博思艾伦2025企业生成式AI应用构建指南从起步规划到落地实践英文版12页.pdf

WhitePaper2025

Building

Enterprise

GenerativeAI

Applications

AGuidetoGettingStartedfrom

BoozAllen’sGenAITeam

TableofContents

3Introduction

4TheGenAIStackArchitecture

5InfrastructureandPlatformLayers

6LLMLayer

6DataandPipelineLayer

7AgentandCapabilityLayer

7UI/ApplicationLayer

8GenAIStackPractices

8LLMOps:Monitoring,EvaluationandContinuous

Improvement

9AIGRC

11Conclusion

BuildingEnterpriseGenerativeAIApplications/CA-2506-126500Copyright©2025BoozAllenHamiltonInc.2

Introduction

Generativeartificialintelligence(GenAI)choosingandorchestratingLLMs

isreshapinghowenterprisesapproachbasedontaskcomplexityandcost,

knowledgemanagement,decisionandpreparinghigh-qualitydata

making,anduserinteraction.Whilepipelinestosupportreal-time

thistechnologyhasimmensepotential,anddomain-specificusecases.

deployingGenAIatscalewithinanWealsoemphasizetheimportance

enterpriserequiresmorethanjustofembeddinghumanoversightinto

modelaccess—itdemandsastrategic,AIworkflows,ratherthanrelying

layeredapproachthatalignswithsolelyonautonomousagents.

businessgoals,datainfrastructure,

andgovernancestandards.Beyondarchitecture,ourreport

outlinesessentialpracticesfor

Thisreportpresentsacomprehensivesuccess:implementingrobustLLM

frameworkforbuildingenterprise-operations(LLMOps)forcontinuous

gradeGenAIapplications,structuredmonitoringandimprovementand

aroundasix-layertechnologystackestablishingstronggovernance,risk,

architecture:Infrastructure,Platform,an