生成式AI在信贷领域的新兴应用案例

EmergingGenerative AI UseCases in Credit Research resultsMarch2025 Context and overview of the research IACPM and McKinsey jointly conducted a research in 2024 to understand key emerging GenAl use cases in Credit and its challenges, risks,investments and returns. This research was conducted using a combination of interviews camried out in Q3 2024 and a flash survey conductedin Dec 2024. 33institutions participated in the interviews and 33 institutions participated in the online survey, representing a total of 44 unique institutions. These financial institutions encompassed Mega Banks, Super Regionals, Core Regionals, and others across North America, Europe, Asia,and Africa In this webinar, we will discuss the findings from our jointly conducted research, including both interview and flash survey results We will also compare the findings with an article McKinsey published inJ uly 2024 based on insights from roundtable conducted in Oct 20231with CROs of 24 North American financial institutions to understand the state of adoption for GenAl use cases within a year of the launch ofChatGPT. CROs were enthusiastic about GenAl use cases: We heard from 80% of the CROs in North America that they were expecting toimplement the first GenAl application by Q4 2024 Since then, LLM models have evolved rapidly and so have the use cases and factors defining success Expectation of CROs from NA in Oct 2023 Executive summary All participating institutions are testing one or more GenAl use cases in Credit, with super regionals leading the deployment. Whileinstitutions are experimenting with a wide range of use cases, most prevalent uses cases are synthesizing for credit decisioning,drafting memos, data assessment and early warning. The progress, however, is slower than what executives expected in Q4 2023 -where 80% of NA Fls expected to implement GenAl use case within a year, while only ~30% of survey respondents from NA havereached the deployment stage 8o% institutions have access to external GenAl tools, though some institutions take caution in making it available to all employees. Toensure safe use, more than 5o% of the institutions require understanding of internal guidelines as a prerequisite to access them Institutions are taking fast actions to address the challenges they anticipated in Q4 2023: "Sixty-seven percent of the participantshighlighted potential shortages of gen Al capabilities inside the organization. Further challenges, cited by around 50 percent ofparticipants, include difficulties defining uses cases and value at stake." To address this, institutions have prioritized investment in building talent (87%), Center of Excellence (CoE) (82%), secureenvironment (94%) and processes (79%). Focus on building CoE is consistent with the research McKinsey conducted earlier in 2024where participants preferred a centralized GenAl operating model With large number of use cases emerging, institutions are considering several factors to determine prioritization with productivityimprovement (47% ranked as top factor) and business need (44% ranked as 2nd top factor) being the top drivers while returmn oninvestment being the least important metric Other important factors that are playing a role in an institution's adoption of GenAl are leadership commitment to prioritize GenAl Institutions are continuing to see challenges they observed in Q4 2023 related to defining uses cases and value at stake. In this survey,we heard institutions that top reasons for abandoning use cases are insufficient performance (41%) and articulation of benefit(41%) 33 institutions participated in the survey with global representationNumber of institutions who responded to the online survey 1. Mega Banks include institutiors with >$1,000Bn in assets; Super Regionals include institutions with $500-$1,000Bn in assets;CoreRegionalsincludeinstitutionswith$1o0-5ooBninassets;Othersincludeinsurancecompanies/brokersanddevelopmentbanks Most institutions are testing GenAI use cases, with super regionalsleading the deployment Question: Which GenAl use cases is your institution currently implementing in commercial credit and what are their development stage?[multiple choice] Institutions are prioritizing use cases like supporting underwriting (synthesizing,drafting memo) and portfolio monitoring (early warning)Mega BanksCore regionals Super RegionalsOtherQuestion: Which GenAl use cases is your institution currenty implementing in commercial credit and what aretheir development stage? [multiple choice] Survey results on use case prioritization aligns with CRO expectation in October2023 A: Case example of how a bank automatedcredit memo drafting using GenAl CA productivity gain ~30%Faster decisioning speed A video will play during the Webinar The use cases vary based on institutional priorities Examples sourced from the 2024 interviews Most institutions have access to external GenAl tools and requireunderstanding of internal guidelines a