AI热潮之下2024年大型语言模型LLMs商业应用现状研究报告 英文版 .docxVIP

with

Behindthe

AIHype

The2024StateofLLMsinBusiness

Processes

I

Contents

01Overview02TopFindings

03Chapter1:What’sDrivingAIAdoption

05Chapter2:ActivatingLLMsinBusinessProcesses

10Chapter3:TheOutcomesofGenerativeAIinBusinessProcesses15Chapter4:GovernanceofAIinBusinessProcesses

18Chapter5:TheRoleofPeopleinaWorldofAutonomousAIandAgents23Chapter6:AIAgentsTheFutureofBusinessProcesses

28Conclusion

29MethodologyDemographics32AboutUserEvidence

Overview

MarketsentimentaboutgenerativeAIhasenteredanewera.Earlyhypeisfading.Businessleadershaveaclearerviewofitsstrengthsandchallenges.

Majorpress,suchastheWallStreetJournal,TheEconomist,andothershavewonderediftheAIhypeover-promisedandunder-delivered.Butveryfewhaveconsideredakeyvaluedriverforcompanies:

Businessprocessesandoperations.AslegendarybusinessthinkerTomDavenportwroteinHBR:“MostAIapplicationstodateseektoimproveagiventask.Butthisismissingthelargerpicture;smartcompaniesareviewingtheintroductionofAIastherationaleforanewlookatend-to-endprocesses.”

Severalfirms,suchasMcDonalds,AirCanada,andIBMhaveexperiencedhighprofilefailureswithearlygenerativeAIprojects.ConventionalwisdomsaysgenAIissuesaremodelchallenges:Waysthatlargelanguagemodels(LLMs)needimprovement.However,McKinseynotesthatitis“importanttorecognizethatthemodelitselfmakesuponly15percentofthesuccess[ofanAIproject].”ThereismuchmoretoanAIstrategythantrainingdata,modelweights,andpromptengineering.

TocreateconsistentbusinessvaluewithAI,companiesmustalsoaddressprocesschallenges.Theseinclude:HowgenAIisactivatedinbusinessprocesses.HowAIaccessesthelatestdata.HowAIrisksaregoverned,andhowhumanoversightaccountabilityismanaged.

WithindependentmarketresearchfromUserEvidence,wesetout