Mcsey

&company

People&OrganizationalPerformancePractice

Theagenticorganization:

Contoursofthenext

paradigmfortheAIera

Companiesaremovingtowardanewparadigmofhumansworkingtogether

withvirtualandphysicalAIagentstocreatevalue.Wesharelessonsfrom

earlyadopters—andwhatyoucandonext.

ThisarticleisacollaborativeeffortbyAlexanderSukharevsky,AlexisKrivkovich,ArneGast,ArsenStorozhev,

DanaMaor,DeepakMahadevan,LariH?m?l?inen,andSandraDurth,representingviewsfromMcKinsey’s

People&OrganizationalPerformancePractice,McKinseyTechnology,andQuantumBlack,AIbyMcKinsey.

September2025

AIisbringingthelargestorganizationalparadigmshiftsincetheindustrialanddigitalrevolutions(seesidebar,“Theevolutionofoperatingmodels”).Thisnewparadigmuniteshumansand

AI

agents

—bothvirtualandphysical—toworksidebysideatscaleatnear-zeromarginalcost.Wecallittheagenticorganization.

McKinsey’sexperienceworkingwithearlyadoptersindicatesthatAIagentscanunlock

significantvalue

.Organizationsarebeginningtodeployvirtual

AIagents

alongaspectrumof

increasingcomplexity:fromsimpletoolsthataugmentexistingactivitiestoend-to-endworkflowautomationtoentire“AI-first”agenticsystems.Inparallel,physicalAIagentsareemerging.

Companiesaremakingstridesindeveloping“bodies”forAI,suchassmartdevices,drones,self-drivingvehicles,andearlyattemptsat

humanoidrobots

.ThesemachinesallowAItointerface

withthephysicalworld.

Theevolutionofoperatingmodels

Intheagriculturalerapriortothe

1800s,operatingmodelsweresimpleandcenteredaroundsmallteamsof

craftspeopleandfarmers.Eightyto

90percentoftheglobalpopulationworkedinagriculture.1

Next,intheindustrialera,peoplemoved

intofactories,andoperatingmodels

shiftedtofunctionalhierarchies.Productsweredesignedformassreplication

bypeopleandmachines,withmajor

upgradeseverythreetotenyears.New

rolesemerged,suchasfactoryworkers,engineers,andshiftsupervisors.By

the1970s,39percentofpeopleinthe

UnitedStatesworkedintheindustrial

sector,withjust4percentinagriculture.2Efficientscalingdrovecompanies’growthandcompetitiveadvantage,and

lean

management

becameastrategictool.

Asthedigitaleralaunchedinthe1990s,industrial-agemaximsbegancrumblingwiththeriseofcomputing.EarlyITefforts

mirroredindustrialthinking,hard-codingbusinessprocessesintomonolithic

systemssupportingproductionand

enterpriseresourceplanning

.Companiessoonshiftedtomodulardigital

products

andplatforms

,updatedmonthlyorevendaily.3Speedrequired

agileoperating

models

withsmall,cross-functionalteams,includingnewrolessuchassoftware

engineers,experiencedesigners,and

productmanagers.Speedandcustomer

accessbecamethekeystocompanies’

competitiveadvantage.Today,5.8percentoftheUSpopulationisemployedintech

jobs,4withonly1.6percentinagriculture5and19.3percentintheindustrialsector.6

Now,theAIeraisbeginningtousherinthenewestevolution,revolutionizing

knowledgeworklikethepreviouserasrevolutionizedphysicalwork,withthe

agenticorganizationbringingtogetherhumans,AIagents,andmachinesintheworkplaceofthefuture.

Thepromiseofthisnewparadigmwill

dependuponthecontinuedgrowthof

AI’scapabilities,aswellasotherfactors

suchasregulation.Thelengthoftasks

thatAIcanreliablycompletedoubled

approximatelyeverysevenmonthssince2019andeveryfourmonthssince2024,reachingroughlytwohoursasofthis

writing.7AIsystemscouldpotentially

completefourdaysofworkwithout

supervisionby2027.Thiswouldbea

phenomenallyacceleratedevolution—

fromanintern-levelemployeerequiring

constantsupervisiontoamid-tenure

employeewhocanoperateindependentlyto,perhaps,aseniorexecutivewhocan

shapeanddrivestrategies.

Organizationalparadigmsdocoexist.Buttheagenticorganizationmayofferthe

keyfortheleaderstogainacompetitiveadvantagebybuildingdecentralized

outcomes-focusedagenticnetworks.

1MarianL.TupyandRonaldBailey,“Thechangingnatureofwork,”HumanProgress,March1,2023.

2VictorR.Fuchs,Economicgrowthandtheriseofserviceemployment,NationalBureauofEconomicResearchworkingpaper,number486,June1980.

3EikiTakeuchi,“WhyAmazonreleasesevery11.6seconds,”Medium,May11,2025.

4Stateofthetechworkforce2025,CompTIA,July1,2025.

5Data360Database,“Employmentbysector(%),”WorldBank,accessedAugust2025.

6Data360Database,“Employmentbysector(%),”WorldBank,accessedAugust2025.

7“MeasuringAIabilitytocompletelongtasks,”METR,accessedSeptember2025.

Theagenticorganization:ContoursofthenextparadigmfortheAIera2

Theagenticorganization:ContoursofthenextparadigmfortheAIera3

Theagenticorganizationwillbebuiltaroundfivepillarsoftheenterprise:businessmodel;

operatingmodel;governance;workforce,people,andculture;andtechnologyanddata

(Exhibit1).Imagine,forinstance,thebankoftomorrow:Whenacustomerwantstobuyahouse,

apersonalAIconciergeactivatesaseriesofagenticworkflowstoservethebuyer.Arealestate

AIagentsuggestsproperties,whileamortgageunderwritingagenttailorsoffersbasedonthe

customer’sfinancialprofile.Complianceagentsensurethatthedealadherestobankpolicies,

andacontractingagentfinalizesagreementsbeforeanotheragentfulfillstheloan.Alltheseworkflowsareoverseenbyanagenticteamofhumansupervisors,mortgageexperts,and

AI-empoweredfrontlineemployees.Insomecases,thebankcouldevenextenditsAI-powered

servicesintofurnishing,renovations,energyupgrades,andmore.Thebankbecomesanetwork

ofagenticteams—anagenticorganization.

Exhibit1

AIisleadingthelargestorganizationalparadigmshiftsincetheIndustrialandDigitalRevolutions.Coreelementsoforganizationalparadigms,byera

Dominantperiod

Era

III

18002000

Craftandagriculture

III

18002000

Industrial

1800200018002000

Digital

AI

Businessmodel

Agriculturalandartisanprod-uctsviadirectchannels,eg,localbread,tailoredclothes

Centuriestochangestandarddesigns

Humancreationanddelivery

Manufacturedgoodsviaphysicalchannels,eg,cars,washing

machines,mass-marketsoap

3–10yearsbetweenmajorupgrades

Linear,repeatablebusinessprocesses

Digitalchannelsandproducts,eg,e-commerceplatforms,

bankingapps,socialmedia,softwareasaservice(SaaS)

Dailyormonthlyproductreleases

Digitaljourneyswithanalytics

AI-nativechannelsandproducts,eg,personalconcierges

Real-timepersonalizationandinnovation

AI-?rstworklowsfueledbyproprietarymultimodaldata

Operatingmodel

Teamsoffarmworkersor

skilledindividuals(artisans),withknowledgetransferfrommastertoapprentice

Functionalhierarchieswithlargefrontlineforrepetitivetasks,

smallwhite-collarteamsfor

managementandengineering

Cross-functionalteamsofknowl-edgeworkersalignedtoprod-

ucts,projects,andsegments,withdigitallyenabledfrontline

Flatnetworksofhybridagentic

teamsstructuredtodriveend-to-endoutcomes

Governance

Localplanninganddirectgovernance

Rigidplans,waterfalldelivery,andmanualgovernance

Iterativeproductdelivery,

quarterlyrealignment,andagilegovernance

Real-time,embeddedgovernanceandagenticcontrolswithhumanaccountability

Workforce,

people,andculture

Deepspecializationandcul-tureofcraftsmanship

Narrowlyspecializedfunctionaltalentworkinginacultureof

planning

KnowledgeworkerswithT-shapedtalentpro?lesworkinginacultureof

experimentation

HybridworkforcewithT-shapedandM-shapedhumantalentpro?les

Cultureofcontinuouschangeandlearning

Technologyanddata

Handtoolsandanimalstohelphumans

Handwrittennotebooksandledgers

Machinesandharnessedenergy

IT“monoliths”(eg,enterprise

resourceplanning,mainframes)ownedby(outsourced)ITdepart-mentswithmanualsoftware

delivery

Gigabytesofstructuredopera-tionaland?nancialdataindatawarehouses

PC,mobile,cloud,industrialrobots,etc

Modularsystems,(micro-)

servicesandAPIsownedby(in-house)cross-functionalteamswithsemiautomateddelivery

Tera/petabytesofsemistruc-

tureddataforadvancedanalyticsindatalakes

Sensors,humanoidrobots,drones,etc

DemocratizedAImeshwithmodularAIagents,agent-to-agentcommuni-cation,anddynamicsourcing

Peta/exabytesofunstructuredmultimodaltacitdata

Iconic

examples

Bread,artisanclothesandshoes,art

Ford,GE,Toyota

Google,Spotify,Facebook

Leadershipintheerastillopen

McKinsey&Company

Theagenticorganization:ContoursofthenextparadigmfortheAIera4

Inthisarticle,weshareearlysignalsfromourworkwithpioneeringcompanies,insightsfrom

techleadersandinvestors,andthequestionsexecutivesareaskingus.Theagenticorganizationparadigmwillundoubtedlyevolve,buttoday’sleaderscannotwaitforperfectclarity.Inthis

article,wepointleaderstowheretheycanactnowtoshapethenewera—refiningtheir

operating

modelstocreatemorevalue

and

rewiring

foranAI-firstapproach—insteadofwaitingtobe

shapedbyit.

Fivepillarsoftheagenticorganization

1.Businessmodel

Intheagenticera,companieswillgainacompetitiveadvantagebygettingclosertocustomersviaAIchannelstoofferreal-timehyperpersonalization,streamliningprocessestobecome

AI-first,andbuildingawalledgardenofproprietarydataastheirsuperpower.AI-nativestart-upsandagenticcompaniescanpotentiallydisruptindustries,withafundamentallydifferentlevelofproductivity(revenueperemployee),costdecoupledfromgrowth,andgreaterspeedtomarketandinnovation.

AI-nativechannelsenablehyperpersonalization

Consumersarealreadybypassingshops,apps,andsearchenginesinfavorofAI-native

interfacessuchasChatGPT.Inthefuture,everyconsumercouldhavea

low-costAIpersonal

assistant

.One

Europeanutilityprovider

hasrolledoutamultimodalAIassistanttoitsthree

millioncustomers.Itsignificantlyreducedaveragehandlingtimes,boostedcustomer

satisfaction,improvedresponsespeed,andresolvedmorecallswithoutahuman.These

assistantswon’tjustrespond;theyarepersonalconciergesthatwillnegotiatewithotheragents24/7,continuouslylearningfromuserbehaviorandmarketsignalstogenerateever-evolving,

hyperpersonalizedproducts.Thisalsounlocksnewopportunitiesforthe

ecosystemeconomy

,inwhichcompaniesthatowncustomercontactcangrowby

meetingvariouscustomerneeds

beyondtheirtraditionalbusinessmodelandindustryboundaries.

AI-firstworkflowsdrivemarginalcoststowardthecostofcompute

Banksalreadyrunmortgageandcomplianceprocesseswithagentsquads.

Insurers

are

reinventingclaimsandunderwriting,whilereimaginingthemselvesasAI-native.Telcosareusingagentsin

customerserviceandbeyond

.Oneglobalbank’s“agentfactory”manages

know-your-

customerprocesses

withtenagentsquads,whichhashelpedachieveasubstantialpositive

impactonthequalityandconsistencyofoutput.

Anotherbank

hasusedhumanstooversee

squadsofAIagentsinmodernizingitslegacycoresystems,enablingupto50percentreductionsintimeandeffort.Thisisnotautomationasusualontopofexistingprocesses—it’saredesign

ofend-to-endprocesseswithhumans“abovetheloop”forstrategicoversight,withpotentialto

bringthemarginalcosttowardthecostofcompute.Goingforward,most,ifnotall,processescanbereimaginedasAI-first,withhumansandtraditionalITsystemsselectivelyintroducedbackin

thelooporabovetheloop.

Proprietarydatabecomesakeydifferentiator

Iftoday’sAIis“aninternwiththeinternetinitspocket,”tomorrow’sedgewillcomefromthe

walleddatagardensthatthepublicinternetdoesn’toffer.Companiescanoutperformtheir

competitionbycontinuouslycapturingandrefiningunique,consented,proprietarydata—suchasstreamsofcustomerbehavior,productusage,andsensordata—andconvertingtheminto

differentiatingpersonalizedproductsandprocesses.AIcanalsohelpbyacceleratingthe

build-

upofdatafoundationsanddataproducts

,aswellasdata-qualityimprovements.

Theagenticorganization:ContoursofthenextparadigmfortheAIera5

2.Operatingmodel

Intheagenticera,howorganizationsarebuiltandoperatewillevolveasmuchastheproductsorservicestheydeliver.WorkandworkflowswillbereimaginedasAI-first,andoperatingmodels

willevolvetoflatnetworksofempowered,outcome-alignedagenticteams.

WorkandworkflowswillbereimaginedasAI-first

TheoperatingmodeloftheagenticerawillbeanchoredaroundreimaginedAI-firstworkflows,withhumansandITsystemsselectivelyreintroducedinAI-nativedesign.AtaEuropean

automakerandapublicsectororganization,squadsofagentsarereverse-engineeringand

modernizinglegacysystemswhilehumanssteerandvalidatework.Inproductdevelopment,

agentscangatherfeedback,analyzedata,testfeatures,andevenruncampaigns.Humanswillbemostlypositionedabovethelooptosteeranddirectoutcomesandselectivelywithintheloopwherehumancontactmatters.

Outcome-alignedagenticteamswillbeorganizationalbuildingblocks

Traditionalorganizationshavebeenbuiltaroundfunctionalsilos.Digitalcompanieshavecross-functionalproductteamsbutarestillconstrainedbyhandoversandhumanteamsizelimitations,suchasthetwo-pizzateam1andDunbar’snumber.2

Intheagenticorganization,structurewillpivottosmall,outcome-focusedagenticteams.

Anagenticteam—asmallergroupofmultidisciplinaryhumanswhoownandsupervisethe

underlyingAIworkflows—canbeorganizedtodeliverclearend-to-endbusinessoutcomes

coveringthefullfunctionalvaluechainofmarketing,productmanagement,technology,data,andoperations.Inourexperience,ahumanteamoftwotofivepeoplecanalreadysuperviseanagentfactoryof50to100specializedagentsrunninganend-to-endprocesssuchasonboarding

acustomer,launchingaproduct,orclosingthebooks.AgenticAIcanextendthescopeandautonomyofaproductteammorethanever.

Winnersorchestrateflatnetworksofagenticteams

ProliferationofAIagentswithouttherightcontext,steering,andorientationcanbearecipeforchaos.Winningoperatingmodelsofthefuturewillempoweragenticteams,withflatdecision

andcommunicationstructuresthatoperatewithhighcontextsharingandalignmentacross

agenticteamstoensuretheymoveinsync.Organizationcharts(basedontraditionalhierarchicaldelegation)willpivottowardagenticnetworksorworkcharts(basedonexchangingtasksand

outcomes).3Finally,agenticnetworksarenotnecessarilylimitedtotheboundariesofasingleorganization,anddifferentoutcomesmaybesourcedfromdifferentparties,openingupnewB2Bopportunities.

3.Governance

Intheagenticorganization,governancecannotremainaperiodic,paper-heavyexercise.As

agentsoperatecontinuously,governancemustbecomerealtime,datadriven,andembedded—withhumansholdingfinalaccountability.

Decision-makingaccelerateswithreal-timedata

Traditionalbudgeting,planning,andperformancemanagementcyclesaretooslowforAI-first

workflows.Earlymoversareexperimentingwith“agenticbudgeting,”inwhichAIagentsproposebudgets,scenarioagentsrunforecasts,andreportingagentsprovidereal-timeinsights.Finance

1MartinFowler,“Twopizzateam,”MartinF,July25,2023.

2“Dunbar’snumber,psychologicalsafetyandteamsize,”PsychSafety,October21,2022.32025:Theyearthefrontierfirmisborn,Microsoft,April23,2025.

Theagenticorganization:ContoursofthenextparadigmfortheAIera6

leadersshiftfromcollectingspreadsheetstointerpretingsignals,stress-testingscenarios,andengagingdirectlywiththebusiness.

Agentscontrolagentsthroughembeddedguardrails

JustasDevSecOps(development,security,andoperations)embeddedautomatedchecksintodigitaldelivery,agenticorganizationswillembedcontrolagentsintoworkflows.Criticagents

willchallengeoutputs,guardrailagentswillenforcepolicy,andcomplianceagentswillmonitorregulation.Everyactioncanbeloggedandexplainedinrealtime—fromdataprivacytofinancialthresholdstobrandvoice.AnAIgovernanceframeworkacrossthelifecycleofAIagents—fromagentdiscoveryandinitiationtodecommissioning—canbalancespeedandscalewiththe

requiredsecurityandcontrolmechanisms.

Humanaccountabilityandoversightremain

Humanaccountabilityandoversightwillremainessential,buttheirnaturewillchange.Ratherthanconductline-by-linereviews,complianceofficersandleaderswilldefinepolicies,monitoroutliers,andadjustthelevelofhumaninvolvement.Thechallengeisfindingthesweetspot:

enoughoversighttomanageriskwithoutpullingagentsbacktohumanspeed.Companies

thatgetthisbalancerightwillcapturemoreoftheagenticadvantage.Ultimately,thescaleof

agenticadoptionwillbecappedbyhowmuchoversightcapacityhumanscanprovide—makinggovernanceitselfapotentialbottlenecktoproductivity.

4.Workforce,people,andculture

Intheagenticorganization,humanswillmovefromexecutingactivitiestoowningandsteeringend-to-endoutcomes.Thatshiftdemandsnewprofileswithdifferentskillsandaculturethatprovidescohesionandpurpose.

Thehybridagenticworkforceneedsanewtalentsystem

Asagentstakeonexecution,peoplewillincreasinglydefinegoals,maketrade-offs,andsteeroutcomes.Thiswillchangehowcompaniesplanforahybridworkforce,whomtheyhire(or

borrow),howtheydeployhumanorAItalent,andhowtheymeasuresuccess.HRsystemsnotonlytrackhumanemployeesbutalsoarearepositoryofagentsandagenticworkflows.

Performancemanagementanchoredintaskcompletionwillgivewaytosystemsthattrackhowwellpeopleorchestrateagents,unlockvalue,anddeliveroutcomes.Inthisnewparadigm,thetalentsystemitselfmustberethought—fromcareerpathstoincentivestoleadershipmodels.

Newtalentprofileswithdifferentskillsemerge

Inourworkwithpioneeringorganizations,weseeAIagentsreplacingtaskshistoricallyhandledbyknowledgeworkers,suchasanalyzingdocumentsandcreatingAPIs.Atthesametime,we

seerisingdemandforotherskills—forexample,deepproblem-solvingwithanend-to-endlens,applicationofsystemdesign,andtheabilitytoapplypatternrecognitiontoedgecaseswhereagentsfail.

Threerolesareemergingashumansworkalongsideagents(Exhibit2):

—M-shapedsupervisors:broadgeneralistsfluentinAI,orchestratingagentsandthehybridworkforceacrossdomains

—T-shapedexperts:deepspecialistswhoreimagineworkflows,handleexceptions,andsafeguardquality

Theagenticorganization:ContoursofthenextparadigmfortheAIera7

Exhibit2

Newtalentpro?lesforsupervisors,specialists,andfrontlineworkerswillemergeintheagenticorganization.

SkillsrequiredforevolvingrolesintheAIera

M-shapedgeneralmanager

Builds,supervises,andoptimizeshybridworklows;hasend-to-endproductorprocessresponsibility

Highercognitive1

AI

Socioemotional

DomainDomain

Multiskilledwithabilitytoapplyknowledgeacrossdomains

T-shapeddeepspecialist

Provideshumanoversightofwork-lows;hasextensivesubjectmatterexpertise

Highercognitive

AI

Domain

Deepknowledgeinaspeci?c

domain

AI-empoweredfrontlineworkerFocusesoninterpersonaltasks

wherehumantouchisrequiredforoptimaloutcomes

Socioemotional

AI

DomainHigher

cognitive

Highsocioemotionalskillswith

basicAIluency

Gapsmeasuredagainstrolestoday

Howdoesitcomparetotoday’smanagerrole?

Needsnovelskillmix,particularlyinintegrativeend-to-end

thinkingandabilitytomanageAI-?rstworklows

Howdoesitcompareto

today’sspecialistrole?

Augmentscurrentexpertise-basedroleswithabilityto

teachand?ne-tuneagentic

systemsandhandleexceptioncasesandagenticlearning

Howdoesitcompareto

today’sfrontlinerole?

Enhancesexistingfrontlineroles—selectedfortheir

socioemotionalskills—byaddingAIcapabilities

1Forexample,criticalthinking,problem-solving,creativity,decision-making,abstractthinking.

McKinsey&Company

—AI-augmentedfrontlineworkers:employeesinsales,service,HR,oroperationswhospendlesstimeonsystemsandmoretimewithhumans

Leadersthemselveswillalsoevolve.CEOs,productofficers,andcomplianceheadswill

increasinglyneedthetechnologyfluencyonceexpectedonlyofchiefinformationofficers.Fillingtheseroleswillrequire

upskillingandreskillingatscale

.Earlyevidenceshowsthatemployees

withouttechnicalbackgroundscanlearntomanageagenticworkflowsasquicklyastrained

engineers.Careerpathsandperformancesystemswillneedtoadaptas“boxesandlines”givewaytoecosystemsofhumananddigitalskills.Astheseprofilestakehold,theconstructsof

“organization”and“employee”willbecomemorefluid,withecosystemsofhumananddigitaltalentblendinginsideandoutsideanorganization.

Cultureactsasglueandethicalcompass

Culturewillbecomeboththeoperatingglueandtheethicalcompassoftheagenticorganization.Pioneeringagenticorganizationshighlighttheneedfororchestration—toalignteamsaround

sharedcontextandoutcomes,identifytherightmixofhumanandAIcapabilities(asnot

everythingneedsagenticAI),andbuildtrustbetweenhumansandagents.Theculturecompassembedsvaluesandlong-termpurposeintoagenticsystems,socompaniesdon’tchaseshort-

Theagenticorganization:ContoursofthenextparadigmfortheAIera8

termefficiencyattheexpenseofcohesionandtrust.Earlypioneersshowthatclarity,decisive

leadership,andcontinuouslearningarecritical—butwhatwilldifferentiatewinnersistheirabilitytopreservecohesionandidentitywhiletransformingatpace.

5.Technologyanddata

Intheagenticorganization,technologyanddatawillgetdemocratized,supportedbyan

agentic

AImesh

.Agent-to-agentprotocolswillmakeintegrationacrosssystems,machines,andhumanseasierandcheaper.Successfulscalerswillbalancebuild-versus-buydecisionsbasedon

sourcesofdistinctivenessandcompetitiveadvantage,avoidingtechnologyorvendorlock-insotheycanadaptquicklytoafast-evolvingofferinglandscape.

DistributedownershipofITanddatabecomesfeasible

Inthedigitalera,technologyanddatasystemsevolvedfromcentralizedmonolithsand

databasessittingfarfromthebusinesstoward

microservices

and

dataproducts

sittingclose

tothebusiness.Thisrequiredsignificantsoftwareanddataengineeringexpertisetodesign,

develop,andmaintaintheunderlyingtechnologyanddata.Intheemergingagenticage,

business-sideemployeeswillbeabletoindependentlycreatesoftwareassetsandmanagedatathroughagenticAI,whichautomatesthesoftwaredevelopmentlifecycle(SDLC)withoversightfromdeepspecialists.Earlyadoptershaveseenproductivityatleastdouble,withemployees

fromdiversebackgrounds—suchasaFrenchliteraturegraduateinoneofourteams—provingascapableassoftwareengineersinbuildingagenticworkflows.

Toscalethistransformationresponsibly,organizationsmustadoptagenticplatformsand

architectures,suchasanagentic

AImesh

.Theseplatformsprovidereusable,high-performing

“atomic”agentsanddataproductsequippedwithtechnicalsafetyguardrailstopreventbuildupof

technicaldebt

orcybersecurityrisks,whileunlockingunprecedentedlevelsofdemocratizationandinnovation.

Agent-to-agentprotocolseaseinteractionsandintegrations

Agent-to-agentprotocolsareredefininginteractionsbetweenhumans,agents,ITsystems,anddevices.RatherthanrelyingontraditionalITsystemintegrationssuchasmiddlewareand

APIs

thatrequireheavyprogrammingandcustomsystem-to-systemconnections,agent-to-agent

protocolsenablesystemstouseagentstocommunicatewithothersystems.Bymovingtoagent-to-agentdialoguethatsitsabovetheunderlyingsystemcomplexity,organizationscanintegratelegacysystems,cloudplatforms,andevenmachinessuchasdronesintocohesiveworkflows

morequicklyandatlowercost.Moreimportant,thisallowsforfasterexperimentation—inwhichnewcapabilitiescanbetested,scaled,ordeprecatedwithoutmonthsofengineeringeffort.

Dynamicsourcingbecomescritical

Manybusiness-criticalplatformswerehistoricallybuiltin-houseorselectedinrigoroussourcingprocesses,withmultiyearimplementationtransformations.Thesesystemswereintendedto

remainlargelystabletosecurecompetitivenessfordecades.Amuchmoreflexiblestrategywillbeneededintheagenticage.LargelanguagemodelsandAIproductsareevolvingsofastthatlockinginonesolutionorvendorcanleadtotechnologythatisoutdatedinamatterofweeks.Atthesametime,organizationswillneedtowallinproprietaryorganizationalcontext,institutionalknowledge,andnonpublicdataforcompetitiveness.Thisrequiresarchitecturethatseparates

theagenticstructure,logic,anddatafromtheunderlyingvendorlandscape.

Theagenticorganization:ContoursofthenextparadigmfortheAIera9

Howtostartthejourney

Themostfrequentquestionweheardinourdiscussionswithexecutiveswas,“HowdoIstart?”ExecutiveswonderhowtocreateaNorthStarvisionwithoutclarityonwhatthefutureholds;

howtoassessmaturityandupgradeneedsfordata,technical,andgovernancefoundations;howtosetprioritiesforvalueandfeasibility;howtobringtheorganizationalongintermsof

skillsandmindset;andhowtoscalefasterthanrivalstocreateacompetitiveadvantage.The