IBM:2024年全球銀行和金融市場展望報告(英文版)(52頁).pdf

編號:153748 PDF  DOCX  中文版 52頁 3.82MB 下載積分:VIP專享
下載報告請您先登錄!

IBM:2024年全球銀行和金融市場展望報告(英文版)(52頁).pdf

1、IBM Institute for Business Value|Expert Insights2024 Global Outlook for Banking and Financial Markets Regenerate banking with AI2 Foreword3 The game-changer:How generative AI can transform the banking and financial sectors5 A critical moment11 Communication:The driving force behind the customer expe

2、rience19 A reimagined workforce experience can boost productivity and re-balance costs27 Risk and compliance:The double-edged sword35 Manage a trusted AI platform for value,innovation,and risk 39 The 10 guiding actionsContents1Almost 8 in 10 institutions(78%)are tactically implementing generative AI

3、 for at least one use case.Their tactical approaches vary but trend higher in the risk and compliance space,as well as in client engagement.Additionally,8%of institutions take a broader,more systematic approach by implementing generative AI across a wider set of business domains scaling throughout t

4、he bank.AI priorities reflect omnipresent concerns about riskand client relationships.Almost 60%of generative AI decision makers see higher value in risk control,compliance reporting,and client engagement.Keeping data private and earning client trust is essential to winning engagements.AI governance

5、 is a must-have.Every banker should be an AI risk manager.More than 60%of banking CEOs indicate new vulnerabilities for cybersecurity(76%),legal uncertainty related to operations(72%),difficulties in controlling outcome accuracy(67%),and prejudice from model bias(65%).Generative AI is more than this

6、 years buzzword.For starters,it can redefine a banks competitive edge in client relationships,evolve and streamline core banking operations,and bolster cybersecurity.Key takeaways2Foreword2023 was a groundbreaking year,with large language models(LLMs)opening unexpected doors for all of us to interac

7、t with AI in personal and business life.Banking executives now find themselves both exhilarated and challenged:how can AI innovations alleviate the structural weaknesses in global banking and unlock opportunities for enhanced productivity and profit?The 2024 Global Outlook for Banking and Financial

8、Markets explores the potentially far-reaching impacts of generative AI and more,starting from in-depth analysis of financial reports from nearly 2,000 institutions worldwide.This global outlook,created with insights from IBM experts and a survey of 600 forward-thinking executives,delves into the ind

9、ustry landscape and examines the emergence of 19 innovative use cases of generative AI.Seldom has there been more to discuss.Generative AI is more than this years buzzword.For starters,it can help to redefine a banks competitive edge in client relationships by elevating communication to new levels o

10、f personalization and effectiveness.This empowers banks to capitalize on their long-term investments in cloud and AI technologies,paving the way for customized and digitized customer interactions that add more value.Furthermore,generative AI can assist to evolve core banking operations,streamlining

11、them like never before,beginning with the understanding of code and process complexities.The technology harbors the potential to reshape the workforce experience,boosting productivity and efficiency.Generative AI can be configured to bolster cybersecurity while alleviating the burden of risk and com

12、pliance management.However,it presents challenges for financial organizations as they navigate the balance between value,innovation,and risk on AI platforms that must be both open and trusted.A pragmatic AI governance approach is indispensable,guiding institutions through continuous innovation and a

13、pplication of this new technology.And this approach gets personal:every employee must not only be a risk manager,but an AI risk manager.How can your organization navigate this level of seismic change?Our comprehensive action guide leads the way.The guide helps you to explore your banks AI priorities

14、,integrate data and AI into your core operations,and scale AI enterprise-wide.Generative AI is undoubtedly a hot topic within your C-Suite at this very moment.The 2024 Global Outlook for Banking and Financial Markets can inform and inspire those discussions,shaping strategies and actions that positi

15、on you to take advantage of these unprecedented opportunities.Intrigued?Let us embark together on this exciting journey.Shanker RamamurthyGlobal Managing PartnerBanking and Financial MarketsIBM ConsultingJohn J.DuigenanGeneral Manager,Financial Services,Banking,Financial Markets,and InsuranceIBM Tec

16、hnology3The game-changer:How generative AI can transform the banking and financial sectorsFollowing the astonishing rise of generative AI,artificial intelligence has seized the worlds attention.Executives are either dazzled by bright futures or dismayed by dystopian scenarios,and polarizing boardroo

17、m discussions proliferate.Many banking executives are brainstorming how to assess and prioritize AIs economic potential,estimate access costs,and manage the risks that come with quickly scaling AI enterprise-wide(see Figure 1).The most essential question of the moment:how can technological innovatio

18、n,such as generative AI,help address and course-correct banks financial performance?This issue requires meticulous assessment.Financial institutions need to reflect on the development and potential of large language models(LLMs)to boost productivity,reduce customer friction,and improve the employee

19、experience.The subsequent insights can help avoid unnecessary hype and assess the real impact of generative AI on bank business modelsand define an action plan that mitigates the associated risks.These considerations are grounded by IBM expertise in providing financial institutions with value-added

20、consulting and breakthrough technology,corroborated by a global survey with 600 executives of financial institutions worldwide.Together,theyve shaped the insights revealed in the 2024 Global Outlook for Banking and Financial Markets.ForewordThe game-changer:How generative AI can transform the bankin

21、g and financial sectorsA critical momentCommunication:The driving force behind the customer experienceA reimagined workforce experience can boost productivity and re-balance costsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actio

22、ns414%78%8%No immediate plans Tactical approachSystemic approachFIGURE 1Almost 8 in 10 institutions are tactically implementing generative AI for at least one use case8%take a more systematic,enterprise-wide approach.Our perspective is organized as follows:A critical moment.Short-term revenue gains

23、can be long-term pains as interest rate hikes factor into business and economic expectations.Communication drives the customer experience.Generative AI can tackle the core of a banks relationship advantage:communication with clients.This elevates decade-long investments in cloud and AI to digitalize

24、 and personalize relationships.A reimagined workforce experience can boost productivity and re-balance costs.Generative AI can facilitate core banking transformation and simplification of operating models.Risk and compliance:The double-edged sword.Generative AI challenges the risk tolerance of finan

25、cial institutions and the composition of risk factors,while strengthening cybersecurity and making compliance less onerous.Manage a trusted AI platform for value,innovation,and risk.Winning the banking future by scaling AI enterprise-wide requires strong technical and cultural foundations.A practica

26、l governance of AI guides banks in the continuum of innovation and use cases application.Action guide.We outline 10 guiding actions across three domainsexplore,integrate,and scaleand how to manage them.86%of banking organizations are in production or preparing to go live with generative AI use cases

27、.8%of them are taking a systematic approach,covering all domains with one or more use cases:client engagement,risk and compliance,information technology,and other support functions.60%of them operate primarily in other advanced and emerging economies.On the contrary,14%of organizations have no immed

28、iate plans to work with generative AI.No typical starting point or implementation pattern emerges.The 78%of banking organizations with a tactical approach work on a small number of use cases or domains without significant preference.Yet,they reveal more traction in the risk and compliance space,as w

29、ell as client engagement.Q.What is your institutions approach to implementing generative AI for each use case presented?Note:For a list of domains and use cases,please see page 44.5A critical momentThe years 2022 and 2023 were marked by unexpected changes in macroeconomic conditions,characterized by

30、 spiking interest rates and sustained inflation.These circumstances were fueled by a post-pandemic reassessment of economic health worldwide and the dramatic resurgence of geopolitical risks.Deglobalization of the international supply chain ignited inflationary forces that had been repressed for ove

31、r a decade by accommodative monetary policies.Major central banks,except those in China and Japan,intervened by swiftly raising interest rates to constrain inflation within functional targets.Central bankers started walking a tightrope,balancing the risk of short-term recession with the goal of prom

32、oting long-term economic growth.The financial services industry,which has long been confined to environments of very low-to-negative interest rates,profited from these interest rate hikes.Institutions operating under negative rates saw a robust resurgence of depleted income statements.This was parti

33、cularly due to the significant gap between the reset of interest earned on lending obligations and the time required to reprice borrowing costs on deposits.In 2023,most banks in major advanced and EU economies reported Return on Average Equity(ROAE)above the post-global financial crisis averagea key

34、 performance metric indicating how much income a bank generates per each dollar of shareholder equity(see Figure 2).1 Conversely,the ROAE impact on banks in other advanced and emerging economies was muted,with these banks experiencing a smaller percentage increase in interest rates.These economies a

35、lso emerged weaker from the pandemic:lack of sufficient post-pandemic government guarantees left local banks more exposed to a deteriorating credit cycle.When adjusting 2023 profit margins for longer-term credit risks,the global outlook may lose its luster.A more rapid-than-expected rise in interest

36、 rates can increase profits but also adversely affect the financial position and operations of banks,prompting significant concerns relative to risk management.This new scenario is already testing client appetite for new loans,impacting banks abilities to service existing debt,and imposing a threat

37、to economic growth,potentially leading to a credit crunch.ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical momentCommunication:The driving force behind the customer experienceA reimagined workforce experience can boost productivity and re-balance c

38、ostsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions6FIGURE 2 ROAE:A complicated storyROAE rises above the global financial crisis average in major advanced and EU economiesbut still lags behind in the rest of the world.Media

39、nAverage of median post GFC 1st-3rd quartile-502006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023510152025Major advanced&EU economies-502006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023510152025Other advanced&emerging eco

40、nomiesNote:Represents banks in each region with total assets of more than USD 50 billion.Source:IBM Institute for Business Value analysis of S&P Global data.7According to the European Central Bank(ECB),Euro-area banks significantly tightened their credit standards(for example,banks internal guidelin

41、es or loan approval criteria)since 2022.Simultaneously,the demand for loans from firms and households has sharply decreased,reaching an all-time low since the ECB launched its bank lending surveys in 2003.2In the US,average credit card balances per cardholder topped$6,000 amid persistent inflation,t

42、he highest in 10 years,according to the Federal Reserve Bank of New York analysis.The delinquency rate increased to 3%in Q3 2023,correlating with an all-time high credit card balance of$1.08 trillion in aggregate.The increase was particularly pronounced among millennials and those with auto or stude

43、nt loans.3As discussed in our 2023 Global Outlook,the current spike in rates is unlikely to herald a return to steadier and healthier profit,given the vastly different macroeconomic conditions and weak efficiency ratios of most banks.The cost-to-income ratio(CIR)a measure of operational efficiencyre

44、mains elevated in major advanced and EU economies(see Figure 3).In 2023,the median is 12%higher compared to the rest of the world(excluding China).When China is included,the gap grows to 16%.4The result can be observed in capital markets,which were quick to discount the economic weaknesses of tradit

45、ional banks business models.Since the onset of the global financial crisis in 2008,investment analysts havent championed banking stocks,pushing down banks price-to-book ratios(PBRs)a key metric of bank franchise value(see Figure 4).PBRs trend below 1 for many institutions,suggesting investor apprehe

46、nsions about shareholder value,leading to higher capital costs for banks that opt to issue additional equity.5 Among major advanced and EU economies,only Canadian and US banks boast a PBR higher than 1.6 These banks benefited from faster restructuring and higher interest rates after the global finan

47、cial crisis erupted.PBRs trend below 1 for many institutions,suggesting apprehensions about shareholder value,leading to higher capital costs for banks that opt to issue additional equity.58FIGURE 3 Persistently high CIRsCIRs remain elevated in major advanced and EU economies.Persistently high CIRsF

48、IGURE 2MedianAverage of median post GFC 1st-3rd quartile35402006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 20234550556065703540455055606570Major advanced&EU economies2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023Other a

49、dvanced&emerging economiesNote:Represents banks in each region with total assets of more than USD 50 billion.Source:IBM Institute for Business Value analysis of S&P Global data.9FIGURE 4 PBRs trend lowerPBRs have fallen below 1 for many institutions.PBRs trend lower FIGURE 3MedianPBR equal to 1.01st

50、-3rd quartile0.00.52006 200720082009201020112012201320142015201620172018201920202021202220231.01.52.02.53.03.54.04.55.0Major advanced&EU economies0.00.52006 200720082009201020112012201320142015201620172018201920202021202220231.01.52.02.53.03.54.04.55.0Other advanced&emerging economies1.451.241.171.1

51、31.071.501.25IndiaTaiwanSingaporeAustraliaBrazilCanadaUSA0.710.620.32EUUKJapan ChinaS.Korea0.570.35Note:Countries depicted have banking systems with more than USD 1 trillion total assets in aggregate.Source:IBM Institute for Business Value analysis of S&P Global data.10To regain investor favor,finan

52、cial institutions need to demonstrate that above average financial performance is not extemporaneous.It must correspond to structural changes in business models and related investment in technology.However,many banks were hesitant to pursue deeper transformation in the last decade.Instead,theyve att

53、empted to digitalize existing methods and operations without a sufficient business and technological redesign.The 2019 adage of Mario Draghi,former Chair of the European Central Bank,still rings true:“The necessity to adjust the business model to the digitalization,to the changes in technology,is so

54、mething much more compelling than being angry about negative rates.”7While not a novel concept,its worthy of emphasis:banks risk losing even more ground as the landscape starts to shift more quickly than expected.The acceleration of AI,particularly following widespread access to LLMs,has propelled g

55、enerative AI to the forefront of boardrooms and executive-level conversations.This years Global Outlook presents an opportunity to reflect on the structural weaknesses of global banking and how recent advances in AI can unlock new possibilities for improved performance and efficiency.And,we explore

56、how to scale AI enterprise-wide while addressing risk and compliance concerns.PerspectiveRisk,compliance reporting,and client engagement lead the way in AI potential We surveyed 600 banking executives with decision-making responsibility for AI from strategy to operations.They reflected on where to f

57、ind the greatest business value for 19 generative AI use cases(see“Domains and use cases”on page 44)across four business domains.The top use cases(32%)correspond to applying the technology to risk and compliance challenges and opportunities,followed by improving client engagement(26%).Our survey rev

58、eals how banks are acting on each use case by domain.How many banks are live or about to go live?The result may surprise you.How many are running a pilot or plan to start one this year?Find out whats happening.How many assessed the feasibility and decided not to proceed?Not everyone is ready.And how

59、 many have not yet formulated a generative AI approach?Generative AI is not a one size fits all.See Perspectives on pages 16,24,and 33 to learn more.Risk and complianceClient engagementIT developmentOther support areas32%26%24%18%Q:Rank the banking functions above in which generative AI unlocks the

60、greatest value for your institution.11Communication:The driving force behind the customer experienceEnhancing the customer experience ranked as a top priority for 53%of the financial services CEOs surveyed by the IBM Institute for Business Value(IBM IBV)in 2023,second only to cybersecurity(58%).8 In

61、 an era of increasing digital disintermediation,a banks capability to sustain a competitive level of customer experience is fundamental.Clients,long accustomed to managing a large part of their personal and business lives on mobile apps,are reshaping the landscape.We asked 12,000 consumers where the

62、y prefer to deposit their salary and keep their savings.9 Already,16%of consumers globally are comfortable with a branchless,fully digital propositionwith that figure significantly higher in other advanced and emerging economies.For example,in Brazil,29%of respondents hold a primary account with a n

63、eobank,defined as a firm offering applications,software,and other technologies to streamline mobile and online banking.10When asked how they prefer to apply for loans,execute investment decisions,or buy/renew car insurance,the majority of consumers cite mobile apps and websites as their most favored

64、 option:63%for loan applications,69%for investment execution,56%for car insurance.When selecting a car insurer,a positive claim experience was 44%more likely to sway decisions toward a specific insurer compared to other drivers,such as brand reputation or comparing coverage.11Clients seek seamless i

65、ntegration between digital and analog life,demanding a consistent omnichannel experience characterized by real-time,comprehensive personal relationships.When contemplating switching their main bank accounts,the quality of customer service is a significant factor.In fact,there was a 20%greater chance

66、 this feature could influence their decision when compared with other elements,such as better access to mobile apps or branch proximity.12ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical momentCommunication:The driving force behind the customer exp

67、erienceA reimagined workforce experience can boost productivity and re-balance costsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions12Although branches still play a crucial role in banking intermediation,their number has been

68、 shrinking since 2012,particularly in major advanced and EU economies (see Figure 5).13 This is primarily due to intense M&A activity,but its also hastened by declining demand and revenue pressures,notably resulting from the low interest rates over the last 10 years.Instead,branch networks are growi

69、ng in other advanced and emerging economies,along with economic wealth,banks earning capacity,and the provision of services to previously unbanked citizens and regions.14FIGURE 5 The rate of commercial bank branch openingsand closuresBranches are declining since 2012,particularly in major advanced a

70、nd EU economies.-10%-5%-10%+11%+9%-1.5%+8%-38%-32%+53%+19%-23%-21%-4.5%-4.3%-3.8%Note:Data for Germany is 2020,not 2021.Source:IBM Institute for Business Value analysis of S&P Global data.13Although higher interest rates are boosting global banking profits,a widespread return to branch network expan

71、sion is unlikely,given client appreciation of the convenience and efficiency of digital banking.How have banks responded to clients increasing digital adaptation,and the need to shift more,better,and reimagined experiences from branches to mobile contact points?As exponential technologies mature,ban

72、ks have addressed three strategies supporting better customer service.Each phase corresponds to a different investment focus in their tech portfolio(see Figure 6).Digitalization as accessBanks initially introduced online journeys to supplement branch access,coexisting without differentiating key pro

73、positions.With the advent of smartphones,digital banking matured as a primary engagement platform.We asked 12,000 consumers about their preferred method to perform basic banking transactions,such as accessing their bank account and verifying balances and transactions.62%already said theyre using a m

74、obile app and 12%report using a bank website.15However,interfaces for basic banking services were built and optimized without resolving core banking complexities,limiting the capability to better satisfy digital requirements of the clientele.This hindered banks effectiveness in engaging clients base

75、d on core banking transactions data,even when enriched by other sources of information.As digital banking grew in usage and relevance,cloud technology became essential to elevating the omnichannel experience,providing advanced analytics for timely insights into client behaviors.Notably,the more open

76、 architecture of hybrid cloud underscores the vital role of innovation and the need for easier integration of the fintech ecosystem to shape a more responsive,customer-centric digital banking landscape.FIGURE 6 Evolving with exponential techThree technology pillars to build better customer service D

77、igitalizationMore digital habitsLess branch visitsHybrid cloudPersonalizationLess frictionMore relevanceAI/machine learning/deep learningCommunicationLess passiveMore conversationsGenerative AIEvolving withexponential techFIGURE 5StrategyCustomer experienceInvestment focusSource:IBM Institute for Bu

78、siness Value.14This is reflected in a banks Net Promoter Score(NPS)a measure of how likely customers are to recommend a banks products and services.A high NPS indicates stronger satisfaction in customer relationships,potentially resulting in more business per client and growth via referrals.Accordin

79、g to Survey Sensum in 2023,specialized retailers and smartphone producers boast the highest NPS(see Figure 7).Specialty stores typically engage in“high touch”personalized customer interactions.And superior customer services usually complement the client-centric focus of high-tech gadgets.On the low

80、end of the spectrum are utility services,such as internet and cable providers.This might indicate a gap in addressing key friction in client consumption of highly commoditized services.16And what about financial services?Banking,together with life and health insurance,has room to improve,while inves

81、tment brokers/advisors and credit cards rank higher.This reflects greater alignment with customer needs.17 Personalization as relevanceProviding digital access doesnt automatically translate to personalized,frictionless digital services.The emergence of AImachine learning,deep learning,and now gener

82、ative AIhas granted banks new capabilities to reimagine the customer experience.By leveraging AI processes in search engine optimization(SEO),banks can identify key frictions across the digital landscape,learning more from each client interaction and redirecting actions to improve NPS.Mobile apps ca

83、nt capture the same soft information about clients as human branch managers.But machine learning offers new avenues to extract client preferences and needs from their transactions.“High touch”service equates to high NPS6%Internet serviceCable&satellite TV serviceLife insuranceHealth insuranceCredit

84、cardsInvestment brokers/advisorsSmartphonesSpecialty storesBanking31%37%49%56%59%35%FIGURE 6FIGURE 7“High touch”service equates to high NPSBanking has room to improve.Source:The Role of NPS in Banking and Other Financial Institutions.Survey Sensum.2023.15Notwithstanding the advantages,better insight

85、s and new information alone failed to drive a comprehensive shift in client intermediation toward digital solutions.Mobile,being demand-driven,is effective for users who can self-direct,but most banking revenues originate from offers.Bank officers typically push products to clients who rely on them

86、to make informed financial decisions(for example,investing,insuring,or funding for business ambitions).Without sufficient conversational touchpoints,many clients might pivot to digital banks before resolving important considerations to financial decisions.For instance,while 69%of consumers prefer to

87、 execute investments online,there is a 29%greater chance that financial advisors trigger investment decisions compared to other factors,such as access to planning tools,mobile support,or timely news.18First-generation chatbots,powered by natural language processing(NLP),empowered banks with rudiment

88、ary conversational capabilities to assist clientsbut rarely provided an adequate conversational experience.This limitation restricted their capability to handle client relationships in a confined,rule-based set of domains.As generative AI matures,this barrier is starting to dissolve.Not only can gen

89、erative AI add value in processing information for better categorization,boosting the quality of customer service(see case study,“Large global payments company turns complaints into actionable insights with generative AI”),but LLMs can empower banks to handle digital conversations that better facili

90、tate value-added relationships.While chatbots answer questions,generative AI engages clients and employees of financial institutionsit works for them by assisting them in managing their tasks.Large global payments company turns complaints into actionable insights with generative AI19A global payment

91、 company sought to improve its customer service processes leveraging generative AI.Previously,the company manually categorized millions of customer complaints across thousands of products into broad groups.This complaint analysis process was time-consuming,error-prone,and unresponsive to emerging is

92、sues.Producing actionable insights required a full three weeksinadequate for trend spotting.This reduced customer satisfaction and prompted the need for compliance reviews.By adopting a generative AI model with more than 100 million parameters,the company was able to overhaul this inefficient proces

93、s.The model was trained on public banking data sets and integrated securely into the companys infrastructure.Reinforcement learning with human feedback(RLHF)was employed to align the model to proprietary context and products.This new process cut the time to complete complaint analysis from three wee

94、ks to under 15 minutes,at the same time improving the granularity of categorization with 91%accuracy.Furthermore,generative AI enabled the extraction of keywords,summarization of call notes,and identification of intentsleading to higher-quality customer service.Case study16Communication as a service

95、Can generative AI make digital platforms the core channel to steward client decision-making?Can mobile technology shift from a demand-led method to an offer-oriented engagement in financial services?The revolution of generative AI has the potential to infuse human language proficiency into machine-d

96、riven interactions,elevating the quality and relevance of engagement.This stems from its capability to create content based on analyzing large amounts of data,including text,image,video,or code.Content can be summarized,and questions answered in chats,all consumable via an AI-generated persona.Its n

97、o surprise that generative AI can find timely application to customer services.For instance,NatWest has invested in elevating the quality of Corathe banks virtual assistantwith LLMs(see case study,“NatWests Cora integrates generative AI to enhance customer experience”on page 17).Based on next-best a

98、ctions,responsive technology reduces the gap between forging client intentions with executing financial decisions.Alternatively,private bankers can also augment their capability to reach out to clients with high-end content delivered in unconventional ways.Generative AI models have been designed to

99、assist a leading European private bank in transforming investment research and financial information into podcasts that high-net-worth individuals can consume on their mobile devices(see case study,“Generative AI transforms a leading European private banks research reports into podcasts”on page 18).

100、Added-value content can be personalized to the preferences,interests,and sentiments of each client.Banks are not only looking at client touch points to improve business experiences.Many are equally invested in implementing generative AI to reimagine the internal operational experience of employees.F

101、or example,conversational capabilities can assist bankers in executing the full lifecycle of software development.As well,generative AI can boost efficiencies and processes in fields such as human resources and administrative requests.For instance,IBM deploys an AI HR assistant to free up a signific

102、ant amount of time in the HR function.20 By focusing on customer services and financial advice,banks can consider two co-existing approaches.They can leverage generative AI to fully automate client conversations powering chats with virtual agents.And they can augment the workforce of human agents to

103、 deliver faster and more relevant aid to clients.Among banks that are in production or preparing to go live with these use cases,digital assistants to the workforce are preferred,compared to fully virtual services.With reference to virtual financial advice,this might reflect higher complexity and gr

104、eater risk of hallucination as revealed by our group of experts(see page 37).There is no single pattern in banks approaches to use cases.Statistical evidence indicates that bankers see use cases as fairly independent from one another.A bank with“no plans”for one of the listed use cases might be“prep

105、aring or live”for another one of the same domain.Q:For the above client engagement use cases,what is your institutions approach to implementing generative AI?Digital assistant to customer servicesDigital assistant to financial advisorsVirtual customer servicesLanding operationsTrade financeVirtual f

106、inancial advisors100%PerspectiveDigital assistants to client-facing workforce are the most common starting point.No Not Still Preparingplans approved assessing or live17NatWests Cora integrates generative AI to enhance customer experience21Case study NatWest will deploy generative AI via its virtual

107、 assistant Cora to provide customers with access to a wider range of information through conversational interactions.This innovative capability has been designed to provide a more accessible,human-like interaction for customers looking to compare products and services across the product suite,or who

108、 research information across the NatWest Group websites.According to Wendy Redshaw,Chief Digital Information Officer of the NatWest Groups Retail Bank:“We are a relationship bank in a digital world,building trusted,long-term relationships with our customers through meaningful and personalized engage

109、ment.Building on Coras success over the last five years,were working to leverage the latest generative AI innovations that will help make Cora feel even more human and,most importantly,a trusted,safe,and reliable digital partner for our customers.”Generative AI will be used to access information fro

110、m multiple secure sources previously inaccessible through chat alone,such as products,services,information about the bank,and career opportunities.As a result,customers can ask questions and receive responses in a more natural,conversational style and are also provided with links to requested inform

111、ation.Generative AI will be used to access information from multiple secure sources previously inaccessible through chat alone.18Case study In todays digital age,information is power.Clients express new preferences to stay up to date with current events and opportunities impacting their financial we

112、ll-being.Relevant information must be communicated at the time of client interest,wherever and whenever needed.As larger chunks of daily routines are conducted on mobile devices,clients increasingly prefer listening to podcasts over reading traditional reports.Sifting through these reports and curat

113、ing relevant information is time-consuming for the client.A leading European private bank identified this shift as a strategic opportunity to accelerate the development of personalized content for its clients.The company is currently testing the use of generative AI to transform research reports int

114、o captivating podcasts in a client-centric way.Private bankers will be able to upload research reports and generate podcast summaries catering to specific client interests.Generative AI proves to be an advantageous technology for promoting inclusion and accessibility to clients with visual disabilit

115、ies.The system uses a two-step approachproviding control and verifiabilityto arrive at final communication.Generative AI transforms a leading European private banks research reports into podcasts.22 A leading European private bank is currently testing the use of generative AI to transform research r

116、eports into captivating podcasts in a client-centric way.19A reimagined workforce experience can boost productivity and re-balance costsBanks investments in technology are not confined to user interfaces but also address the need to redesign architectures end to end.These investments account for 14%

117、of a total worldwide expenditure in technology that exceeds$4.5 trillion,according to 2023 analysis by Gartner.23 In 2022,technology expenses in the banking industry averaged 7%of total operating expenses(see Figure 8),and this percentage varied across different regions(see Figure 9).24 Besides tech

118、nology,total operating expenses include workforce,real estate,and other expenses.Yet,correlating the impact of spending in banking technology to financial performance is not easy.More so than in other industries,traditional business models strongly depend on macroeconomic conditions,and regulation m

119、ore clearly defines the spaces of application.Most importantly,the total amount spent does not singularly determine positive or negligible outcomes.Rather,the utilization of technology expenses holds greater significance.For example,effective technology spend can increase the value of services that

120、motivate client willingness to pay for prime experiences,protect economic value from cyberattacks,and embed ecosystem interactions into nonbanking platforms.Contrary to assumptions,research commissioned by the ECON Committee of the European Parliament reveals some banks with lower IT spend outperfor

121、med their higher-spending counterparts,emphasizing the importance of efficient IT utilization.25 Additionally,CIR dynamics(see Figure 3 on page 8)indicate many banks struggled to translate efficiency-oriented initiatives into structural gains.Similarly,investments in innovation did not achieve expec

122、ted productivity outcomes.While total operating expenses have been growing in the last 15 years,the percentage directed to technology changed less than other expense domains.For instance,workforce costs increased by almost 5%of the total amount(see Figure 8).26To significantly impact financial perfo

123、rmance,technology investment must concurrently address automation and augmentation to re-balance bankers contributions to the bottom line and generate more business value per workforce unit.ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical momentCom

124、munication:The driving force behind the customer experienceA reimagined workforce experience can boost productivity and re-balance costsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions20FIGURE 8 Evolution of workforce and tec

125、h expenses since 2007Total operating expenses have increased,but the percentage of operating expenses directed to technology changed less than other expense domains.5.4%-1%20072008200920102011201220132014201520162017201820192020202120221%2%3%4%5%Operating expenses in USD trillions54.6%54.7%54.5%54.8

126、%54.2%53.8%53.2%51.2%51.4%51.4%51.4%51.7%50.6%49.6%50.2%50.0%6.6%6.3%6.3%5.4%5.4%5.6%5.6%6.0%0.971.071.121.211.221.251.281.281.301.301.411.411.321.341.371.374.9%5.2%5.5%5.2%5.2%5.3%4.9%Workforce as%operating expensesTech&comms as%operating expensesWorkforce as%operating expensesTech&comms as%operati

127、ng expensesThe recent advancements in AI significantly expanded the breadth of its capabilities.While the precise impact on productivity is yet to be realized,early signs indicate the potential for this to be a game changer.Yet fully exploiting the benefits is likely to require a very human response

128、:a collaborative effort from industries and regulators,and a wholesale reimagining of business models,and workflows.27Source:IBM Institute for Business Value analysis of S&P Global data.21FIGURE 9 Tech&comms expense ratio by buckets across regionsInstitutions in other advanced and emerging economies

129、 tend to invest a smaller portion than banks in major advanced and EU economies.10-15%32%0-5%28%53%5-10%17%20%15-20%17%4%20-25%6%3%20%Tech&comms expenses over total operating expensesMajor advanced and EU ecomomies Other advanced and emerging economiesPercentage of banks with this ratioSource:IBM In

130、stitute for Business Value analysis of S&P Global data.22The potential of generative AI,automation,and increased productivityAgainst this backdrop,the ascent of generative AI offers new potential for transforming complex,content-driven applications.Its primary objective:augmenting the workforce and

131、delivering better customer services per unit price,more informed advice per conversation,and accelerated time-to-market of application development.In a recent IBM IBV survey on embedded finance,we identified a constraint in banks capabilities to achieve return on investment(ROI).Bankers experience a

132、 gap in how they design services to meet client needs in nonbanking ecosystems and the needs of their digital partners.Nonbanking executives highlight inadequate operational support and slowness to develop APIs.28 Yet,a majority of banking executives dont consider these elements critical factors to

133、succeed in ecosystem-oriented partnerships.When asked what hinders banks capabilities to open architectures that embed their products and services in external digital platforms,72%of executives were still more likely to identify internal roadblocks such as lack of modularity in core banking systems

134、and suboptimal API standards.29Complex monolithic processes persistently impede banks capabilities to redesign value streams for a more dynamic cloud consumption,resulting in a convergence of process debt and technical debt.This is precisely where generative AI emerges as pivotal to achieving full t

135、ransformation of legacy IT automating systems.It starts from understanding and documenting code that already exists,buried under complex processes that lost causality.Expediting software development with generative AIGenerative AI tools,when integrated into a developers process,can document pre-exis

136、ting code,reverse-engineer requirements,and write new ones.Based on user input,generative AI tools can produce high-quality code recommendations.Auto-generated code suggestions can then increase developer productivity by providing straightforward answers and handling routine coding tasks,reducing th

137、e need to“context switch”and conserving mental energy.These code development tools can also help identify coding errors and potential security vulnerabilities.30Using AI code generation software is generally straightforward and available for many programming languages and frameworks,and its accessib

138、le to both experts and junior developers.These tools offer three broad benefits.Typically,they can:Enable developers to generate code faster,reducing the manual work of writing lines of code and freeing developers to focus on higher-value work Test and debug computer code quickly and efficiently Mak

139、e advanced code development accessible to junior developers.31With modernization proving lengthy and more costly than banks estimated,generative AI can also translate legacy code base(expensive to maintain)into newer software languages while driving new ways of working on application modernization.G

140、erman core banking provider Finanz Informatik has demonstrated this well(see case study,“Finanz Informatik is on a modernization journey of its core banking platform”on page 25).23The impact of generative AI on the skills gapFIGURE 10 Evolving priorities in skills and talentEstablishing priorities a

141、nd untangling complexities require strategic skills.201620182023Time management skills and ability to prioritizeAbility to work effectively in team environmentsAbility to communicate effectively Willingness to be flexible,agile,adaptable to changeAnalytics skills with business acumenEthics and integ

142、rityIndustry/occupation specific skillsProficiency in reading,writing,and mathematicsForeign languageCapacity for innovation and creativityBasic computer and software application skills Proficiency in STEM42%40%38%38%35%33%33%32%32%31%31%28%In a banking world eager to embrace technology yet facing s

143、killed labor shortages,the application of generative AI can reduce the skills gapand enhance competitiveness.In the absence of a comprehensive plan for redefining workforce engagement,improved efficiency doesnt translate immediately into operational savings.Realizing the full potential of AI-driven

144、efficiencies necessitates not only reskilling but also adopting a new talent mix.Crucially,unlocking the value of new skills and talent requires a change in the way people work,co-create,collaborate,and execute.Before fully unlocking the potential of STEM skills,institutions must first establish pri

145、orities and untangle organizational complexitiesendeavors that require a strategic approach.That could be one reason why STEM and IT skills have decreased in importance since 2016(see Figure 10).32 AI applications are not immune from risks and compliance hurdles.In the age of generative AI,banks app

146、roaches to data and privacy must be rethought,with considerations such as EU GDPR and IP protection assumed as givens.Source:Augmented work for an automated,AI-driven world:Boost performance with human-machine partnerships.IBM Institute for Business Value.August 2023.24Over the last year,52%of finan

147、cial institutions executed feasibility studies in support areas such as human resources,finance,and procurement.This equals the percentage of studies conducted in IT development(52%),and its followed by 48%in risk and compliance,and 48%for customer engagement.There is no single pattern in banks appr

148、oaches to use cases.Statistical evidence indicates that bankers see use cases as fairly independent from one another.A bank with“no plans”for one of the listed use cases might be“preparing or live”for another one of the same domain.IT development lifecycle is a growing area for immediate application

149、 of generative AI.This is being addressed by 17%of institutions globally.There is no single pattern in banks approaches to use cases.Statistical evidence indicates that bankers see use cases as fairly independent from one another.A bank with“no plans”for one of the listed use cases might be“preparin

150、g or live”for another one of the same domain.PerspectivesGenerative AI is gaining momentum in IT development.Finance and auditing applications of generative AI attract the greatest attention from banks.Q:For the above use cases,what is your institutions approach to implementing generative AI?Develop

151、ment lifecycleTest and bug discoveryCore banking modernization100%No Not Still Preparingplans approved assessing or liveQ:For the above human resources,marketing,and other operations use cases,what is your institutions approach to implementing generative AI?Digital assistant to finance/auditingDigit

152、al assistant to procurementVirtual HRDigital assistant to marketingDigital assistant to HR recruitment services100%No Not Still Preparingplans approved assessing or live25Case study Finanz Informatik is on a modernization journey of its core banking platform33Finanz Informatik embarked on a journey

153、to modernize its central core banking platform OSPlus,a pivotal system supporting over 350 financial institutions.Transitioning from a COBOL-coded structure to a Java-based architecture could prove challenging,yet it was deemed essential to interoperate seamlessly on both IBM z/OS mainframe and Linu

154、x platforms.This initially involved the mapping and prioritization of business processes,helping ensure clarity in the transition from the legacy system to a new platform.The complexity extended beyond technology,entailing a concurrent skill transformation across the organization,leading to a better

155、 developer experience and acquiring young talent in the market.The first investment prepares Finanz Informatik for the potential future introduction of generative AI to assist and further accelerate development design and execution,while reducing risk for the end-to-end lifecycle.An LLM calibrated t

156、o banking-specific coding domains is in discussion.This LLM could support development,from application discovery and analysis to automated code refactoring,and assistance in the COBOL to Java conversion.The first investment prepares Finanz Informatik for the introduction of generative AI to assist a

157、nd further accelerate development design and execution.The first investment prepared Finanz Informatik for the introduction of generative AI to assist and further accelerate development design and execution.2627Risk and compliance:The double-edged swordSound financial institutions navigate financial

158、 and economic uncertainties for and with their clientsand effective risk management is a fundamental tenant.Banking profits rely on adept risk management.This inherently makes every banking employee a risk manager,responsible for generating sustainable risk-adjusted valuewhether in approving loans,o

159、ffering investment advice,or managing IT operations.A key challenge is to risk-manage the complexity of operations.Since the introduction of interest rate derivatives in the 1980s,when interest rates skyrocketed,financial innovation took hold at an accelerated pace,opening a gap between risk managem

160、ent capabilities and IT responsiveness.Investments in technology were often a catch-up game,with business requirements lacking an enterprise-wide perspective and consistent architectural design.The result:banks not only layered new technology onto weak architectural foundations;they ended up develop

161、ing more tactically than strategically and building siloes.The more banks innovated,the more the complexity of their back-and middle-office operations increased,limiting their capability to reign in CIR.The pace of“financial innovation”faltered in the aftermath of the financial crisis,following regu

162、latory requirements for the simplification and central clearing of most derivative contracts.And banks didnt fully grab the opportunity to swiftly simplify spaghetti-like back-and middle-office architectures.With the advent of“fintech innovation”which focuses less on contractual payoffs and more on

163、transformed client engagement with technologybanks realized a profound architectural gap in what makes their operations fit for an open digital world.ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical momentCommunication:The driving force behind the

164、customer experienceA reimagined workforce experience can boost productivity and re-balance costsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions28As many institutions were only midstream in modernizing their core environments

165、,banks capabilities to transform at speed have encountered another roadblock.More than 50%of banking executives reported insufficient API standards and insufficient modularity in core banking systems as the main challenges to compete with ecosystem-oriented integration strategies such as embedded fi

166、nance.34The stratification of operational risks and constraints on both the business and the technology sides makes software development more expensiveand much slowerfor banks compared to their Silicon Valley competitors and partners.With its ability to create,test,and deploy code,generative AI can

167、revitalize software development,which is what banks need at scale.However,an accelerated development experience,grounded on weak architectural foundations,risks a rapid increase in system complexity.Theres potential but also peril.Imagine a scenario in which generative AI accelerates the software de

168、velopment lifecyclebut without corresponding architectural clarity to master business domains,avoid compliance loopholes,and avoid enterprise-wide dependencies.The result could be heightened security and resiliency risks.For example,consider the exponential curve of complexity created from compoundi

169、ng relationships.That is why architectural design and platform governance are more business-critical than ever in gauging operational risks while transformation accelerates,avoiding what could be considered an“event horizon”a black hole that rapidly consumes any possibility of added value.The rapid

170、advancements in generative AI also introduce a mix of challenging operational risks,especially when LLMs are consumed through“as-a-service”models.Limited visibility into the interactions of LLMs within an extensive data supply chain,often comprised of substantial unfiltered data from public domains,

171、necessitates dedicated supervision and stringent controls of outputs (see Perspective,“The generative AI supply chain”).As well,this expanded attack surface heightens third-party risk and cybersecurity considerations.In fact,when asked which concerns slow the adoption of generative AI in their enter

172、prise,88%of banking CEOs identified difficulties in controlling data lineage as the top compliance challenge.This is 27 percentage points more than CEOs in other industries.3529Perspective The generative AI supply chainThe generative AI supply chain refers to the process and flow of components,data,

173、and activities involved in the development,deployment,and maintenance of generative AI models.It encompasses various stages,including data collection and preprocessing,model training and fine-tuning,model validation and testing,deployment and integration into systems,and ongoing monitoring and updat

174、es.The supply chain involves sourcing and curating training data,selecting and configuring AI models,implementing infrastructure and computing resources,and facilitating compliance with ethical and legal considerations.It also includes collaboration with data providers,AI developers,researchers,and

175、stakeholders throughout the lifecycle of generative AI models.The generative AI supply chain refers to the process and flow of components,data,and activities involved in the development,deployment,and maintenance of generative AI models.30Risk management of AIand with AIBanking is and always has bee

176、n a data-intensive industry.The new news:generative AI models employ different data and more varied methods to source data,including online repositories,public data sets,web scraping,APIs,surveys,images,and partner data feeding.An AI model is not coded like other quantitative techniques but learns d

177、irectly from data,recognizing patterns or executing insights-driven decisions without human intervention.Therefore,the data model is intrinsically linked to the AI model.This changes the way banks deal with data management and sheds a spotlight on previously peripheral risks dealing with confidentia

178、lity,integrity,and availability.In 2023,we asked banking executives worldwide which AI-related risks pose severe threat to the economic viability,should their enterprise lack proper forms of governance.36 More than 60%indicated new vulnerabilities for cybersecurity(76%),legal uncertainty related to

179、operations(72%),difficulties in controlling outcome accuracy(67%),and prejudice from model bias(65%)(see Figure 11).FIGURE 11 CEOs weigh in on AI risksGovernance of AI becomes strategic.Operations face higher legal liabilityAccuracy is hard to be assessed:incorrect info as factBias can cause prejudi

180、ce or preferential propositionsCybersecurity faces new vulnerabilities76%67%65%72%Source:Generative AIImpact on hybrid cloud pulse survey.IBM Institute for Business Value.2023.IBM internal information.31For instance,legal rights for intellectual property(IP)must be carefully considered before implem

181、enting any generative AI application,which means due diligence on compliance in distant corners of the AI supply chain.After a decade of digital innovation,AI algorithms are not new to leading banks.Solutions range from automation of mortgage approval,analysis of customer sentiment,or monitoring tra

182、nsaction data to detect fraud.However,development has typically occurred inside siloed banking organizations around localized use cases.As AI takes hold enterprise-wide,this segmentation constrains the governance of a system that blurs the line between business and technology accountability.Lack of

183、clarity about how banks must practically govern and risk-manage AI from a centralized perspective hinders the ability to move forward quickly.Uncurtailed AI systems can exhibit unexpected behaviors or make incorrect predictions in certain scenarios,leading to adverse consequences or impacts.Human ov

184、ersight is the key mitigation action for the most concerning errors,biases,or unintended consequences that may go otherwise unnoticed or unaddressed.But generative AI can also boost the efficiency and quality of human oversight for compliance.By enhancing the understanding of complex regulation,and

185、the application of demanding requirements and controls,generative AI can assist in smoothing the impact of mandatory changes.Another potential impact:generative AIs hunger for data and training comes with substantial energy consumption,creating a potentially negative impact on an institutions sustai

186、nability metrics.However,improvements in extreme ultraviolet lithography(EUV)are expected to reduce environmental,social,and governance(ESG)impact by packing more transistors onto a single chip,granting AI more processing brainpower and performance while using less energy.37Still,generative AI can b

187、e useful to augment the human oversight of risks too complex to manage traditionally.Some banking domains have already embraced AI to support risk mitigation and improve compliance effectivenessfor example,automation for the identification,investigation,and remediation of data breaches.Generative AI

188、 can be useful to augment the human oversight of risks too complex to manage traditionally.32Case study Leading Nordic bank dramatically improves fraud detection with AI in real time38A leading Nordic bank has always prioritized fraud containment with detection systems.And to support that objective,

189、they are investing in AI models.Typically,card and payment transactions run on core banking platforms,while AI models scan transactions through off-platform detection systems.High latency between on-premises core systems and off-premises cloud technology can generate“time-outs,”which could halt the

190、execution of transactions when customers execute payments.To avoid time-outs,real-time checks tend to be executed on a limited number of transactions(estimated to be approximately 10%).The question:Can fraud-detection AI models be run in real-time on 100%of transactions?The bank established their ab

191、ility to develop and train their machine learning models on cloud and deploy them on-premises where the transactions reside.By embedding AI models in core banking platforms,vast amounts of data could be deciphered with business-critical response time,improving from more than 50 milliseconds to 1 mil

192、lisecond.This performance enables fraud detection scan for 100%of transactions.Using AI inferencing on core platforms demonstrated that AI can serve two objectives:improving fraud detection while maintaining a smooth customer experience.33Across business domains,risk and compliance is expected to ge

193、nerate the greatest value for the banks.Its no surprise:this is where more banks are production-oriented.But generative AI is a tale of two cities.On average,15%of organizations are preparing to go live or already live with risk and compliance use cases.Yet,56%will not proceed at this time.There is

194、no single pattern in banks approaches to use cases.Statistical evidence indicates that bankers see use cases as fairly independent from one another.A bank with“no plans”for one of the listed use cases might be“preparing or live”for another one of the same domain.A key consideration:for enterprise ov

195、ersights and application of AI to be effectiveas well as suitable governance deployedorganizations need a new collaboration between humans and machines.Q:For the above risk,compliance,and security processes use cases,what is your institutions approach to implementing generative AI?Transaction monito

196、ringSecurityRisk modellingForecastingCompliance reporting100%Risk and compliance can generate valueyet more than 50%of banks hesitatePerspectiveNo Not Still Preparingplans approved assessing or live3435Manage a trusted AI platform for value,innovation,and riskAI governance enables trustand thats not

197、 an easy task.Financial institutions already possess risk management frameworks for processes that may incorporate AI,including third-party risk,model risk management,and change management processes.Standards and best practices are still taking shape for governance and management of AI risks.Organiz

198、ations are forging broader AI regulation and sorting out banking and financial markets idiosyncrasies.Financial institutions dont have to reinvent the wheel when it comes to their risk management framework in the era of AI.Rather,they can add the new“spoke”of AI and refine as needed to properly acco

199、unt for transparency,robustness,explicability,fairness,and privacy.For instance,agreements with customers and third parties might already be in place,defined before AI took center stage.This demands an understanding of how data can be anonymizedand agreements updated responsiblyto foster compliant d

200、eployment of AI.Technology choice also significantly impacts governance effectiveness.Early adopters of AI often manage models across various tools,applications,and platforms both internal and external.Standardizing these on a common platform is essential for consistent governance.Open platform tran

201、sformations in particular offer advantages by facilitating continuous and consistent integration of proprietary and third-party modelsrecognizing no single AI model can address all use cases.Ultimately,AI governance is a collaborative process involving all functions of a banking organization.It requ

202、ires new ways of working,amplified by changes in data and unstable models still subject to exponential innovation.Practically,good governance is essential in the discovery phase of use cases to help manage compliance against the expected goals and risk profile of the bank.It is also essential to man

203、age implementation and scaling enterprise-wide.ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical momentCommunication:The driving force behind the customer experienceA reimagined workforce experience can boost productivity and re-balance costsRisk an

204、d compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions36Four guiding principlesRisk management and governance,end to endWe identified four guiding principles to manage the end-to-end process.Manage for value Clearly identify,articulate,an

205、d quantify the business and economic value of AI initiatives across the enterprise or the specific domain of application.Align initiatives on a prioritized roadmap,focusing on potential value at scale.Given the multitude of use cases and the high costs that challenge the ROI of AI innovation,investm

206、ent discipline is paramount.Manage for the complexity of innovation Recognize that use cases may vary in terms of innovation and corresponding feasibility.The most innovative use cases might pose greater challenges in terms of data and control.Support federated innovation across business units and f

207、unctions,allowing different parts of the business to explore ideas within a controlled,cohesive environment supported by a Center of Excellence.Manage for risk Define and communicate a risk profile as the core element of governance.Enforce a consistent level of acceptable risk through adequate guidi

208、ng principles,processes,and the IT system configuration.As new types of risk emerge,the risk and compliance processes require continuous calibration to include hallucination,bias,legal,and regulatory compliance.Manage for scale Establish an effective governance framework relying on a common enterpri

209、se platform for data accessibility,model,and use case approval.As AI capabilities might require costly,tailored tools(for example,building,tuning,and/or maintaining large foundation models),they are best developed and risk-managed at scale across an enterprise-wide common platform definition.123437G

210、overnance is not a one-size-fits-all approach,and every bank must cater to its own perspective.Well-conducted AI governance is the foundation of a well-conducted businessand a compelling competitive advantageas it directs business action with clarity and effectiveness.In this regard,we reflected wit

211、h 110 IBM experts about a high-level mapping to value,complexity,and risk for the same 19 use cases included in the international survey of 600 banking executives.The exercise is only illustrative but serves as a worthy starting point for custom reviews inside individual organizations(see Figure 12)

212、.A variety of elements like data access,choice of AI platform,talent availability,and local regulation characterize your banks trade-offs for each use case.Customer engagementLower riskHigher valueDigital assistant to customer servicesVirtual customer servicesLower complexityInformation technologyOt

213、her business areasLower riskHigher valueDigital assistant to financial advisorsVirtual financial advisorsLower complexityLower riskHigher valueDevelopment lifecycleCore banking modernizationLower complexityLower riskHigher valueVirtual HR Digital assistant to finance and auditLower complexityMore fa

214、vorable(higher value,lower complexity,lower risk)Less favorable(lower value,higher complexity,higher risk)FIGURE 12 Mapping use cases to value,complexity,and riskrepresentative samplesExamples to use as starting points in each analysis of use caseswith full data set below.Source:Based on an internal

215、 survey of 110 IBM banking experts to assess a high-level mapping of risk use cases.38Generative AI use casesValueComplexityRiskMore favorable(higher value,lower complexity,lower risk)Less favorable(lower value,higher complexity,higher risk)Customer engagementDigital assistant to customer servicesDi

216、gital assistant to financial advisorsVirtual customer servicesVirtual financial advisorsLending operationsTrade finance796976516464133722454246415854656453Risk and complianceTransaction monitoringRisk modelingForecastingCompliance reportingSecurity746358747244584443376366495359Information technology

217、Development lifecycleCore banking modernizationTest and bug discovery736571244518303423Other business areasVirtual HR Digital assistant to HR recruitment services Digital assistant to marketing Digital assistant to procurementDigital assistant to finance and audit655762575315161620413050373653FIGURE

218、 13 Mapping use cases to value,complexity,and riskfull data setEach use case has been evaluated in terms of potential trade-offs between expected value(new revenue,improved user experience,cost reduction,and better risk management of operations),complexity(readiness of the organization,talent gap,te

219、chnology maturity),and risk(hallucination,bias,security,and privacy).Each evaluation ranges from 0 to 100(see Figure 13).Use cases with value assessed between 0 to 40 are deemed less favorable(in blue)and from 60 to 100 more favorable(in green).Use cases with complexity or risk are estimated between

220、 0 to 40 are deemed more favorable(in green)and from 60 to 100 less favorable(in blue).Source:Based on an internal survey of 110 IBM banking experts to assess a high-level mapping of risk use cases.39The 10 guiding actions Banks are facing shorter-than-ever technology cycles.To manage,they need an e

221、xpedited,structured process to not only move forward,but to move forward quickly.Based on IBMs experience with leading financial institutions working with AI,we share 10 actions that can guide decisions about building generative AI foundations.These actions align with the broader processes of scalin

222、g AI across the firm,including machine learning,deep learning,and NLP(see Figure 14).FIGURE 14Action planExplore,integrate,and scale.Source:IBM Institute for Business Value.Define the AI governance and risk profile of the bank.Prioritize the selected use cases.Formalize the AI strategy for the bank.

223、FIGURE 12123Implement and learn from the first use cases.8Integrate the AI strategy with a scale-up approach.9Organize an“AI factory”to deploy AI confidence at scale.10Infuse generative AI in the software development lifecycle.7Choose the most adequate AI models to match the selected uses.6Choose yo

224、ur AI platform and design appropriate governance.5Explore your banks AI prioritiesIntegrate data and AI into core operationsScale AI(AI+)in your bankEstablish your data foundation on a data fabric.4ForewordThe game-changer:How generative AI can transform the banking and financial sectorsA critical m

225、omentCommunication:The driving force behind the customer experienceA reimagined workforce experience can boost productivity and re-balance costsRisk and compliance:The double-edged swordManage a trusted AI platform for value,innovation,and risk The 10 guiding actions40 Explore your banks AI prioriti

226、es.Define the AI governance and risk profile of the bank.In defining your AI governance,include key criteria for managing your banks risk appetite and assess the potential value of AI implementations adjusted for risk and complexity.Enterprise-wide AI redefines your banks operational risk profile in

227、 terms of new risk factors and their relative relevance.As well,banks need to outline expectations of regulators to facilitate risk-controlled adoption across the enterprise.Demonstrating centralized AI governance of the complete AI model lifecycle is essential for regulatory compliance,transparency

228、,explainability of results,and data security.Key factors here can include data security,ethical considerations,regulatory compliance,and the integration of AI into operations.Successfully identifying and addressing these risks is critical to maintaining customer trust,regulatory adherence,and helpin

229、g ensure responsible use of AI.Prioritize the selected use cases.Prioritize identified use cases based on your banks risk appetite,weighing internally perceived value,feasibility within an acceptable timeframe,and impact on risk profile(see example on page 37).At the same time,encourage creativity t

230、hrough pilot projects that address business themes(for example,enhancement of customer services)and IT topics(for example,assisting code creation).Formalize the AI strategy for the bank,grounded in use case prioritization and planning for skills and resources.Define a master plan for enterprise AI t

231、hat covers all selected use cases,changes in the organization,funding needs,and plans for resource management.Put each use case in the broader context of banking transformation to mutualize costs for AI platform transformation,and gain clarity about marginal contribution to enterprise value.At this

232、stage,its crucial to create a framework and doctrine that guide teams in AI-related work.Key considerations include establishing goals.Does the bank aim to be a“simple AI user”through APIs accessing“black box”models?Or is the goal to be an“AI creator”and gain autonomy in building competitive advanta

233、ges?These decisions identify potential skill gaps.01020341 Integrate data and AI into core operations.0405Establish your data foundation on a data fabric.A strong foundation for AI success requires more than just a methodology or set of principles;organizations also need to modernize their informati

234、on architecture technology.Banks need an architecture designed for AIone that will help them optimize and automate data access and availability,deliver high-quality governed data,and manage privacy and compliance.Simplifying,unifying,and connecting a data fabric across complex,dispersed environments

235、 is critical to building a foundation for successful,timely AI initiatives in a transforming banking business.Choose a trusted AI platform and design appropriate governance.Scaling AI enterprise-wide means applying multiple AI models,since no single model is a best fit in the execution of all use ca

236、ses.Banks cannot advance without a clear decision on which AI platform aligns with their role expectations in the AI value chain.The chosen mix between open-source models and models built in-house is a pivotal factor in platform selection.And independence from any specific cloud provider must be con

237、sidered to preserve autonomy and flexibility to deploy future models.Choose the most adequate AI models to match the selected uses in a multimodal world.Once the platform is chosen,define which generative AI model to onboard for each use case.Consider criteria that includes accuracy,costs,performanc

238、e,carbon footprint,and the modalities by which generative AI can be accessed.This is because generative AI is not only about multiple models but also about multiple modalities.The platform must allow for a broad range of formats for both inputs and outputs.Formats can include text,images,video,voice

239、,audio,and combinations thereof.Other important considerations include the model publishers commitment to quality,the enforcement of privacy protection rules and IP clearance for training data,and the ability to verify the absence of ethical bias in the models used.Infuse generative AI in the softwa

240、re development lifecycle.To leverage generative AI in software development successfully,the technology must be seen as more than just a basic tool for code generation.Generative AI tools,when integrated into a developers process,can document pre-existing code,reverse-engineer require-ments,and write

241、 new ones.Based on user input,generative AI tools can produce high-quality code recommendations.By offering clear solutions and managing routine programming tasks,generative AI can streamline and drive new ways of working.Additionally,it can detect coding errors and expose possible security risksbot

242、h significant advantages.Implement initial use cases and learn from a precise impact analysis to prepare for scale-up(processes,organization,skills,and change management).The deployment of initial use cases provides an opportunity to assess impacts on the operating model,preparing for more structura

243、l updates to organizations and skills.This is when a banks AI practice is forged by concrete actions.Lessons learned are not failures but risk-controlled opportunities to assess and refine your banks AI culture.Immediately promote feedback loops addressing change management and skill transformation.

244、Generative AI itself can be used to scan large volumes of client inquiries,complaints,and communications to build action-oriented summaries for management decisions.06070842 Integrate the AI strategy with a scale-up approach to infuse AI into business processes.The initial use cases that adhere to t

245、he banks risk profile have been validated,and the AI platform transformation has demonstrated its efficacy,laying the foundations for progress.Now,the focus is on momentum and formalizing a scaling strategy for the entire organization,with some adopting an“AI-first”approach for all new initiatives.O

246、rganize an“AI factory”to deploy AI confidence at scale,promoting short cycles and cost control.Recommendations include organizing around an“AI factory”for iterative testing and deployment.Now your bank can change pace from the initial+AI approach,which means adding AI to your operations,to an AI+ope

247、rating model in which the AI platform is central to all development and business methods.Scale AI in your bank.0910After reading our extensive research,youve undoubtedly realized that banks must keep ahead of this rapidly evolving landscape.As our guiding actions outline,this requires continuous tho

248、ught and a consultative approach to harnessing generative AIand its imperative to get started now.43Shanker RamamurthyGlobal Managing Partner,Banking and Financial MarketsIBM C J.DuigenanGeneral Manager,Financial Services,Banking,Financial Markets,and InsuranceIBM Technology,Global I is the Global M

249、anaging Partner for Banking in IBM Consulting,with a particular focus on bank transformation,core banking,and payments.He is also the President of the IBM Industry Academy and a member of the IBM Acceleration Team.He is a well-known thought leader with five patents relating to addressing complexity,

250、has authored several white papers,and been ranked as one of the 50 most influential financial services consultants worldwide by Euromoney magazine.Paolo Sironi Global Research Leader,Banking and Financial MarketsIBM Institute for Business V is a Global Research Leader in Banking and Financial Market

251、s at the IBM Institute for Business Value.He is senior advisor for selected global accounts,assisting service teams in Board-level and C-level conversations about business model adaptation in platform economies.He is one of the most respected Fintech voices worldwide and co-hosts the European editio

252、n of the Breaking Banks podcast.He is a celebrated book author on digital transformation and quantitative finance and economics,and a keynote speaker at major international events.John is the General Manager for the Financial Services sector within IBM Global Industries.His mission is to align the v

253、alue creation from IBM Technology with current industryimperatives that span clients across the sector.John and his team partner closely with clients to show how industry-specific use cases map to IBM technology capabilities,solutions,and outcomes.Use cases are a vehicle to demonstrate relevance,sca

254、le re-use,and create client references.As an IBM Distinguished Engineer,John also leads by example by being a hands-on technologist.Since joining IBM in 1998,he held technical and business roles spanning the entire product lifecycle,with extensive development,service,sales,and technology leadership.

255、He led multibillion-dollar transformation programs at IBMs customers.John is based in New York City.About the authorsContributorsMany thanks to the following for their contributions and insights:Sara Aboulhosn,Swati Bhide,Diane Connelly,Michael Conway,Angela Finley,Leah Generao,Vivek Kapur,Connor Lo

256、essl,Nicolas Meyerhoffer,Lucy Sieger,and Saket Sinha.44Domains and use casesPlease see below for domain categories and specific use cases.They are not exhaustive of all potential opportunities to leverage generative AI in banking.Client engagementVirtual customer services.Generative AI enhances the

257、client experience with virtual agents that reduce customer service call volumes.Digital assistant to customer services.Generative AI enhances classification of client complaints and automates content summarization to augment agent productivity and quality of live conversations.Virtual financial advi

258、sors(retail or corporate banking).Generative AI powers virtual financial advisors to digitize the advisory interaction and reduce human touch-points.Digital assistant to financial advisors(retail or corporate banking).Generative AI enhances search and summarization of financial information to improv

259、e the quality of advisory conversations with clients.Lending operations.Generative AI enhances approval processes to reduce time to serve customers.Trade finance.Generative AI enhances classification and summarization of counterparty agreements to speed processing time of trade finance operations.Ri

260、sk and securityTransaction monitoring.Generative AI empowers KYC/AML processes to reveal intricate webs of transactions and criminal networks.Risk modeling.Generative AI enriches the discovery of behavioral patterns to improve client and counterparty segmentation for risk management.Forecasting.Gene

261、rative AI integrates financial models with analysis of alternative data to enhance forecasting.Compliance reporting.Generative AI can ingest existing regulation and upcoming regulatory requirements to automate monitoring and compliance alerts.Security.Generative AI can scan logs,data,and software to

262、 find vulnerabilities to address cybersecurity proactively.IT developmentDevelopment lifecycle.Generative AI enhances software development to reduce the time and cost of building applications.Test and bug discovery.Generative AI creates synthetic data and tests scripts for efficient deployment of ne

263、w code.Core banking modernization.Generative AI enhances developers capability to convert software code,such as COBOL to Java,to reduce costs in modernizing monolithic core banking.Other support areasVirtual HR.Generative AI improves the self-service options of employees with better knowledge search

264、 and accuracy of classification,freeing up HR capacity.Digital assistant to HR recruitment services.Generative AI improves summarization and content creation to perform recruitment campaigns.Digital assistant to Marketing.Generative AI creates content,images,and personas to enhance marketing campaig

265、ns.Digital assistant to Procurement.Generative AI improves the summarization of vendor qualification and the analysis of legal documentation to streamline procurement.Digital assistant to Finance and Auditing.Generative AI improves the identification and reconciliation of gaps in financial data to s

266、treamline audit processes.45Research and methodologyIn collaboration with Oxford Economics,the IBM Institute for Business Value surveyed 600 bankers with executive responsibilities in the application of data and AI across their institutions.Their organizations each had more than$10 billion in total

267、assets in the latest fiscal year.As well,they operate in 17 countries:Australia,Brazil,Canada,China,France,Germany,Hong Kong,India,Italy,Japan,Mexico,Singapore,Spain,Sweden,UAE,UK,and US.This research is complemented by the financial performance analysis of worldwide banks with total assets larger t

268、han$50 billion.Annual reports are sourced from S&P Global Market Intelligence.Regional comparisons are provided based on the country classification in the International Monetary Fund(IMF)World Economic Outlook database.39 We distinguish between“major advanced and EU economies”(corresponding to G7 co

269、untries,with all EU member states)and“other advanced and emerging economies.”In addition,we conducted an internal survey of 110 IBM banking experts to assess a high-level mapping of risk use cases,keeping in mind the potential AI risk profile of a common financial institution,operating in a highly r

270、egulated environment,bound to adequate but finite investment resources.IBM Institute for Business ValueFor two decades,the IBM Institute for Business Value has served as the thought leadership think tank for IBM.What inspires us is producing research-backed,technology-informed strategic insights tha

271、t help leaders make smarter business decisions.From our unique position at the intersection of business,technology,and society,we survey,interview,and engage with thousands of executives,consumers,and experts each year,synthesizing their perspectives into credible,inspiring,and actionable insights.T

272、o stay connected and informed,sign up to receive IBVs email newsletter at can also find us on LinkedIn at https:/ibm.co/ibv-linkedin.The right partner for a changing worldAt IBM,we collaborate with our clients,bringing together business insight,advanced research,and technology to give them a distinc

273、t advantage in todays rapidly changing environment.About Expert InsightsExpert Insights represent the opinions of thought leaders on newsworthy business and related technology topics.They are based on conversations with leading subject-matter experts from around the globe.For more information,contac

274、t the IBM Institute for Business Value at .Related reportsEmbedded finance:Creating the everywhere,every day bankIBM Institute for Business Value in collaboration with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-financeEmbedded finance:The voice of the makersIBM Institute for Business Val

275、ue in collaboration with BIAN and Red Hat.October 2023.https:/ibm.co/embedded-finance-makersFoundations of banking excellence:Practices and priorities to accelerate digital transformationIBM Institute for Business Value in collaboration with BIAN.October 2022.https:/ibm.co/foundations-banking-excell

276、ence46Notes and sources 11 Ramamurthy,Shanker,John J.Duigenan,Hans Tesselaar,Hctor Arias,and Paolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Institute for Business Value in partnership with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-finance12 Ibid.13 IBM Institute

277、 for Business Value analysis of S&P Global data.Unpublished information.14 Ibid.15 Ramamurthy,Shanker,John J.Duigenan,Hans Tesselaar,Hctor Arias,and Paolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Institute for Business Value in partnership with BIAN and Red Hat.September 202

278、3.https:/ibm.co/embedded-finance16 Khandelwal,Manisha.The Role of NPS in Banking and Other Financial Institutions.Survey Sensum.May 19,2023.https:/ Ibid.18 Ramamurthy,Shanker,John J.Duigenan,Hans Tesselaar,Hctor Arias,and Paolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Instit

279、ute for Business Value in partnership with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-finance19 Based on internal IBM client information.20 Krishna,Arvind.“IBM CEO:Todays workforce should prepare to work hand in hand with A.I.”Fortune.April 20,2023.https:/ and IBM Collaborate on Generati

280、ve AI Initiative to Enhance Customer Experience.”IBM Newsroom.November 6,2023.https:/ IBM Institute for Business Value analysis of S&P Global data.Unpublished information.2“October 2023 euro area bank lending survey.”Press release.European Central Bank.October 24,2023.https:/www.ecb.europa.eu/press/

281、pr/date/2023/html/ecb.pr231024c42cea39db.en.html3“Household debt and credit report.”Center for Microeconomic Data.Federal Reserve Bank of New York.https:/www.newyorkfed.org/microeconomics/hhdc.html;“Credit Balances on the Rise as Consumers Manage Higher Costs.”TransUnion.November 9,2023.https:/ IBM

282、Institute for Business Value analysis of S&P Global data.Unpublished information.5 Ibid.6 Ibid.7“Mario Draghi,President of the ECB,and Luis de Guindos,Vice-President of the ECB.Frankfurt am Main.”Introductory statement.Press conference.European Central Bank.September 12,2019.https:/www.ecb.europa.eu

283、/press/pressconf/2019/html/ecb.is190912658eb51d68.en.html8 CEO decision-making in the age of AI:Act with intention.Global C-suite Series:28th Edition.IBM Institute for Business Value.June 2023.Unpublished information based on banking and financial markets data.https:/ Ramamurthy,Shanker,John J.Duige

284、nan,Hans Tesselaar,Hctor Arias,and Paolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Institute for Business Value in partnership with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-finance10 Walden,Stephanie and Mitch Strohm.“What Is A Neobank?”Forbes.June 24,2021.https

285、:/ Based on IBM client experience.23“Gartner Forecasts Worldwide Banking and Investment Services IT Spending to Reach$652 Billion in 2023.”Gartner Newsroom.June 21,2023.https:/ Forecasts Worldwide IT Spending to Grow 4.3%in 2023.”Gartner Newsroom.July 19,2023.https:/ IBM Institute for Business Value

286、 analysis of S&P Global data.Unpublished information.25 De Haan,Jakob.“Low IT spending by banks:Reason for concern?”ECON Committee.European Parliament.June 2021.https:/www.europarl.europa.eu/RegData/etudes/IDAN/2021/689439/IPOL_IDA(2021)689439_EN.pdf26 IBM Institute for Business Value analysis of S&

287、P Global data.Unpublished information.27 Keller,Christian,Mark Cus Babic,Akash Utsav,Ana Paula de Jesus Assis,and Brian Goehring.AI revolution:productivity boom and beyond.Barclays Research in partnership the IBM Institute for Business Value.January 11,2024.https:/www.cib.barclays/our-insights/AI-pr

288、oductivity-boom.html28 Sironi,Paolo,Diane Connelly,and Connor Loessl.Embedded Finance:The Voice of the Makers.IBM Institute for Business Value in partnership with BIAN and Red Hat.October 2023.https:/ibm.co/embedded-finance-makers29 Ramamurthy,Shanker,John J.Duigenan,Hans Tesselaar,Hctor Arias,and P

289、aolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Institute for Business Value in partnership with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-finance30“AI code-generation software:What it is and how it works.”IBM.September 19,2023.https:/ code-generation software:Wha

290、t it is and how it works.”IBM.September 19,2023.https:/ Goldstein,Jill,Bill Lobig,Cathy Fillare,and Christopher Nowak.Augmented work for an automated,AI-driven world:Boost performance with human-machine partnerships.IBM Institute for Business Value.August 2023.https:/ibm.co/augmented-workforce33 Ibi

291、d.34 Ramamurthy,Shanker,John J.Duigenan,Hans Tesselaar,Hctor Arias,and Paolo Sironi.Embedded finance:Creating the everywhere,everyday bank.IBM Institute for Business Value in partnership with BIAN and Red Hat.September 2023.https:/ibm.co/embedded-finance35 Generative AI:Impact on hybrid cloud pulse

292、survey.2023.IBM internal information.36 Ibid.37 Moore,Samuel K.“The Semiconductor Industrys Most Important Tool Goes Green.”IEEE Spectrum.June 12,2023.https:/spectrum.ieee.org/euv-lithography38 Based on internal IBM information.39“Groups and Aggregates Information.”World Economic Outlook Database.In

293、ternational Monetary Fund.April 2023.https:/www.imf.org/en/%20Publications/WEO/weo-database/2023/April/%20groups-and-aggregates4849 Copyright IBM Corporation 2024IBM Corporation New Orchard Road Armonk,NY 10504Produced in the United States of America|January 2024Java and all Java-based trademarks an

294、d logos are trademarks or registered trademarks of Oracle and/or its affiliates.The registered trademark Linux is used pursuant to a sublicense from the Linux Foundation,the exclusive licensee of Linus Torvalds,owner of the mark on a world wide basis.IBM,the IBM logo,Watson,and IBM z/OS mainframe ar

295、e trademarks of International Business Machines Corp.,registered in many jurisdictions worldwide.Other product and service names might be trademarks of IBM or other companies.A current list of IBM trademarks is available on the web at“Copyright and trademark information”at: document is current as of

296、 the initial date of publication and may be changed by IBM at any time.Not all offerings are available in every country in which IBM operates.THE INFORMATION IN THIS DOCUMENT IS PROVIDED“AS IS”WITHOUT ANY WARRANTY,EXPRESS OR IMPLIED,INCLUDING WITHOUT ANY WARRANTIES OF MERCHANTABILITY,FITNESS FOR A P

297、ARTICULAR PURPOSE AND ANY WARRANTY OR CONDITION OF NON-INFRINGEMENT.IBM products are warranted according to the terms and conditions of the agreements under which they are provided.This report is intended for general guidance only.It is not intended to be a substitute for detailed research or the ex

298、ercise of professional judgment.IBM shall not be responsible for any loss whatsoever sustained by any organization or person who relies on this publication.The data used in this report may be derived from third-party sources and IBM does not independently verify,validate or audit such data.The results from the use of such data are provided on an“as is”basis and IBM makes no representations or warranties,express or implied.1J0RA4OA-USEN-01

友情提示

1、下載報告失敗解決辦法
2、PDF文件下載后,可能會被瀏覽器默認打開,此種情況可以點擊瀏覽器菜單,保存網頁到桌面,就可以正常下載了。
3、本站不支持迅雷下載,請使用電腦自帶的IE瀏覽器,或者360瀏覽器、谷歌瀏覽器下載即可。
4、本站報告下載后的文檔和圖紙-無水印,預覽文檔經過壓縮,下載后原文更清晰。

本文(IBM:2024年全球銀行和金融市場展望報告(英文版)(52頁).pdf)為本站 (Yoomi) 主動上傳,三個皮匠報告文庫僅提供信息存儲空間,僅對用戶上傳內容的表現方式做保護處理,對上載內容本身不做任何修改或編輯。 若此文所含內容侵犯了您的版權或隱私,請立即通知三個皮匠報告文庫(點擊聯系客服),我們立即給予刪除!

溫馨提示:如果因為網速或其他原因下載失敗請重新下載,重復下載不扣分。
客服
商務合作
小程序
服務號
折疊
午夜网日韩中文字幕,日韩Av中文字幕久久,亚洲中文字幕在线一区二区,最新中文字幕在线视频网站