人工智能與研發(fā)革命

FT

AIand

theR&D

revolution

TheFTTechforGrowthForumissupportedby

Foreword

T

hislatestreportfromtheFinancialTimesTechfor

GrowthForumaddressesafundamentalquestion

facingcompaniestoday:howdowefosterinnovationinordertodrivegrowth?

Businessesalreadyknowthatthereturnswhichflowfromusingadvancedtechnologiessuchasartificialintelligence

areconsiderable.Thekeyforleadersisunderstandinghowto

unlockthatpotentialbyoptimisingtheirinvestmentinresearchanddevelopment.

Thisreportdrawsonexperttestimonytooffercompaniesa

strategythatwilltakethemtothecuttingedgeoftechnologyandinnovation.

JonathanDerbyshire

Visittheeditorialhubat

/tech-for-growth-forum

TechforGrowthForumEditorFinancialTimes

AIandthe

R&Drevolution

Howcanthisfast-changingtechnologyhelpprovidersupdateproductsmoreefficientlyandsuccessfullyateachstageofR&D.ByLucyColback

andconsumerproducts,anticipatethatathirdofsales,

worth$30tnoverfiveyears,willcomefromnewproducts.

I

naworldoffast-changingconsumerpreferencesandincreasingchoice,companiesthatwanttostayaheadmustworkcontinuouslytoensurethattheirgoodsandservicessatisfythecustomer.In2023McKinseysaidthatbigindustries,includingautomotive,telecoms

Advancementiskeyandtheleveloffreshfundsflowing

intoresearchanddevelopmentisconsiderable.According

tothelatestUKstatistics,￡71bnwasspentonR&Din2022,ofwhich￡50bncamefromthebusinesssector.IntheUSthefigureisestimatedtobe$886bnwithbusinessaccountingfor$690bn.

Thereturnfromusingadvancedtechnologiesis

considerable.Lookingatthepharmaindustry,Accenture

estimatesthatscaleduseandredesignedworkflowswill

meanthatmedicinescanbebroughttomarketfouryears

faster,earninganextra$2bnforeachsuccessfuldrug.Costsof$2.6bnto$6.7bncouldalsobescythedbyupto45percent.

Withsuchlargesumsatstake,companieswillhaveto

thinkcarefullyaboutwhereandhowtheyspendtheircashtoreachtheconsumer.

Inapreviousreportwestudiedtheroleoftechnologyinmarketingandhowbusinessescanusetechnologytoolstofosterloyaltywithpersonalisedmessagingandtocreateholisticcross-channelstrategies.

Turningtothefrontendoftheproductcycle,thisreportwillexaminehowtechnologycanhelpprovidersupdate

productsmoreefficientlyandsuccessfullyateachstageofR&D.

Thebusinesscaseforsimulationisshifting,withfastertime-to-marketandreducedproductcost

askeyfuturevaluedrivers

Pastvaluedriversofapplyingsimulationtechniques(%)Futurevaluedrivers(%)

Improvedproductperformance

Reductionofproductcost

Reductionofengineeringcost

Fastertime-to-market

Zeroprotoyping

87

73

73

58

67

Fastertime-to-market

Reductionofproductcost

Improvedproductperformance

Reductionofengineeringcost

Zeroprotoyping

73

72

72

70

65

Source:McKinsey

Wearealltechnologycompanies

Mindsetisanimportantfactorininnovation.Sean

Ammirati,professorofentrepreneurshipatCarnegieMellonUniversity,Pennsylvania,saysanentrepreneurialculture

insidetheR&Dfunctionhelpscompaniestoinnovate—eventhelargeones.Teamswithsuchamindsetaremorelikely

tocomeupwithtransformationalratherthanincrementalproductdevelopments.

Ammirati,whohasfoundedseveralmachine-learning

start-ups,saysmanycompaniesdonotmakethenecessaryinvestmentinadjacentandtransformationalinnovation.Heciteda2012paperbyHarvardBusinessReview,whichstated:“Companiesthatallocated70percentofinnovationactivitytocoreinitiatives,20percenttoadjacentonesand10percenttotransformationalonesoutperformedtheir

peers.”

Technologycompaniesshouldinvestmoreinthesetwo

areas,thesameresearchsaid.Giventoday’subiquityof

technology,AmmiratisaysthateverybusinessshouldthinkofitselfasatechnologycompanywhenitdecidesitsR&D

budget.

Identifyyourtarget

Itisimportanttoidentifywhataproductistryingto

achieveandwhoisthetarget—thismightnotalways

beobvious.Pella,theIowawindowanddoormaker,has

designedamechanismthattreatstheinstallerratherthan

thehomeownerasitscustomer.Basedonobservationsandresponses,Pella’snewwindowiseasytoinstallfromthe

insideratherthantheoutsideofaproperty,reducingtherisktoworkerswhenputtingwindowsintotallbuildings.

TargetsareimportantfortheR&Dprocess,too.This

shouldincludedecidingiftheobjectivesincludecuttingthecostofmaterials,thecostofengineering,timetomarket,

orallthree.Havingthisinmindhelpswithsettingkey

performanceindicatorstoassesswhetheraprocessworks.

ConfidenceinthecapabilitiesofAIandML

isdrivingadoption

‘Accuracyofresults’asacriterionfortoolselection(%ofsurveyresults)

AI/ML-basedsimulationoperationalised

mmNoorverylimitedusageofAI/ML-basedsimulation

20

80

9

91

Accuracyofresults

isacriterionfor

toolselection

(n=70)

Accuracyofresults

isnotacriterion

fortoolselection

(n=30)

Source:McKinsey

Theroleofdata

Customerneedsshouldalwaysbetheinspirationforproductdevelopment.Themoredatathatisavailable,theeasier

thesearetodefine.Dataiscriticaltoanytechnologystrategy.Itcan,forinstance,alertcompaniestowhatcustomersare

seekingandwhatisindemand.

Onlinemarketplaceshaveextensiveaccesstodataaboutpurchasinginformationandshoppingsearcheswhichcan

givethemanadvantageoverthevendorswhousetheirsites.

Asweobservedinourreportontheplatformeconomy,thereisvalueinbrandshostingtheirownwebsitesasa

meanstoretainandinterpretcustomerdata.Thiscan

giveinstantinsightsintomodifyingoraddingtowhattheyoffer.ForexamplethebracompanyLivelyintroduced

straplessbrastoitsrangeafteritfoundthatmanywomenweresearchingforthese.TommyJohn,whichmakesmen’sunderwear,foundthatwomenlikeditsproductstooand

itaddedsecond-skin“boyshorts”,teesandpantsaimedatfemales.

Whencollectingorcompilingdata,itisimportantforittobeclean.Thisisespeciallythecasewhendeployingartificialintelligencesystemswhichareinessencestatisticalmodelsthatfeedondata.“Thismeansyourteamneedstocareaboutdata,”Ammiratisays.“Ifyourteamarebeingsloppywith

howtheyenterdatayou’reactuallycausingdownstreamproblems.”

Increasinglyproductscanbedevelopedfora“segmentofone”usingcustomerinformation.SkincarerangessuchasFentyBeautyofferproductstosuitacustomer’sskintone

usingacombinationofimagerecognitionandmachine

learning.Othercompaniesusecustomer-supplieddata

orinsightsgleanedfrompreviouspurchasestosuggestor

customiseproducts.NikeByYouallowscustomerstocreatetheirowntrainersfromapaletteofstyles,coloursand

designs.

Theco-creationofproductsgoesbeyondpersonalisation,Thisiswherecustomersadvisecompanieswhattheywantbyspecifyingmodificationsbeyondthosegiveninadrop-downlist.Ikeaisamongthegroupsthathaveco-createdproductswithcustomersforseveralyears.

AtasummitonAIhostedbyBloomreach,AzitaMartin,

vice-presidentofAIforretailatNvidia,saidthenextchallengeformanycompanieswouldbeinfulfillinghighlypersonalisedorders—essentiallyforamarketofone—quicklyenoughto

satisfytheconsumer.Shesaidthecorrectforecasting,fasterthroughputindistributioncentres(usingcomputervision,roboticsandsimulation)andlast-miledeliverywereareaswhereAIcouldgiveretailersmoreagility.

‘Ifyourteamarebeing

sloppywithhowthey

enterdatayou’re

actuallycausing

downstreamproblems’

Fromconcepttoreality

Peoplecanstillcomeupwithproductideasindependentlyofdata.TechnologycanhelptoacceleratethesetomarketandAIisapowerfultoolinthatprocess.

WhilehumanoversightwillremainasimportantforR&Dasitisforthecreationofmarketingcontent,generative

AIbasedonlargelanguagemodelswillbeinvaluableasa

“conversationstarter”.Presentedwithanideaorevena

vagueconcept,generativeAIcanhelpwithbrainstormingtohelpdevelopaproduct,undertakemarketresearchtofindiftherearesimilaritems,identifyandanalysecompetitorsandgivethoughtsonhowtocreateadifference.AIcanprovide

manyversionsofaproductandsuggestmodificationstosuitanichethatahumandesignermightnothaveconsidered.

Oncetheproductconcepthasbeenhoned,AIcanhelp

todevisemarket-testingstrategiesaswellasaccelerate

producttestinganddesign.Itcancreateandtestiterationsofaproductataspeedfarfasterthanahuman.Itcansuggestmaterialsandsourcingaswellasmanufacturingprocesses.

Thisisparticularlyhelpfulforstart-upsandsmaller

businesses.ThefounderofSkittenz,acompanythatmakesmittencoverstoenlivenskigloves,usedAItoinvestigatewhichmaterialsandmanufacturingprocessesmightbe

suitabletobringtheproducttomarket.

‘Thisisnotsomething

thatyoucanonlytrain

yourexecutiveson.This

issomethingthat

everybodyinthe

organisationneedsto

beawareof’

SuperchargingwithAI

AIisalsoinvaluableforanestablishedR&Dteam,eventhosethatdeploymoretraditionalsystemsandprograms,which

wewillcoverbelow.TheseareincreasinglyaugmentedbyAI,enablingdesignerstodevisemoreoptionsmorequicklythantheunderlyingsoftwarealone.Ammirati,whohasspecialisedinbothcorporateinnovationandAIintheenterprise,saysthatinthepastthreeyearsthesetwoworkstreamshavemerged

andcannolongerbeconsideredseparateareasofexpertise.

BasedonfindingsfromaforumofR&DleadersinJuly

2024,McKinseysaysthatbothanalyticalandgenerative

AIarelikelytohaveabigeffectoninnovationoutcomes.It

estimatesincreasedmarketfitofupto50percent,productperformanceimprovementsof15to60percent,increased

workplaceproductivityofupto50percentandupto40

percentreductionintimetomarket.Simulationsanddeeplearningsurrogatesinparticularincreasetestingcapabilities.Bainresearchmakessimilarfindings,withengineeringhoursreducedbyuptoafifthandcostfallingbybetween5and30percent.

Bothstressthatstakeholders’buy-inisessentialinthe

digitalisationprocess.Thismightbemoreeasilyachievedinaworkforceifleadersareclearthattech,andAIspecifically,

isalongwayfromreplacingpeople.Amongotherreasons,

thisisduetothecontinued,ifdiminished,incidenceof

hallucinations(whenAIcreatesinformationandpresentsitasfact).“HumanplusAI”willbethefuture,especiallyforR&D.

Ammiratisays:“Therightmentalmodelistothinkof

generativeAIasaco-founder,orwhatisoftendescribedas

‘humanintheloopAI’.Sothinkofitmorelikeabrainstormingbuddy,morelikesomeonewhoautomatesfirstdraftsthan

doesalltheworkforyou.”

WhileAIwillnottakeyourjob,hesays,companiesusingthetoolwillbe“incrediblydisruptive”tothosewhoarenot.“Thisisnotsomethingthatyoucanonlytrainyourexecutiveson.

Thisissomethingthateverybodyintheorganisationneedstobeawareof.”Tokeepupwiththecompetition,nevermindstayahead,businessleadersmusteducatetheirworkforceinhowtouseAI.

Thisdoesnotnecessarilymeanteachingprompt

engineering,whichAmmiratibelieveswillbecomeobsoleteasinterfaceschange.IndeedanewinterfaceisalreadyhereintheformofAIagents.TheseLLM-basedAIappscansummonthehelpoftoolsanddataandsoarenotboundbytheconstraintsofthemostrecentupdate.Overtimetheylearnuser

preferences,retainconversationsforcontextandcanaccessanddeployotherprogramsandinformationautonomouslyinresponsetouserrequests,optimisingworkflowsandcreatingsubtasks.

ToomanybusinessesarenotreadytoharnessAIeffectively.TheStateofAI2024,aglobalsurveyof300seniorexecutivesbySearce,themoderntechnologyconsultant,foundthatwhileallUKandUSrespondentswereusingorplanningtouseAI,

only14percentofcompanieshadmanagedtoscalebeyondthepilotstage.

Impedimentsincludedlackoftalent,especiallyforUK

respondents,whichsuggeststhosecompaniesthatarequicktoinvestineducatingtheircurrentworkforcewillhaveanadvantage.Badlymanageddataandpoorlyprioritiseduseswerefurtherbarrierstosuccessfuladoption.

Despitethehurdles,adoptionisnolongeroptional.For

thosethatarestillnotconvinced,Ammiratisayswehave

reacheda“generaltechnologyinflectionpoint”.Thinkofitassimilartowhentheinternetarrived.“Thebusinesseswhosaidno,we’rejustgoingtoavoidthatinternetthing—mostofthem,unfortunately,arenotaround.”

ClassicsimulationuseissignificantlyaheadofAl-andML-basedsimulationuse

Companies’usesofclassicalvsAl/MLsimulationtechniquesbytechnologymaturitylevel(%ofcompanies)

None

Mostlyproofofconceptsorpilots

Useofhighlymaturevirtualengineeringtechnologiesatscale

10x

morecompaniesuseclassicsimulationcomparedtoAl/ML-basedsimulationatahighmaturitylevel

1

1

48

51

34

61

5

2/3

organisationsalreadyuseAI/MLfor

simulations

Classicalsimulation

techniques

(egMBS,FEA,CFD)

AI/ML-basedsimulation techniques(egneuralnetwork,deeplearning)

Source:McKinsey

AutomotivepullsawayastheleadingindustrytoputAlandMLtouse

Al/MLusagebyindustry(%)

Simulationusers

withinautomotiveare1.6xmore

likelytouse

Al/ML-basedsimulationscomparedtomachinery

59

58

50

47

76

Automotive

(n=42)

Aerospace

anddefence

(n=16)

Basicmaterials

(n=27)

Medicaldevices

(n=19)

Machineryand

fabricatedmetals

(n=18)

Source:McKinsey

TechnologiesavailableforR&D

StepsintheR&DprocesssuchastestingandpredictivemodellingincreasinglytakeadvantageofAI,includingmachinelearningandstatisticalmodelling.Theseareenabledatgreaterspeedandscalebytheincreased

processingpowerofcloudcomputing.Whilethisisnotareplacementforexpertoversight,GenAIprovidesan

intuitiveinterfaceforthosewhomaynothavethetechnicaltrainingthatwasonceessentialevenforearlyproduct

development.Thetechnologiesinclude:

MarketResearch

Digitallydistributedmarketsurveys,suchasthoseofferedbySurveyMonkey,GoogleFormsandTypeformtonamejustafew,offerhugereachandenhancedanalytics.

Design

CAD-CAMsoftwaresuchasthatprovidedbyAutodesk,

SiemensandTrimbleisusedfordesigninfieldssuchas

engineering,architectureandmanufacturing.ManyoftheprogramsincorporateAIwhichquicklyprovidesagreater

numberofalternativedesigns.So-calledgenerativedesigndeploysAIalgorithmsderivedbymachinelearning,whichcancomeupwithdesignsoptimisedfordifferentparameterssuchasmaterialeconomiesorstrength.Providedthe

engineerspecifiesminimumrequirementsandconstraintssuchasmanufacturingprocess,loadsandflexibility,the

systemcanoffermanyvariations,someofwhicharelikelytobenovel.

Prototyping

Thistoo“willnowalmostalwaysstartinthedigital

environment”,saysMarkRidley,aformerchieftechnologyofficeroftheFinancialTimeswhoisnowatechnology

consultant.Puredigitalprototyping,hesays,canmassivelyreducecost,whichwillallowmanymoreattempts.Machine-learningmodelsfacilitatedbycloudcapacityhaveenabledmoredigitalexperimentationintheR&Dprocess.Cloud

computingmeanscompaniesnolongerneedbigbudgets

tobuyin-houseserversforcomplexcomputations.“The

technologyofthecloudallowedforreallysignificant

innovation...becauseitprovideddemocratisedtechnology,whichcouldthenbeusedbysmallentrepreneurialteams.”

Ridleyadds:“Thebeautyhereisthatwiththeincreaseincomputationalpower[and]theavailabilityofsoftwaretodothesethings,designerscannowaugmentmoretraditional

modelslikecomputationalfluiddynamicswithML-basedgenerativedesign.”

‘Youwanttokeep

everyfailed

experiment—because

thefailedexperiments

aresourcesoflearning’

Simulations

Theseareincreasinglypopularfortestingmarketing

strategies,suchasdecidingwhichwordingbesttargetsthecustomerandwhichcommunicationsmediumtochoose.

SimulationsuseA/Btesting(comparingdifferentversionsofastrategy).Theyassumethataproductisalreadyviableandthatthesalesperformancewillbeaffectedbymarketingratherthanhowtheproductisdesignedandperforms.

Simulationscanalsobeusedforproductdevelopment,

testingforvariationsinmaterialsanddesign—oftenusing

thesamesoftwareasthatusedtodesigntheproduct.

EngineeringsoftwarefromproviderssuchasAnsysand

Matlabenablesdesignerstorenderobjectsvirtuallyand

simulatesystemsinwhichtotestandanalyseaspectsoftheirbehaviourindifferentenvironments,forinstancetheirfluidandthermaldynamics.Thevalueinsuchsoftwareliesin

beingabletotestwithoutaphysicalprototype.

Understandingmarketreceptivenesstotheproductisthenextstep.Whileadjacenttomarketing,thisislessbasedontestingastrategyandmoredesignedtodetermineviability.Modellingsoftwarecanhelpinareassuchascomparing

pricingandperformanceagainstcompetitors.Itcanalso

examinetheeffectofdifferenteconomicenvironmentsandcompetitoractions.Itiseasiertotestmaterialordesign

behaviourinasimulationthanitistotrytoassesshumanbehaviour,whichissomuchmoreunpredictable.

Whilesimulationscanbeaprocessingburden,augmentedcapacitycancomefromcloudcomputing.

3Dprinting

Oncesimulationsandtestshaverunsuccessfullythereis

stillmorethatcanbedonewithoutnecessarilyinvestingintheactualproduct.Ridleysaysprocessesusedtorequire

drawingsfollowedbyclaymodelling.Digitalisationhas

significantlyloweredthebarrierstoproduction.“Theclaymodelrequiresartistryanddeepexpertise.Andwhatwe’reseeingnowisthatitiseasiertobuypatternsandcheap

technologyofftheshelfonacreditcard.Thatenablesmoreusersandsmallgroupstodotheinnovationthemselves.”

GenAIinterfacingwith3Dprintingmighteventually

allowacreatortodiscusstheattributesitwantsfroma

givenproduct,Ridleysays.Theymightforinstanceaskthecomputertocomeupwithvariantsofamopthatsimplifyaspectssuchascleaningintocorners,rinsingorapplyingdetergent.

Toproduceinjection-mouldedplasticscomponents,

normallyrequiringexpensiveandspecialisedmachinery,

Boschhasdevelopedadvanced3DprintingwithAI

algorithms.Thesecanadjustsubstrateinputs,heatand

pressureinrealtime,ensuringthatprototypequalityisasgoodastheendproduct.Thisalsoallowsforsmallbatch

productionbeforeinvestmentinlarge-scalefacilities.

SeparatelyBoschhasdevelopedceramic3Dprintingwhichmodelsthedifferentshrinkagesexperiencedwhenanobjectiskiln-fired.

Digitaltwinning

Adigitaltwinisavirtualmodelofaplannedoractualreal

worldproductorprocess.Itisastepupfromastraightdigitalsimulationasitcanoptimiseintegration,testing,monitoringandmaintenanceoffacilitiessuchassupplychainsorpowerplants.Atwinisparticularlyusefulinworkingoutlifecyclemanagement,modellingintheoryhowaproductmight

behaveovertime,updatingitselfasnecessary.Thetwinemulates,replicates,observesandevaluatesaproductorprocessandhighlightsopportunitiesforchange.

Demandisgrowing.FortuneBusinessInsightsestimatesthatdigitaltwinningwillpropelan82percentincreaseinthemarketforcomputer-assisteddesignandproductlifecyclemanagementsoftware,helpingittoreach$26.4bnby2030.NorthAmericaisexpectedtoaccountforone-thirdofthe

overallmarket.

Whilestillearlydays,EYhighlightstheindustrial

metaverseasthenextphaseindigitaltwintechnology,withanassetrendereddigitallyinvirtualspace.Siemens,aproviderofdigitaltwintechnology,partneredin2022withNvidiato

createanindustrialmetaverse,allowinguserstoseetheircreationsinanimmersivedigitaluniverse.

Collaborationsoftware

Communicationandcollaborationacrossanorganisationbetweenmembersofdifferentunitsisfareasierwith

technology.Atthemostbasiclevel,Slackallowsfordiscussionbetweencolleagues.Thereareseveral

morespecialisedprogrammesforinnovationandidea

management,suchasMiro,BraineetandIdeanote.Itis

worthnotingthatwhiletheseprogramshelphumansto

shareknowledgeandadvanceideas,theydonotgenerateordevelopconceptsthemselves.

Thistypeofknowledgebankingisstillhelpfulifit

allowscompaniestofindhistoricaldataeasily.“Thekeytoinnovationisnotsuccess,”saysRidley.

“Thekeyisthatyoufailveryquicklyandyoulearnfromeveryoneofthoseexperiments.Youwanttokeepevery

failedexperiment—becausethefailedexperimentsaresourcesoflearning.”

Italsohelpstokeeparecordofhowandwhychangesweremadetoproductstoavoidspuriousreversals—andtobe

surethatthedataisaccurate.

Finally,AIcanhelpwithcheckingthatexistingpatentsorregulationsarenotbeinginfringed.

Talentandtruststandinthewayofgreateradoptionofsimulation

KeyimpedimentsforbroadadoptionofAI/ML-basedsimulationversusclassicsimulation(%ofsurveyresponses)

61

47

44

40

36

ClassicAI/ML-based

61

LimitedtrustinAI/ML-basedmethodologies

Lackoftalent/skillLimitedunderstanding

ofthebenefitsof

AI/ML-basedsimulation

3/5

oforganisations

consideralackof

talentandalimitedunderstandingof

thebenefitsasa

criticalimpedimenttowardsadoptionofAI/ML-based

simulation

Source:McKinsey

What’snext?

EvenasAImakesstridesinboostingcapacitytocreate

moreproductsathigherspeed,quantumtechnologiescouldaddnewdimensions.TheUKhasinvestedinthisareafor

adecadeandinJulythegovernmentannouncedfundingof￡100mntoestablishquantumhubstopushforwarddevelopment.

Wearestillsomewayfromunderstandingthefull

potentialofquantumcomputingbutitcouldhelpwith

materialanddrugdiscoveries,modellingexistingmaterialsandchemicalprocesses.

Quantumsensors,whichcanbefarmoresensitivethan

regularsensors,arealreadyusedinsomecommercial

applications.Theseincludebrain-scanning,wherequantum

sensorsallowforawiderrangeofdiagnosticenvironments,andgravitysensing,whichcanbeusedtoscanforsubsurfacecomposition,forinstanceintheconstructionindustry.

Electricbatteryresearchershavealreadydeployedquantumsensorstoanalysemicrocurrentsandimproveproduction

yields.

Therearesomeobstructionstowider-scaleapplication,includingthehypersensitivityofquantumsensors,

whichmakesthemdifficulttouseinsomeenvironments,

theirdemandforpowerandtheircurrentcomplexity.

Neverthelesswecouldseedevelopmentsthatallowfor

morewidespreadcommercialusagewithintwotofiveyears,dependingontheapplication.

PremiumMembers

Beyondtheproduct:Whylearningtolisteniskeytoinnovation

DilipBhatia,chiefexperienceofficer,Lenovo

TheFTTechforGrowthForumissupportedby

HCLTechandLenovo,ourpremiummembers,who

helptofundthereports.

Ourmemberssharetheirbusinessperspectiveon

theforumadvisoryboard.Theydiscusstopicsthat

theforumshouldcoverbutthefinaldecisionrests

withtheeditorialdirector.Thereportsarewritten

byaFinancialTimesjournalistandareeditorially

independent.

Members’viewsstandalone.Theyareseparatefrom

theFTandtheFTTechforGrowthForum.

Howcanbusinessesgetbetteratproductinnovation?Itsoundsparadoxical,butthefirstandmostimportantstepistostop

thinkingsomuchaboutproductsandtechnology.Instead,

companiesshouldfocusmoreonunderstandingeverything

aboutthecustomerexperience—andgettingthatrightmeanslearninghowtolisten.

AtLenovo,welistentofeedbackfromacrossthecustomer

journey—throughadvisorycouncils,customerpanelsand

millionsofproduct-ownershipsurveys—andusethistopowerourinnovation.Ourapproachisagile:wezeroinonspecific

groupstoimprovetheirindividualexperienceofourproducts,ratherthancarryingoutmassive,costlyprojectsthatcantakeyearstocomplete.

Whenprofessionalstoldustheylovedthe“sleeknessand

sturdiness”oftheThinkPad,welistenedandbuiltonit.To

dev