on data,AI and development tools up to USD13.6 billion by 2028Source:Analysys MasonAI in telecoms:a strategic guide for operato

117、rs and vendorsExamples of strategic partnerships between vendors and operators within the AI ecosystemThe complexity and continuous changes associated with the telco AI value chain,alongside other market trends such as operators increasing adoption of cloud data and AI platforms services,highlight t

118、he need to invest in partnerships across the AI ecosystem to maximise their telco AI market opportunities.Analysys Masons report,The rise of platform services:application vendors need a new strategy,emphasises the changes that telco AI application providers will need to make to remain relevant inclu

119、ding establishing partnerships with cloud platform services providers.Other players such as NVIDIA and PCPs are aggressively signing partnership deals,despite being competitors across some layers of the value chain.No single vendor can win in the telco AI market on their own.25To succeed in the telc

120、o AI market,operators and vendors should assess which roles to play and identify strategic partnerships across the broader AI ecosystemEcosystem playerPhysical location providerAI cloud infra-structure providerPlatformproviderAI app providerServices providerExamples of strategic partnerships with op

121、erators and vendorsAWSYesYes/PartnerYes/PartnerYes/PartnerYes/PartnerPartnership with Malaysian operators,Maxis Telecom and CelcomDigi to support AI activities for new revenues and customer experiencePartnership with Ericsson to customise FMs/LLMs for network related GenAI use casesAzureYesYes/Partn

122、erYes/PartnerYes/PartnerYes/PartnerPartnerships with FM providers;Cohere,Meta and Mistral AI,and Model Hub providers;Hugging Face to increase the number of FMs/LLMs that developers can access on AzurePartnerships with several telecoms application vendors such as Amdocs,Netcrcaker and ServiceNow to d

123、eliver GenAI solutionsGCPYesYes/PartnerYes/PartnerYes/PartnerYes/PartnerPartnership with operators such as Telefnica and Orange to explore projects related to AI,GenAI and data services.Partnerships with B/OSS vendors such as Amdocs and Netcracker.NokiaPartnerPartnerYes/PartnerYesYes/PartnerNokia ha

124、s a partnership with Azure and GCP to develop AI and data related solutions using the PCPs cloud infrastructure and platform services.NVIDIAPartnerYesYes/PartnerPartnerYes/PartnerPartnerships with several operators looking to equip their data centres with GPUs to deliver AI-based solutions to enterp

125、rises Partnerships with PCPs and telecoms application players to scale AI adoption in telecoms.AI in telecoms:a strategic guide for operators and vendorsContents26IntroductionDrivers for AI investment in telecoms and use casesA blueprint for AI adoption in telecomsThe AI ecosystem for telecomsAbout

126、Analysys Masons AI-related research and insightsAI in telecoms:a strategic guide for operators and vendorsAnalysys Masons AI-related research programmes give operators an understanding of AI architecture and its potential across the telecoms sectorSource:Analysys MasonAI applications:delivery layerA

127、I platforms:development layerAI infrastructure:compute layerOrganisational changeWireless Technologies(formerly NGW)Automated assuranceCustomer EngagementService Design and OrchestrationNetwork Automation and OrchestrationMonetisation PlatformsEarth ObservationAI and Data PlatformsCloud and AIInfras

128、tructureAnalysys Masons AI-related research programmes help operators and vendors to navigate AI challengesAI in telecoms:a strategic guide for operators and vendors28Connect with usAdaora OkelekePrincipal Analyst,expert in AI and data managementAI is exciting,fast-paced and disruptive.I am passiona

129、te about engaging clients and helping them to navigate this new and complex area so they can make informed choices.Adaora is one of Analysys Masons foremost experts on AI and leads our AI and Data Platforms research and insights programme.Her research focuses on service providers adoption and use of

130、 data management,AI,analytics tools to support the digital transformation of network,customer and other business operations.Adaora also tracks vendor strategies for the telecoms industry to understand how they are evolving their product portfolios to include data,AI and development capabilities.She provides key industry insights to operators and vendors on strategies for adopting these