2024年智能煉金術(shù)：生成式人工智能如何徹底變革現(xiàn)代企業(yè)中的商業(yè)智能和分析白皮書

TheAlchemyofIntelligence:

HowGenerativeAIcan

revolutionizeBusiness

IntelligenceandAnalyticsinModernEnterprises

TableofContent

Introduction 03

BusinessUser 04

Opportunities 04

Challenges 04

Recommendations 07

BusinessAnalyst 08

Opportunities 08

Productivity 08

ProgrammingforNon-Programmers 10

Insights 10

Beautification 11

Challenges 12

Usefulness 12

Trust 13

HumanErrorandDocumentation 14

Security 14

Recommendations 15

Test 15

Adopt 16

Train 16

DataAnalyst/CitizenDataScientist 17

Dylan’sTransformation 18

ArrivalofAIAgent 19

ATeamofAgentsEmerges 21

Summary 21

ITAdministrator 22

InfrastructureDemands 22

DataGovernanceandSecurity 23

Observations 24

SystemArchitect 25

Opportunities 25

Challenges 26

Recommendations 26

Summary 27

Conclusion 27

Authors 28

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA2

Introduction

Intherapidlyevolvinglandscapeoftechnology,businessesare

constantlysearchingforinnovativewaystostayaheadofthe

curve.OnesuchgroundbreakingadvancementisGenerativeAI,atechnologythathasthepotentialtoreshapethefutureofBusinessIntelligence(BI)andanalytics.Imagineaworldwheredataspeaksdirectlytoyou,whereyouranalyticstoolsnotonlyansweryour

queriesbutalsoanticipateyourneeds,providinginsightsyouhadn’tevenconsidered.ThisisthepromiseofGenerativeAI–atoolthattransformsrawdataintorich,actionableintelligence,empoweringbusinessestomakesmarter,fasterdecisions.

Thejourneythroughthiswhitepaperwilltakeyouintotheheartofthisrevolution.We’llexplorereal-worldscenarioswhere

GenerativeAIactsasacatalystforenhancedproductivity,sharperinsights,andmorebeautifuldatavisualizations.Frombusiness

userslikePeggySue,whoexperiencethemagicofAI-powered

chatbots,todatascientistslikeDylanDawson,wholeverage

generativemodelsforunprecedenteddataanalysis,thenarrativeunfoldstorevealbothopportunitiesandchallenges.Bytheendofthisexploration,youwillunderstandnotonlythetransformativepowerofGenerativeAIbutalsohowtoharnessiteffectively

withinyourenterprise.Forsimplicity,wehavebrokenthisintovariousreal-worldpersonas.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA3

BusinessUser

UsesdashboardsandreportsgeneratedbyBItoolstomakeinformedstrategicdecisions.

Businessusers,likePeggySue,aretheworkerbeesofany

corporatehive.Checkingnumbershere.Doingtheworkthat

needstodobedonethere.Buzz.Buzz.Buzz.Thissectionexplorestheuniqueopportunities,challengesandrecommendationsfor

otherslikePeggySue.

Opportunities

PeggySuewasthrilledtohavethischancetolaunchhercareerwithaglobalbeerdistributioncompanyknownforthequalityofitsbeersandforbeingarealhigh-techleaderintheindustry.ShehadmanycoursesattheUniversitysheattendedondashboardsandanalytics,andtheyreallypaidoffforherduringherfirst6

months.Neverdidadaygobywhenshedidn’tseepostsonherLinkedInfromthisgoodfriend,orthatfriend,ravingabouttheirexperienceswithsometypeofgenerativeaichatbots.

Shewasthrilledthedayshereceivedanemailstatingthat

herorganizationwouldbegettingachatbotalongsidetheir

dashboards.Suddenlythereitwas,andPeggySue’sheartwasallaflutterwiththepossibilities.

Everythingshereadusedphraseslike“Gamechanging”“makeslifesomucheasier”“willreplaceallworkerseverywhere”somepostersmightaswellhaveusedthewords“hocuspocusdominocus”

becauseitsoundedlikemagic.

PeggySue’smindwasracing“LookatthebeautifulinputboxwhereitsaysIcanaskanything.”UnfortunatelyforPeggySueanother

thoughtstruck,“Icanaskanything,butIhavenoideawhattoask”

Challenges

Whilemanyorganizationsrushtogetabotintothehandsof

businessusers,blankcanvasparalysiscantakeoverbecausetheyfocusedonthetechnology,andnottrainingtheirstaffhowtouseit.

EventuallyPeggySuebeganaskingthequestionsastheycame

tohermind“TellmethetotalsalesforourbeerinSouthAmerica.”

“WhichlocationissellingthemostofourPorters?”“Whichdivisionisn’tdoingwellfinancially?”Eachofherquestionsreceivedananswer.

Thechallengeforherwasthatmostanswersjustseemedwrong.Whenshedugintothedetailedrecordsinherdashboard,she

confirmedtheywerewrong.“Well,Ireckonthisthingisn’tverygoodatmath.WhydidtheygivethisthingtomeifIcan’taskittoaddupnumbers?”

Othertimesthefigureswereaboutmeasuresthatsheknewthecompanyhadmultiplewaysofcalculating.“Thisanswermayberightforoneofthemeasures,buttheanswerdoesn’texplainwhichcalculationmethodisevenused.Evenifitsaccurateforonemethod,Ihavenowayofknowingforsureit’sthemethodmybossexpectstosee.”

“Maybeit’stheresoIcanaskquestionsaboutthedashboarditself”

shethoughttoherself.Whichwasgoodbecausealthoughshehadreceived10minutesoftrainingfromafranticallybusytrainer,shedidn’tremembereverything.So,sheasked“HowdoIfigureout

whichdivisionisstrugglingonmydashboard?”

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA4

Whilethoughtprovoking,shewashopingforspecificinformationaboutthedashboardshewaslookingat.Afterafewquestionslikethisshegotalittleworriedthatperhapsmanagementwastrackingherquestionsandthatifshekeptaskingquestionslikethis,she

mightbereprimandedforhavingnotalreadylearnedeverythingaboutthedashboardevenafterthewhopping10minutesof

trainingshehadreceived.

Onedayasshewasreviewingsomequarterlyfiguresandhercolleagueswereoutoftheoffice,somethingstruckher:

“MaybeIshouldbeaskingthesametypeofquestionsInormallyaskthem.”So,shedid:“WhataresomereasonsthatcouldexplainwhywearesellingsomuchmoreBrownAlethanotherbeers?”

Atthatmomentalightbulbwentoff,andachoruswassinginginPeggySue’shead.Assheproceededwithheranalysis,shewasagaincuriousaboutthedata.Althoughsalesweresohighfor

BrownAle,theprofitsweren’t.

Shequicklytyped“Whataresomereasonswearenotmaking

muchprofitonbrownaleconsideringwesellsomuchofit?”intothehandylittle“AskAnything”inputboxandwasagainimpressedwiththeresponse.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA5

PeggySuewasinspiredbythispatternofaskingwhenshewas

puzzledaboutwhatcouldexplainthingsthatshedidn’tseeinthebarchartsandpiechartsandlinechartsonthescreen.Aftera

meetingonedaywheresheheardaboutacontestthecompanywashavingwhereanyemployeescouldmakesuggestions

abouthowtoincreasesalesshedecidedtogetreallyboldinherquestioning:“CanyoutellmeculturallywhywearesellingsomuchBrownAlewherewedoandwhatotherculturesaresimilarthatwecouldstartsellingitto?”

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA6

Recommendations

Whileonthecruiseshetookafterwinninghercompany’s

suggestioncontest,PeggySuehadmanychancestorecount

herexperiencestodatewithGenerativeAIinsidehercompany’sBusinessIntelligencetooltootherpassengers.

?Don’taskquestionsofanykindthatinvolvemath.

?Realanswerstorealbusinessproblemstypicallyinvolvecomplicatedbooleanlogicthatturnthemillionsofrows/columnsofdataintotruth,thatyourmodelmaynothaveaccessto.

?Don’taskforanswers,askforadvice.Answersimplyyouaredoneandwillact,butadviceimpliesyouwillaugmentthe

inputwithyourownknowledgethenact.

Onepassengershetalkedtooveroneofthosetalldrinkswith

fruitwedgesandanumbrellasaid“Wehave175differentBI

applicationsthatIworkwith.Whichoneofthemdoyouthinkis

therightonetostartusingwithoneofthoselargelanguagemodel

chatbotthingamajiggies?”PeggySuehadafewbitsofadviceforhim:

?TheoneusedbythegroupthatyouhaveprovidedsomeAILiteracytrainingtobeforehand.

?Theonethatyourbusinessuserspeekoverthecubiclewallsandchatwitheachotherthemostabout.

Storytellingasideforamoment...thebiggestrecommendationwecanofferforBusinessUsersistothinkofyourGenerativeAIchatbotslikeyouwouldanyothertrustedadvisorinyourlife.

?Theyaren’tgoingtodoyourworkforyou.

?Theywon’talwaysprovideadviceyouagreewith.

?Unlikeotheradvisorsinyourlife,theyarenevertoobusyforyoutoask,andtheynevergetoffendedwhenyouaskthe

samequestion10differentways.

?Youarestillultimatelyresponsibleforyourwork,soalwaysuseyourownintelligencetoaugmentanyadviceyoumayreceive.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA7

BusinessAnalyst

WorkscloselywithstakeholderstounderstandbusinessrequirementsandusesBItoolstocreatereports,dashboards,andvisualizations.

Let’srewindtheclocksixmonthspriorandlookathowPeggy

Sue’snewBIcopilotcametobe.SallySue,theunstoppable

business-analyst-turned-datascientist,hasbeenexperimentingwithGenerativeAIforhercodingtasks.Copilotsareexcellentatgeneratingcodeandsummarizinglargeamountsoftext,andherbusinessrecentlyadoptedaBItoolthathasacopilotbuiltintoit.“Wow!”shethought.“Icananalyzemydataandbuilddashboardsjustbyaskingquestions?”Sallywasthrilledattheidea–aswas

herCIO.Canyouimaginethenumberofquestionsthatcouldbequicklyansweredifpeoplecouldchatwiththeirdataanddashboards?

Beyondtheexcitement,Sallyrealizedthatthereareseveral

potentialrisks.She’staskedwithevaluatingthiscopilotfor

productionandsendingitovertobusinessuserslikePeggy

Sue.WhatkindsofquestionsmightPeggyask?Whatkindof

dashboardswouldpeoplebuildwiththis?Howdowecertifythisforproductionuse?Whataboutdatasecurity?Isthereavariablecosttousethis?ThereareanumberofquestionsthatcametoSally’smind.Shebrokeherquestionsdownintotwomainareas:opportunitiesandchallenges.

Opportunities

GenerativeAIbringsampleopportunitiesforworkingwithdataanddashboardsbyhavingaconversationwithit.Sally’sgoingtofocusonthreeofthesepotentialopportunities:

1.Productivity-CanGenerativeAIimprovetheproductivityofbothmyjuniorandseniorbusinessanalystswhenworkingwithaBItool?

2.Insights-Canmystakeholders“chatwiththeirdashboard”togetfastertimetoinsight?

3.Beautification–CanGenerativeAIhelpcreatebetterlookingbeautifuldashboardswithbest-practicesautomatically

builtin?

Let’sexplorethesethreeconcepts.

Productivity

Buildingdashboardsisnoeasytask.Therearemanyconsiderationsthatmustbeaccountedfor:

?Who’stheaudience?Anexecutive?Abusinessunit?Anotheranalyst?Yourself?

?Whatmetricsdotheycareabout?

?Doesthedatasupportthosemetrics?

?Howoftenwilltheybeviewingthedashboard?

?Whatfollow-upquestionsdoyouanticipatethemasking?

?Doyouneedtosplitthisintomultipledashboards?

Theanswerstothesequestionswillgreatlychangethedesignofthedashboard.Understandingtheoverallbusinessproblemandhowthedatacansupportthosemetricsis,firstandforemost,

whatmustbedone.Forabrand-newbusinessanalyst,thisis

tough.Thismaymeansendingoutalotofemailstryingtogetanunderstandingofwhatmetricspeoplecareabout,wherethatdatalives,andwhatdocumentationtoread.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA8

SallySue’sBIcopilotenableshertouploaddocumentationto

createagoverned,customcopilotcapableofansweringspecificquestionsmoreeffectively.Herorganization’sinternalwiki,whichhasimproveddocumentationpracticesovertime,servesasa

primaryresource.However,manyanswerstobusinessquestionsremainburiedindecade-oldemailsandtribalknowledge.

Duringhertestingphase,Sallytakesadvantageofthecontent

inthiswiki.Largelanguagemodelsexcelatparsingand

summarizingvastamountsoftext:themorequalityandaccuratedocumentationsheprovides,thebetterthecopilotwillperform.Sallypoursthroughthewiki,confirmsitsaccuracybycross-

checkingwithotherdepartments,scrapescurateddocumentation,carefullycleansit,andformatsitintoaJSONfilewithinformation

suchasthetitle,sectionname,andtext.ThisJSONisthen

uploadedintotheBItool’scopilotwherethesoftwarehandlestherest.

Customizingthecopilotwiththiscomprehensiveandaccurate

documentationenhancesitsabilitytoanswerbusinessquestions.Forexample,itcannowprovidedetailedexplanationsonlegacyprocessesthatareonlydocumentedinthewiki.Thisimprovementbooststheproductivityofjuniorandnewcolleagues,whousethecopilotforcreatingdashboardswithoutneedingtosiftthrough

extensivehistoricaldata.Notethatachievingthesebenefits

involvesongoingeffortstomaintainandupdatetheAImodeltoreflectcurrentbusinessneedsanddataqualitystandards.

Considerthisexample:

SallySuehasloadedinaknowledgebasethatdescribestheTTCKPI.Throughouttheoffice,everyonetalksaboutTTC–it’sthe

mostimportantmetricforpredictingprofitability.Butwhatdoes

TTCactuallymean?Intheknowledgebase,TTCisdefinedas“TotalTimetoCompletion.”Thelowerthisnumberis,thefasterbeer

canbottledanddistributed.ConsiderascenariowhereajuniorbusinessanalystistaskedwithcreatingadashboardaboutTTC.ThatjunioranalysthasthedatawhichholdsavariablecalledTTC.Theysimplyasktheircopilotthevaguequestion:

“Explaintomewhat’sinthedata.”Thecopilotthenreturns:

Yourdataappearstocontainanumberofbeerproductsproducedovertimeaswellasattributesaboutthoseproducts,suchas

weight,size,andcosttoproduce.TheTotalTimetoCompletion(TTC)isalsoincludedasoneoftheseattributesandisthe

business’smostimportantmetricbecauseitishighlynegatively

correlatedwithrevenue.ThelowerTTCis,themorerevenuetendstoincrease.Herearesomesuggestionsforvisualizingthedata:

Thekeythinghereisthatthecopilot:

?KnowswhatTTCisandcandecodewhattheacronymmeans

?KnowsthatTTCisimportantforrevenue

?Givesinitialvisualizationsuggestions

RatherthanrandomlyaddingvisualsintoaBItool,thecopilotis

givingtheuserastrongstartingpointandhelpsreducetheblankcanvaseffect.SallySueisalltoofamiliarwiththeblankcanvas

effect:it’sthefeelingyougetwhenyou’retaskedwithstartinga

brand-newpresentation,paper,ordashboard.You’represented

withablankcanvas,whichcanbeeitheragreatthingthatinspirescreativity,oraterrifyingthingthatsuccumbsyourbraintothe

dreadedwriter’sblock.AgoodBIcopilotcan,andshould,eliminatethiseffectandgivetheuseragoodstartingpoint.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA9

StartingfromscratchcanbeadifficulttaskforeventhemostseasonedBIveterans

Somequestionstheusermightaskare:

?“GivemesomesuggestedvisualizationsforTTC.”

?“BuildmeastarterdashboardforaCEOwhocaresaboutrevenueasitrelatestoTTC.Includeothermetricsthatmaybeusefultoknow.”

?“Modifymydashboardsothatit’smoreaboutTTCovertimeratherthanTTCasawhole.”

SallySuetriesallthesequestionsandevaluateshowtheBItool

does.Ifit’swell-tuned,itshouldgivestrongstartingvisualizationsandmetrics.Shefindsthatitdoesanokayjobcreatingastarterdashboard.It’snotperfectandsomeoftheKPIsseemabitoff,

butit’scertainlynotbad,either.Thecopilotcoulddowithalittleimprovementfromuserfeedbackandadditionaldocumentation,butshe’llgettothatlater.Thevisualizationsitbuildsinitscurrentstateareatleastgoodforeditingandspurringnewideas–exactlywhatitshouldbedoing.

ProgrammingforNon-Programmers

Most,ifnotall,BItoolshavesomesortofprogrammingor

scriptinglanguagebuiltintothemsothatuniquemetricscanbe

createdonthefly.Thisiscrucialforcreatinghighlycustomized

dashboardsandgeneratingtheneededmetricsdirectlyinthe

toolwithoutthetedioustaskofleavingit,usinganothertool

orlanguage,thenreloadingthedata.SallySueiswell-versedin

programming,butherbusinessusersarenot–infact,she’sluckyiftheyknowSQL.Timeandtimeagainshegetsquestionsfromherusersonhowtocreatesomeofthemostbasiccalculations:True/Falseflags,summationsovertime,summationsbygroups,nestedcalculationsandmore.Sallynoticedthathercopilotincludesa

placetodescribecalculationstogeneratethem.Intrigued,shetriedasimpleprompt:

“AverageTTCbyregion.”

ThecopilotreturnsafewoptionsofaverageTTCgroupedby

region,allvariableswithinthedata.Thecodeitreturnsiswell-

formatted,commented,andevenincludesafewexamplevaluesforverification.Sallyisextremelyhappytoseethis,asitgivesherbusinessusersasignificantlyeasierwaytocreatemetricsand

customcalculations.Shesuspectsthatthiswillgreatlyreducetheamountofquestionsthatshegetsandimprovethespeedand

accuracyofdashboardcreation.

Insights

Picturesareworthathousandwords,andadashboardismade

ofmanyinteractivepictures.Peoplelovedashboardsbecause,

whendoneright,theycanproduceawealthofinformationina

compactspace.Ifyou’reabusyexecutive,youmighthaveaccesstodozensofdashboards.Somedashboardsarelargerthanothers,andsomerequireyoutoclicktoaspecificlocationandhighlight

specificpartsofthedashboardtogettheinsightsyouneed.

Sometimesyoudon’thavethetimenorthepatiencetogothroughthatdashboardtogetwhatyouneed.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA10

Whatifyoucouldsimplyaskforit?

Whileplayingaroundwiththisnewtool,SallySuenoticedthat

hercopilotwasavailablenotjustfordashboardcreators,but

forreportviewersaswell.Shecouldaskquestionsaboutthe

dashboardtofigureoutwhatsheshouldreallyknowin30secondsfromthisseaofinformation.ShebuiltabasicTTCdashboardthatshewouldexpectexecutivestolookat.Itusedcommonmetrics

thatsheensuredwereapartofthedocumentationsheuploaded,andtherewasnocomplexmath.Sheincludedsomevisualslike

timeseriesgraphsofaverageTTCovertime,thecurrentaverageTTC,averageTTCfromthedayprior,andaverageTTCacross

variousregions.Sheaskedthecopilot:

“Givemethe3mostimportantthingsIshouldknowfromthisdashboard.”

Thecopilotreturned:

ThisdashboardappearstodisplaytheaverageTTCmetricovertimeandacrossregions.Thesearethemostimportantthingstoknowfromthisdashboard:

1.AverageTTCis3.5minutesandhasincreasedby30s(+20%)comparedtoyesterday.Thismayaffectrevenue.

2.AverageTTChasbeentrendingdownsinceJan2024,

buttherewasatemporaryspikeonMar23rdwhereTTCincreasedbynearly90s,butthenreturnedtonormalthefollowingday.

3.TheSoutheastregionhadthehighestincreaseinTTConMar23rd,contributingabout75%totheoverallriseinTTConthatday

SallySuethenasked,

“TellmethedriversofTTCfortheSoutheastregion.”Thecopilotreturned:

ThethreemostimportantdriversofTTCfortheSoutheastregionare:

?Bottlesavailable:thismetricappearstoberelatedtothe

numberofbottlesthatareavailableforproductiononadailybasis.Thisnumberrangesfrom0to10,000andispositivelyrelatedtoTTC.

?Totalworkers:thismetricindicatesthenumberofworkersonthewarehouseflooratthetimeandrangesfrom3to24.

?Unittemperature:thismetricindicatesthetemperatureofeachproductionunitandrangesfrom100Fto230F.Unitsover175Fareconsideredoverheating.

Sallywasimpressedwiththeperformanceofthecopilottograbinsightsfromthedashboard,showingthingsthatarebothdirectlyshownwithinthedashboardandthingsthatmaybehidden;

however,thisisjustfromherinitialtesting.Whileitcertainly

lookedconvincing,shestillneedstospendtimeverifyingtheaccuracyoftheseresultswhichsheplansondoinginafocusedtrustandsecuritytest.

Beautification

Sallyknowsalltoowellhoweasyitistodrag-and-droptobuild

dashboards.ModernBItoolsgenerallyhaveanoptimalsetof

colorsandsettingsturnedonforyoubydefault.Thesetendto

workwellandareusuallysetbyUXtoenablepeopletocreate

decent-lookingdashboardswithoutneedingtothinkasmuch

abouttherightcolorsorgraphsettings.WhatSallySuealsoknewishoweasyitistobuildbaddashboards.

Whatisabaddashboard?You’veprobablyencounteredone.Toomanymetrics.Numberseverywhere.Dozensofpages.Somanygraphscrammedintoasinglepagethanitbringsan8Kmonitortoitsknees.Colorsthatmakeyouwanttowatchblack-and-whitemoviesjusttorelaxyoureyeballs.You’vemostcertainlyseenabaddashboard.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA11

Whatisevenhappeninghere?

Nobodygoesoutoftheirwaytobuildabaddashboard–likethatever-growingjunkdrawerinyourkitchen,itjusthappensover

time.Onenewmetrichere.Onenewgraphthere.OnenewpagefortheaccountantoverinBusinessUnit274.Itgrowsandgrows.Themoreeyesareonadashboard,themorelikelyitistogetthisway.GenerativeAIhasthepotentialtocurbthis.

AnygoodcopilotinaBItoolwillhavebeentrainedon

dashboardingbestpractices.AsSallywenttobuilddashboards,shepaidspecialattentiontothegraphsitcreated:

?Didtheymakesense?

?Arethecolorsappropriate?

?Aretheretitleswheretheyshouldbe?

?Diditcreatetheoptimalnumberofpages?

?Diditfollowbestpracticesformetricsonasinglepage?

?Isitaccessible?

AgoodBIcopilotfollowsdashboardingbestpracticesandgivesastrongstartingpoint

Thankfully,hercopilotfollowedallthesebestpracticeswhen

buildingadashboard,andevenhadtheabilitytogivesuggestionsonhowtoimproveexistingdashboards.Itseemsthatthe

designersofthiscopilotthoughtwellaboutthis.

Challenges

Overall,SallywashappywiththeBIcopilot’scapabilities.Hertestsweresimple,butsheneededawideraudiencetoreallytestitout.Assherolledoutteststoherotherbusinessanalysts,shehad

threeissuesinmind:

?Howusefulisthis?

?Canitbetrusted?

?Isitsecure?

Usefulness

ABIcopilotisanoptionalfeature,firstandforemost.Itsgoalistoassistyoutoexploreyourdataandbuilddashboardsfaster.

THEALCHEMYOFINTELLIGENCE:HOWGENERATIVEAICANREVOLUTIONIZEBUSINESSINTELLIGENCEANDANALYTICSINMODERNENTERPRISES|LFAI&DATA12

Sallyknewthat,likeanyotheroptionalfeature,itwillgocompletely

unusedifit’snotactuallyhelpful.Whensherolledoutthecopilottomorebusinessanalysts,sheaskedthemtopaycloseattentiontothefollowingquestion:doesthisfeaturemakebuilding

dashboardsfasterforyou,orisitafrustratinghindrance?

Ifyoufindyourselfgoingbacktothedrag-and-dropmethod,youprobablyfindthecopilottonotbeveryhelpful.Ifyouavoidthecopilotbecauseyoucan’ttrustitsanswers,thenit’snotagreat

copilot.Copilotsshouldbeconsistentintheiranswersandhavebestpracticesbuiltin.Ifit’screatinguselessdashboardsthat

aren’tevengoodforediting,thenthecopilothasfaileditsgoal.

Thecopilotshouldhelpreducetheblankcanvaseffect.Editingis,ingeneral,fasterthanstartingfromablankcanvas.Ifeditingis

harderthandragginganddropping,thenthecopilotisnotagoodfit.

Considercreatingasurveyorevenaworkshopforagroupof

users.Givethemsomesimpledatatoworkwithandaskthemtobuildadashboardoutofitusingthecopilotinalimitedamountoftime.Thedatashouldbeneutralandideallyhasnotbeen

seenbyanybody,butalsoeasytounderstand.Onegreatway

tofinddatalikethisistosearchforopendatasetsfrom

https://

.

Splitthegroupintotwo:onewhichhasaccesstothecopilot,andonewhichdoesn’t.Askthegroupwhodoeshavethecopilottouseittotheiradvantageto

createdashboardsoutofthedata.Aftertimeisup,allthepeopleintheworkshopshouldsendtheirdashboardstoyouforreview.Comparewhichoneswerebetter.Thisissubjective,soconsiderrecruitingotherstovote.

SendasurveyouttothegroupwhohadaccesstothegenerativeBIdashboard.Askthemquestionssuchas:

?Didyouusethenewcopilottobuildyourdashboard?

?Didyoufindithelpful?Ifso,howdidyouuseit?

?Didyougiveupusingitatanypoint?Ifso,why?

?Ifyoudidnotfindithelpful,whatdidyounotlikeaboutit?

?Ifyoudidnotuseit,whynot?

?Didyoutrusttheresultsthatitgaveyou?

?Wereanyresultsinaccurate?Ifso,whatwerethey?

?Wouldyouusethecopilotagaininthefuturetobuilddashboards?

?Onascaleof1to10,howdoyouratethecopilotoverall?

Performinganexerciselikethiscouldhelpidentifytheusefulnessofthecopilotandgiveyourselfsomeobjectivedatathathelpsyoudeterminewhetheryoushouldmoveforwardwithitsadoption.

Trust

Copi