通過(guò)代理型人工智能重新定義工作、決策和創(chuàng)新Redefining+work,+decision-making,+and+innovation+with+agentic+AI

Redefiningwork,

decision-making,and

innovationwithagenticAI

?2025DellInc.oritssubsidiaries.AllRightsReserved.Dell,EMCandothertrademarksaretrademarksofDellInc.oritssubsidiaries.Othertrademarksmaybetrademarksoftheirrespectiveowners.

PrepareyourorganizationforthenextstepofAI

AgenticAIistheevolutionbeyond

generativeAIthatbringsautonomy,

adaptability,andgoal-oriented

intelligencetobusinessoperationsandthefutureofwork.Asorganizations

navigateanincreasinglycomplex

digitallandscape,theabilitytodeploy

AIthatactswithintent,learnsfrom

itsenvironment,andcontinuously

optimizesprocessesisagame-changer.

AgenticAIhasthepotentialto

superchargeeveryemployeeand

everyprocess,andtheimpacton

workstreamswilllikelybeprofound.

Withtheintroductionofpurpose-

drivenAIagents,thefutureofwork

willrequireemployeestoleverageandorchestrateaconstellationofdiverseAIAgentsacrossmultipleusecasesandworkstreams.AgenticAIcanaccelerateinformeddecision-making,automate

low-risktasksandprocesses,and

begintobridgethegapbetweenhumanexpertiseandmachineintelligenceinnewandcompellingways.

Withtherightstrategyandplanning,

organizationscanimplementagenticAIeffectivelywithoutintroducing

unnecessaryrisk.Bypartneringwith

DellandNVIDIA,organizationscan

accesstherightinfrastructure,data

transformationservices,andplatformstoharnessthebenefitsofagenticAI.

1.ForresterTrendsReport“AgenticAIIsRisingAndWillReforgeBusinessesThatEmbraceit.”March7,2025.

ForCEOs,agenticAIrepresentsmorethanjustanefficiencyplay—itsignalsafundamentalshiftinhowenterprisescompete,operate,anddelivervalue.1

2

AIagentsasyour

competitiveadvantage

WhatisagenticAI?

AgenticAIgivesorganizationsanopportunitytocombinemassivequantitiesoftheirvaluabledata,bestpractices,andprocedureswithautonomousintelligentsystems.UnliketraditionalAI,whichpassively

processesinputstoproducefixedoutputs,agenticAIfunctionsmorelikeanintelligentcollaboratorthatactivelyworkstowardsgoalswithinyourbusinesscontextratherthanjustfollowinginstructions.

Thiscollaborativecapabilityallowsorganizationsandtheiremployeestooperateatunprecedented

levelsofefficiencyandcreativitywhileensuringalignmentwithorganizationalgoalsandexpectations.

Byhavingthepowertoreason,perceivetheenvironment,learn,andadapt,agentscanbegivenagoal

andthenindependentlyperformcomplextasksandsolveproblemstoreachthatgoal,potentiallywithouthumaninteraction.

TheevolutionofgenerativeAI

Digitalassistant

Solvesspecific,task-

focusedquestions

withinanarrowdomain.

AgenticAI

Autonomouslymanagesawiderangeoftasksandadaptsasneedschange.

Chatbot

Engagesusersthrough

promptstoanswer

questionsorguideactions.

3

4

LevelsofAI

AgenticAIcanbestbeunderstoodbyviewingitwithinthebroadercontinuumofartificialintelligence.AsAIhasevolvedthroughdistinctstages,eachofthosestagesaddsnewcapabilities.Thisprogressioncanbeseenacrossthesecategories:

Higherlevelofconfidence/riskmitigationneeded

ArtificialGeneralIntelligence

Sciencefiction

(fornow)

Potentialvalue

01110

PhysicalAI

AgenticAI

GenerativeAI

TraditionalAI

Effortrequiredtobuild

TraditionalAI

Rule-basedsystemsthatrely

onpre-programmedlogicand

requirehumaninterventionfor

adjustments.Effectiveforrepetitive,structuredtasksbutlacksflexibility.

GenerativeAI

Usesdeeplearningmodelsto

createtext,images,code,andothercontentbasedonexistingdataanduser-drivenpromptinputs.

AgenticAI

AIagentscanreason,perceive

environments,learn,andadapt.

Whenprovidedwithagoaltheycansolvecomplexproblemswithlittletonohumaninteraction.

PhysicalAI

Embodiesintelligenceinrobotic

systemsthatinteractwiththe

physicalworld,suchaswithrobots,sensors,andotherequipment.

ArtificialGeneralIntelligence(AGI)

Representstheaspirationalgoal

ofAIdevelopers,researchers,andscience-fictionenthusiasts,i.e.

machineswithhuman-likecognitiveabilities.AGIremainstheoretical.

WhyAIlevelsmatter

Autonomy

Creativity

Goals

BusinessleadersmustunderstandthedistinctionsbetweenstagesofAItomake

strategicinvestmentsandassessthecomplexityofdifferentAIimplementationsorusecases.AgenticAI,inparticular,introducesanewlevelofcomplexityandsophistication,whichwillrequireadeeperalignmentwithbusinessprocesses,businesslogic,andtheabilitytoactautonomouslyacrossscenarios.

Inaddition,asagenticAIsolutionsareadoptedbyorganizationsalreadyinvestedintraditionalandgenerativeAI,it’simperativetoimplementrobustguardrails,humanoversight,andtightersecuritytoensuresafe,reliableoutcomes.

Goals,autonomy,andcreativityarenotjusttraitsofagenticAI,theycanbeactively

definedbyorganizations.SettingcleargoalsvalidatesthatAIisalignedwithbusinessprioritiesfromthestart.Definingboundariesforautonomyallowsthesystemtoact

independentlywhilestayingwithinlimits.Guidinghowagencyisappliedviaconstraints,context,orpreferredmethodsensuresthatsolutionsstillsupportstrategicintent.

5

5

Risksandrewards:Thehumanimpact

AgenticAIisreshapingtheworkplace,transforminghowtasksare

completed,decisionsaremade,andteamscollaborate.Thisshiftofferssignificantrewards,butitalsointroducesnewcomplexitiesthatmust

becarefullymanaged.Organizationsneedaclearstrategythatbalancesautomationwithoversight,ensuringsystemsarecalibratedforboth

performanceandaccountability.

AdvancingtoagenticAIisasteptowardamoreintelligent,adaptive,

andresiliententerprise.Butgettingtheremeansleadersmustdeterminewhatfunctionsarebestsuitedforautonomy,wherehumanjudgment

shouldremaincentral,andhowtoensuretransparencyandtrustinAI-drivendecisions.

Risks:Automationrequiresoversight

AgenticAIdeliverspowerfulcapabilities,butwithoutproperoversight,itcanintroducechallengesthatimpacttransparency,accountability,andjudgment.

Rewards:AIasaforcemultiplier

Whenstrategicallydeployed,agenticAIcanbecomeaforcemultiplier,amplifyingemployeepotential,acceleratingdecisions,andimprovingbusinessoutcomesusingdata-driveninsights.

6

Exploringcorecharacteristics

Coretraitsthatdefineagents

AgenticAIdeliverstransformativebusinessvaluebycombiningautonomy,adaptability,and

goal-drivenintelligence,enablingITsystemstoproactivelydriveoutcomes,optimizeprocesses,andlearninrealtime.ThefollowingcorecharacteristicsofagenticAIcanhelpempower

organizationstooperatewithgreaterspeed,precision,andscalability,creatingarepeatable,sustainablecompetitiveedgeinanAI-driveneconomy.

ThesecoretraitsenableagenticAItoprovidescalable,intelligentsystemscapableof

drivingefficiency,optimizingdecision-making,andreducingfrictionacrossworkstreams.Organizationsthatleveragethesecapabilitiescanachievegreateragility,innovation,andresilienceinanincreasinglyAI-driveneconomy.

Perceiveenvironmentsthroughdatainputs,

sensors,andmodelstoinforminteractions.

Learnandadaptby

evolvingbehavior

throughongoinganalysisandexperience.

Operateautonomouslytoexecutetaskswithouthumanintervention.

Interfacewithsystemstotakemeaningful

actionsacrossdigitalenvironments.

Pursuedefined

objectiveswithout

requiringdetailed

executioninstructions.

Analyze,reason,and

actbasedoncontextualunderstandingand

availabledata.

7

8

8

WhyagentsarethenextstageofAI

Theimpactsofautonomousoperations

UnlikeAImodelsthatrequirefrequenthumanintervention,agenticAIoperatesindependentlybycontinuouslyrefiningitsprocessesbasedonreal-timedataandorganizationalobjectives.BelowarekeycapabilitiesthatdefinehowagenticAIenhancesbusinessoperations.

Passivemonitoring

AgenticAIcanscandigitalandphysicalenvironmentstoidentifyrelevantdata

pointsoranomaliesandintelligentlyreactbasedondefinedrulesandprocessesanddecision-making

Activequeryresponse

WhiletraditionalAIsystemsoften

requirestructuredinput,agenticAI

candynamicallyrespondtocomplex,

unstructuredqueriestointerpretcontextandretrieverelevantinformation.

Event-driventriggers

AgenticAIcanexecutepredefinedactionswhenspecificconditionsaremet,reducingresponsetimesandleadingtotimely

actionsincrucialscenarios.

Scheduledtasks

AgenticAIcanoptimizeefficiencyby

intelligentlymanagingtime-sensitive

workflowsandadjustingtimelines.Thishelpsenhancecompliancereporting,predictive

maintenance,andinventoryrestocking.

Organizationalworkflowintegration

ForagenticAItodeliveritsfullpotential,itmustbeseamlesslyintegratedintoexisting

workflows,datasources,platforms,orinfrastructure.Ratherthanfunctioningasa

standalonesystem,agenticAIenhancestechnologyecosystemsbyaugmentingemployeecapabilities,improvingdecision-making,andproactivelyaddressinginefficiencies.

WhenconsideringhowtodeployAIagents,it’simportanttodeterminehowspecificusecasesimpacttheunderlyingITarchitectureandplatforms.Forexample,anagenticAIsystemthatispurelydigitalwilldependondifferentinfrastructurethanagenticAIwhichrequiresphysicalcomponents.

Examplesinclude:

Informationsystems

Businesssystems

AgenticAIpersonalizescustomer

engagement,streamlinesworkforceplanning,andautomatessupplychainoperationsby

enablingCRM,HR,andERPsystemstoadaptandactautonomously.

IToperations

AIagentscanproactivelymonitorITsystems,identifythreats,andautomateresponses,

ultimatelyhelpingmaintainsecurity,meetSLAs,andreducedowntime.

Operationalsystems

In-storeorreal-timecustomerinteraction

Monitorandrespondtoin-storecustomerinteractionsinrealtimeusingsentiment

analysisandcomputervisiontopersonalizecustomeroffers.

Factory,warehouse,logistics

CombinecomputervisionwithIoTdevicesandagenticAItomanagecomplexphysicaloperationssuchasautonomousvisual

equipmentinspection.

9

Governanceandriskmitigation

MaintainingdataprivacyandcomplianceisacornerstoneofresponsibleAIadoption.

Organizationsmustimplementgovernanceandmulti-layeredsecuritymeasurestosafeguardsensitiveinformationwhileenablingAIsystemstooperateeffectively.Strategiesinclude:

Controlledaccess

RestrictAIaccesstoonlynecessarydatasets,ensuringthatinformationremainsprotectedandconfidential.

Encryption

Utilizeadvancedencryptionmethodstosecuredataatrestandintransit,reducingexposuretocyberthreats.

TransparencyinAIdecision-making

DocumenthowAImodelsmakedecisions,providingattributiontostakeholdersandregulatorybodies.

Defineandenforceguardrails

Limitscopeoffunctionality,definepurposeclearly,andprovideanarrowoperatingwindowforagents.

Riskmonitoringandmeasurement

ContinuouslymonitorAIsystemsforvulnerabilitiesandperformancedeviations,usingdefinedmetricstoassessriskandinformtimelyresponseactions.

Byembeddingstronggovernance,security,andprivacymeasuresintoagenticAIdeployment,organizationscanconfidentlyharnessAI’spotentialwhilemaintainingethicalintegrityand

regulatorycompliance.

10

11

11

AgenticAIusecaseexamples

AgenticAItransformsindustriesbyautomatingdecision-making,optimizingworkflows,and

drivingefficiency.WhileagenticAIapplicationsmayvaryacrosssectors,thefollowingshort-listofusecasesillustrateAI’sbroadimpact.

CustomerService

AgenticAIcantransformcustomerservicebydeploying

agentsthatautonomouslyresolveinquiries,personalize

interactions,andanalyzesentimentinrealtime.Integrated

withCRMsystems,thisreducescustomerwaittimes,

improvessatisfaction,andhandlesroutinecustomersupport.

Operationsandlogistics

AgenticAIcanoptimizetheentiresupplychainbypredictingdemand,managinginventorylevels,andcoordinatinglogisticswithprecision.Itreactstoreal-timeconditions—rerouting

shipments,adjustingschedules,andlearningfromeverydatapointtoimproveefficiencyovertime.

Cybersecurity

AgenticAIcanstrengthencybersecuritybyautonomouslymonitoringITenvironments,detectinganomalies,and

executingresponsesbasedonpre-establishedsecurity

protocols.Thisproactiveapproachreducesrisk,shortensincidentresponsetimes,andensurescontinuousprotectionatscale.

Finance

Infinance,agenticAIcanidentifysuspicioustransactions,analyzecomplexriskpatterns,andpreventfraudinreal

time.Thisenhancescompliance,reducesmanualoversight,andsupportsmoreaccuratefinancialdecision-makingwithcontinuouslyupdatedintelligence.

Smartcitiesanddigitaltwins

AgenticAIcanenablesmarterurbaninfrastructureby

analyzingreal-timesensordata,optimizingtrafficflow,andpredictingmaintenanceneeds.Whenintegratedintodigitaltwinenvironments,itcansimulatecityoperations,enablingproactiveurbanplanningandcrisispreparedness.

Continuousforecastingandplanning

AgenticAIcanfacilitatedynamicforecastingbyconstantlyrefiningpredictionsbasedonshiftinginternalandexternaldata.Itcanauto-adjustplans,detectemergingtrends,

andsupportmoreaccurate,agiledecision-makingacrossbusinessfunctions.

PrioritizingagenticAIusecases

SuccessfulimplementationrequiresastructuredapproachtodeterminewhereAI-drivenautomation

deliversthemostsignificantvalue.Generally,businessleadersshouldevaluatepotentialAIusecase

feasibilitybasedontheeffort,cost,andcomplexityofdeployingagenticAI.Thisincludesfactoringinsystemintegration,datapreparation,modeltraining,andongoingmaintenance.Prioritizesolutionsthatfitexistingworkflowswithminimaldisruption.

TheLEARNSacronymcanbeusedtoidentifytherighttypesofagenticAItasksandactivitiesthat

delivervalue.Itcanhelpdecision-makersunderstandthebusinessvalueofAIsolutionsbyviewingusecasesthroughthelensofthefollowingcriteria:

Lowrisk

TasksthathaveminimalnegativeimpactifautomatedorifAImakesamistake.

Emerging

AreaswhereAI-drivenautomationstillevolvesandshowsstrongpotential.

Arduous

Repetitive,time-consumingprocessesthatbenefitfromautomation.

Remedial

Error-pronetasksthatAIcanoptimizewithenhancedaccuracy.

Notworthit

FunctionswhereAIaddsvaluebyenabling

employeestoperformother,morevaluablework.

Speed

ProcesseswhereAIcanaccomplishthetaskfasterthanemployees.

Byapplyingthisapproach,organizationscanstrategicallyprioritizewhichactivitiesarebestsuitedforagenticAI,establishingabalancebetweenautomationandhumanoversight.

12

12

13

HowtobuildagenticAI

Corecomponents

BuildingarobustagenticAIsystemrequiresacombinationoffoundationalAItechnologies,scalableinfrastructure,andstrategicimplementation.BelowarethekeycomponentsthatenableagenticAItofunctioneffectively:

Dataaccess

ForagenticAItooperateeffectively,itmusthave

seamlessaccesstotimely,relevant,andcontext-richdataalongsideestablishedbusinessprocessesandlogic.ThisaccessenablesAIagentstomakeinformeddecisions,

adapttochangingenvironments,andcontinuouslyoptimizeperformancebasedonthemostcurrentinformationavailable.

Infrastructure

DeployingagenticAIatscalerequiresarobust,scalableinfrastructureandintegrationwithexistingtools

thatsupportbothcloud-basedandedgecomputing

environments.Someusecasescanrelyonthecloud

whereasedgecomputingenablesreal-timeprocessingatthepointofaction,reducinglatency.

LargeLanguageModels(LLMs)

LLMsarecoretoagenticAI,providingtheabilityto

process,understand,andgeneratehuman-liketext.ThesemodelsenableAIsystemstointerpretcomplexqueries,synthesizeinformation,anddelivercontextualinsights

tailoredtobusinessneeds.

Multi-agentsystems

RatherthanrelyingonasingleAImodel,multi-agentsystemsinvolvemultipletask-specializedAIentitiesworkingtogethertoachievecomplexobjectives.

Investmentsindataaccess,infrastructure,andLLMsinyourinitialagenticAIdeploymentswilllaythe

foundationformulti-agentsystemsinthefuture.

Crawl,walk,runapproach

ImplementingAIandgenerativeAIsuccessfullyrequiresaphasedapproachtodrivealignmentwithbusinessobjectivesandriskmanagement.OrganizationscanadoptaCrawl,Walk,RunmethodologytoscaleAIcapabilitiesgradually.

Crawl

Focusonsmall-scaletraditionalAIandgenerativeAIprojectsfocusedonlow-risk,high-valueusecases.

IdentifyareaswhereAI-drivenautomationcandeliverquickwinstohelpyouvalidateAIeffectivenessbeforecommittingtobroaderdeployment.

Walk

Beginpilotingsmall-scaleagenticprojects,integratingAI-drivendecision-makingintocriticalworkflows,governanceframeworks,andmodels.

Run

Deployandintegrateagenticacrossyourorganizationasacorecomponentofbusinessprocessestoaddressyourmostcriticalneeds.

14

15

DellandNVIDIA:YourtrustedAIinnovationpartners

TheDellAIFactorywithNVIDIA

AgenticAIdemandsmodernITandscalableinfrastructuretosupportautomationatscale.TospeedAIadoption,organizationscanleveragetheDellAIFactorywithNVIDIA,anend-to-endframeworkthatincorporatesservices,AIsoftware,andinfrastructure.

TheDellAIFactorywithNVIDIAhelpsaccelerateAIadoptionwithpre-builtarchitectures,high-performancesystems,andAI-optimized,integratedsoftwarestacksthatenable

efficient,scalabledeploymentofagenticAImodels.

SERVICES

DATA

USECASES

StrategyData

Content&codegeneration

Platform