优化治理以加速生成式人工智能的部署.pptxVIP

OptimizinggovernancetoaccelerateGenAl

deployment

3Thebigpicture

6Keyfindings

7Section1:AreCSPsdeployingGenAIinproduction,atscale?

11Section2:DataandAImodelgovernanceareinextricablylinked

17Section3:ExplainableAIprovidescriticalguardrails

21Section4:AIcentersofexcellenceareasteppingstonetobecomingAInative

27Section5:HowODAisevolvingtooptimizeAIgovernance

33Additionalsponsorfeature:AccelerateL4autonomy:ScaleagenticAIsolutionsintoproduction

37Additionalresources

Contents

Thebigpicture

inform.tmforum.org3

“Ifyoulookatprojectsthatdonothavegovernance,

youwillseealotoffailures,becausetheywon’tget

pastlegal;theywon’tgetpastariskassessment;and

theywon’tgointoproduction,”saysEoinCoughlan,CTOandIndustryLeadforTelecommunications,MediaandEntertainmentatIBM.

Indeed,tworecentstudiesillustratehowdifficult

itistoscaleGenAIinproduction.AFebruary2025

reportcommissionedbyLenovoandauthoredbyIDCfoundthat88%ofgenerativeAI(GenAI)pilotsnever

reachproduction,whileamorerecentstudyfromMIT,publishedinJuly,highlightedthatastaggering95%ofGenAItoolsfailtodeliverexpectedreturnoninvestment(ROI).

“Andhereinthetelcoindustry,weknowthatoverhalfoftelcoAIinitiativesneverreallygetoffthestarting

blocks,”saidTMForumCEONikWillettsduringhiskeynoteatInnovateAmericasearlierthismonth.

“That’snotyetscale–thatisjustfrustration.”

Governanceisimperative

Telcoexecutivesareoptimisticthattheircompaniescanmovebeyondge