opentext"

WHITEPAPER

Smartsuccess

IgnitethepowerofAlanddatascienceacrosstheorganization

opentext"

Contents

Theimportanceofartificialintelligence4

ThebenefitsofintroducingAl5

Keystageswithinthedatascienceprojectlifecycle6

Creatingahigh-performingdatascienceteam8

Thecriticalroleofinternalandexternalpartnerships9

FiveAlusecasesyourcompanycanconsider:PoweredbyOpenTextMagellan10

RealizingthebenefitsofusingMagellanforAl11

Resources12

Smartsuccess:IgnitethepowerofAlarxidatascienceacrosstheorganization212

opentext"

Summary

Artificialintelligence(Al)transformsorganizationsbyempoweringthemtomake

moredata-drivendecisionsandeliminatestediousandrepetitiveprocesses.Yet,

manycompaniesstruggletoseehowtodeploytherightAlsolutionfortheir

organization.WhiletheymayhaveageneralunderstandingofwhatAlcando,it

remainsunclearastohowtheycanleveragedatascienceandadvancedanalytics

toimproveefficiency.

ThiswhitepaperwillhelpyouidentifywhetherAlistherightfitforyourorganization

andhowtosetupanAlstructurethataddsvaluetothebusiness.Itoutlinesthe

benefitsofAlfortheenterprise,providesalistofquestionstoaskbeforegetting

startedandoffersguidanceforstructuringahigh-performingdatascienceteam.In

addition,itdetailstheiterativeprocessformanagingasuccessfulAlprojectwitha

comprehensive,step-by-stepguide.

ItalsoprovidesfiveusecasestoalloworganizationstoexplorethepotentialofAl

withintheirownorganizations.

Smartsuccess:IgnitethepowerofAlarxidatascienceacrosstheorganization3”2

opentext"

ThebenefitsofintroducingAl

Withevolvingcustomerdemandsandagrowingsenseofimmediacyaffectingall

industries,businessleadersmustmakedecisionsquickerthanever.Ofcourse,

informeddecisionsarebornfromtheabilitytoanalyzeasmuchdataaspossible,

butcollecting,contextualizing,analyzingandderivinginsightsfromhugevolumesof

informationistootime-consuming.

Machinelearningautomatesthetediouselementsofhumanresearchandreview

toprovidevaluableanalysesandrecommendations,amplifyingtheroleofhuman

employeesacrossbusinessprocesses—noteliminatingit,assomemightfear.

Withthesecapabilities,AI-poweredsolutionscan:

呼

SavetimeDriveHelpyoustayImprove

andmoneyefficiencycompetitivesecurity

Maximizingthebenefitsforyourenterprise

TheAljourneywillbedifferentforeachorganizationbasedonitsassessmentsof

needs,strengths,processes,goals,etc.Itshouldthereforenotbeexpectedthat

allorganizationsadoptAlinthesameway.Considerthefollowingquestionswhen

determiningwhetheranAl-poweredsolutionwillnotonlybefeasible,butalso

beneficial,foryourorganization:

WhatspecificobjectivescanweachievebyintroducingAl?

Alwillonlyyieldorganization-widesuccessifyouareabletolookatitthrough

astrategicbusinesslens—notasanITproject.Thisrequirescross-functional

collaborationtoidentifytheorganization'soverarchingobjectives.Youmightchoose

toimplementtechnologyinawaythatreducesrisk,improvesproductivityorhelps

youmakemoreaccuratepredictions.WhileAlcandoallofthisandmore,starting

withaspecificgoalwillhelpyoudevelopadetailedandactionableplantogetthere.

HowwillAlhelpussecureacompetitiveedge?

IfyouarenotsurehowtoleverageAltohelpsecurebetterperformance,consider

anyindustrytrendsthatmightbetakingshape.Analyzingthedirectionyour

businesslineismovinginshouldhelpyouidentifywaystomeetcustomerdemands

andstaycompetitivewithAl.InFinancialServices,forexample,banksareusing

Altohelpthemmakemoretailoredproductandservicerccommcndationo.

Usecasesfromyoursectorwillalsoprovetobeavaluableresourcetohelpyou

envisiontherightsolutionsforyourneeds.

Doyouhavetherightdata?

Doyouhavedatarelevanttoansweringthebusinessproblem?Howmuchdatado

youhave?Whereisyourdata?Isdatareadilyaccessibleforanalysisandmodelling?

Withtheanswerstothesequestionsinmind,youcanbeginpursuingasuccessful

Alprojectinyourorganization.

Smartsuccess:IgnitethepowerofAlarxldatascienceacrosstheorganization5/12

opentext"

1.DefinetaskKeystageswithinthedatascienceprojectlifecycle

2.CollectdataTheprocessofdevelopingamachinelearningmodelishighlyiterative.Often,you

3.Modelexplorationwillfindyourselfreturningtopreviousstepsbeforeproceedingtoasubsequent

one.Afterall,amachinelearningprojectisnotconsideredcompleteafterthe

4.ModelrefinementfirstversionhasbeendeployedInstead,thefeedbackyoucollectaftertheinitial

5.Testingandevaluationversionwillhelpyoushapenewgoalsandimprovementsforthenextiteration.

6.Deploymentandintegration1.Projectplanning/Definetask

7.ContinuousmaintenanceManyAlprojectsfallshortbecauseorganizationsjumpdirectlyintomodeling

withouttakingthetimetoplan.Startingwithacohesiveplangetseveryoneon

thesamepageandmaypreventroadblocksdowntheroad.Whileitisnotalways

simpletodefineamodeltaskfromthebeginning,brainstormingshouldhelpyou

identifyprojectswithhighpotential,includingthosethatcouldmakeabigimpact,

yetstillbefeasible.Wheremightpredictiondrivethelargestvalue?

Duringthisstep,youshouldalsoestablishwhatkindofdatayouwillneedto

supportyourAlproject,howyouwillacquireitandhowmuchdataisneeded.For

example,ifyouarethinkingofdeployingasolutionforpredictivemaintenanceona

train,youmightpullinformationfromloTsensors,weatherpatternsandpassenger

travelpatterns.

Definetask

Machinelearningdevelopmentlifecycle

2.Dataacquisition/Collectdata

Howyoulabelyourdatacanhaveahugeimpactonmodelperformance.Insome

cases,youmaybeabletouseaself-labelingsystem,butinothers,manually

labelingdatamaybenecessary.Beconsistentinyourlabelingcriteriaandbesure

toassociatespecificmodelswithdatasetversionstoavoidconfusion.

Smartsuccess:IgnitethepowerofAlarxldatascienceacrosstheorganization6/12

opentext"

3.Modelexploration

Todeterminewhetherthemodelsyourdatascientistsarecreatingwillbeeffective,

youmustfirstdeterminebaselinesagainstwhichmodelperformancewillbe

measured.Thebestapproachistostartsmallwithaninitialdatapipeline.Youwill

likelytestmanyideasthroughoutthisstep,butultimately,oneleadingmodelwill

emerge.Youcanreproducetheresultsofthismodel,thenapplyttoyourdataset

forasecondbaseline.Thereafter,youcanrevisitstepsoneandtwotoensureitfits.

4.Modelrefinement

Oncethemodelisinplace,refineitbydebuggingitandperforminganerror

analysistopinpointpotentialfailures.Asyouenhancethemodel,considerthe

followingquestions:

?Canthedatascientistexplaintheirchoicesandaretheyreproducible?

?Howistheaccuracymeasurec?

?Haveyoumeasuredthepotentialprofitsavingsoftheproject?

-Canothersunderstandthenotesandprojectmethodology?

?Whatisthereleaseplanandhowwillyoumeasuresuccess?

?Hasdatasecuritybeenreviewedonthealgorithmsandpipelines?

5.Testingandevaluation

Often,datascientistsfindthat"datainthewild"tendstobedifferentfromthe

datatheyusedfortrainingpurposes.Forthisreason,itisimportanttorevisitthe

metricformodelevaluationtoensurethattheoneyouhavechosenisappropriate

forfacilitatingthedesiredactions.Nowisalsothetimetowritetestsfortheinput

datapipeline,modelinferencefjnctionalityandspecificscenariosexpectedin

production.

6.Deploymentandintegration

Forbestresults,startwithamodeldeploymentamongasmallgroupofusers.If

everythingrunssmoothly,rollitouttoallusers.Youshouldcontinuetomonitorlive

dataandmodelpredictiondistrioutions.

7.Continuousmaintenance

Followingdeployment,besuretoaddressanychangesthatimpactthesystem

andretrainthemodelperiodicallytopreventitfrombecomingwornout.Should

modelownershiptransfer,youshouldensurethatthenewteamisproperly

educatedonthemodel.

Ofcourse,toexecutethesestepseffectively,youmusthavetheproperteaminplace.

Smartsuccess:IgnitethepowerofAlarxldatascienceacrosstheorganization7/12

opentext"

Creatingahigh-performingdatascienceteam

Alprojectsarenotsolosports.Theyrequiretheongoingcontributionsand

collaborationfromanumberofkeyteammembers.Becausedatascienceisan

intersectionofvariousfunctions,includingmathandstatistics,computerscience

andbusinessknowledge,youwillneedadynamicgroupofindividualstoforma

high-performingmachinelearningteam.

Typically,teamsinvolveindividualsfromdatascienceanddataengineering,as

wellassubjectmatterexperts.Hereisalookateachoftheserolesinmoredetail:

Dataengineers:Thesespecialistsaretypicallytaskedwithbuildngthedata

infrastructureofanorganization.Theyhavestrongprogrammingandhardware

skills,arefamiliarwithbigdatatechnologiesandexcelinbuildingdatapipelines.

Theyarenotnecessarilyexpertsinanalyzingandmodellingdatabutshouldbeable

toworkwithrelevantbusinessdivisionstodeterminethedatacharacteristicsfor

eachusecase.

Datascientists:Afteryouhavebuiltoutyourdatainfrastructure,youwill

needpeoplewhocantakethatdata,cleanit,analyzeit,applyalgorithms,

runcxpcrimcntGandcommunicatercGultGeffectively.ThcccprofcGGionalG

typicallyusetools,suchasJupyter'notebookandRStudio?,haveknowledgeof

programminglanguages,suchasRandPython?,andexperienceworkingwithbig

datatechnologies,suchasApacheHadoop?andApacheSpark".haddition,they

haveastrongbackgroundinstatsties,programmingandmachinelearning.

Subjectmatterexperts:Subjectmatterexpertshelpcultivatestrategies,generate

ideasandexaminefactorsnecessaryforsupportingdifferentusecases.They

understandthekeybusinessproblem,processesandchallenges.Theirinput

andknowledgearecrucialinhelpingtheteamdevelopmeaningfulsolutionsthat

providebusinessvalueandovercomechallenges.

Datasciencemanager:Datasciencemanagersarehands-onleaderswhohelpbuild

thefoundationofanorganization'sdatasciencestrategy,recruitandcreatetalent

teams,ensureeffectivecommunicationamongmembersanddevelopprocessesfor

theteamtofollow.Theyareinchargeofconnectingthedatascienceandanalytics

teamwiththerestoftheorganization,otherdivisionsandexecutives.Oneoftheircore

responsibilitiesistotranslatecomplexAlandMLterminologytononexpertsandmake

suretheteamworksinalignmentwiththestrategyoftheoverallorganization.

DataengineeringDatascientistSubjectmatterexperts

?Helpsidentifythe?Cleans,massages?Providesfeedback

rightdatasetsandorganizes(big)dataonoutouts

?Helpsbuildthedata?Appliesdescriptive,?Helpsensureanalysis

infrastructureofpredictiveandprescriptiveisapptopriate/moving

anorganizationanalyticaltechniquestointherightdirection

gaininsights,build

modelsandsolve

abusinessproblem

Datasciencemanager

?Guidestheteam-egardingthebestuseoftoolsandresources

?HelpstranslatecomplexAlandMLterminologyfornon-experts

?Ensurestheteamworksinalignmentwiththestrategyoftheoverallorganization

Datasciencecoreteam

Smartsuccess:IgnitethepowerofAlarxldatascienceacrosstheorganizationai2

opentext"

Thecriticalroleofinternalandexternalpartnerships

Foreffectiveexecution,Almustbebuiltonstrongpartnerships,bothinternallyand

externally.Eachpartyinvolvedmustbealignedwiththeoverarchingdatastrategy

tokeepalleffortsfocusedonthesamegoal.Wheneveryoneisalignedandknows

theirrole,theycanachievereal,measurableresultsfromtheAlpractice.

Internalpartnerships

Notonlyishavingtherightpeopleonyourteamimportant,soisensuringthey

havetheresourcesneededtoadoptAltechnologyandtocollaborate.Tofacilitate

broadAltraining,makesureyojhaveinputfromanAlexpertwhocanguide

yourorganization.Theycanworkwithyourchieflearningofficer(CLO)tofindor

developtrainingmaterials,includingvideotutorialsandcourses.Theorganization

shouldthendevelopproceduresforeachroleinvolvedwithAl,includingexecutive

leaders,businessunitmanagersanddatascientists.

Before,duringandafterdeployment,theremustbeconstantcorrmunication

betweentheAlteamandthebusinessunit.Eachspecialisthasstrengthsand

boundariestotheirskillsets,soworkingwithotherswillbeanessentialaspectof

projectsuccess.Typically,datascientistsordatasciencemanagerswillensure

cross-functionalcollaborationtocreatebusinessvalueandensurecomplexAlterms

arebeingtranslatedtousableinsights.Theseindividualsmustalsoensurethattheir

workisalignedwiththeorganization'soverarchinggoal.ThisiswhereAlsystems

empowerteamstobreakdownsilosandcommunicateinsightsacrossteams.

Externalpartnerships

FindingtherightAlpartnerforyourorganizationisanothercomponent

integraltothesuccessofyourAlpractice.AtrustedAlpartnershojldallowyou

toleverageunderusedyetvaluabledata,allowdatascientiststoworkintheir

familiarenvironments,suchasJupyternotebookandR,andprovidetheability

tooperationalizemodelsinbusiness-friendlyinterfaces.Establishngaproject

managerwhocancollaboratewithyourAlpartnerwill

alsobecriticaltosuccess.

Theimportanceofexecutivebuy-in

OnecommoncauseoffailureamongAlprojectsislackofbuy-in.Wniledatascientists

andsubjectmatterexpertsmaybeabletoseethevalueofAlintheenterprise,

supportmuststartatthetop.CommunicatinghowanAlinvestmentcouldhelpthe

companysavemoneyandpilotingsolutionscanhelpteamsestablishcredibilityand

attractsupportfromleadership.Moreover,explainingmodeloutputsinconciseand

tangibletermscanhelpaccelerateadoptionandbuildtrustamongs:akeholders.

Smartsuccess:IgnitethepowerofAlarxldatascienceacrosstheorganization9/12

opentext"

FiveAlusecasesyourcompanycanconsider:

PoweredbyOpenTextMagellan

Al-augmentedrecommendations

Consumersareinundatedwithfrequentandirrelevantproductoffers.To

understandhowtheycouldbetterservetheircustomers,amajorbankdeployed

acustomizedpersonalizationmodelwithintelligentrecommencations.The

bankimplementedasystemthatharnessesAlandbigdatatomonitorcustomer

actions,predicttheirneedsanddeliverhighlypersonalizedexperiencesacross

everychannel.Withadeeperinsightintocustomerpreferencesandbehaviors,

thebankhasimproveddealsandoffersfromretailpartners.Withfewerannoying

andirrelevantoffers,thecustomerexperiencehasimprovedandthebankhasseen

higherclick-throughandconversionratesontheiroffers.

Intelligentcapture

Whiletraditionalcontentcapturesystemshelpbusinessesmanagetheircontent,

thereisagrowingneedtoautomatetheclassificationandroutingofcontentatscale.

Al-enhancedcapturecanextractinformationfromenterpriseconlentandroute

documentstotheappropriateworkflow.Withthesolution,organizations,suchas

largebanks,canintelligentlyclassifydocumentsandfreeupstafffromsortingthrough

thousandsofincomingdocumentseveryday.

Classification/Smartmigration

FromspreadsheetstoMicrosoft8Worddocumentsandemail,organizationsare

swarmingwithbillionsofpiecesofcontent.Managingthisunstructured,textual

dataisnotoriouslydifficult,butmachinelearningcanhelporganizationscategorize

informationandtriggerprocesses,therebyautomatingknowledgeclassification.

Thisisespeciallybeneficialfororganizationswithsensitivedata,includi