Business Horizons (2019) 62, 819e829

Available online at www.sciencedirect.com

ScienceDirectwww.journals.elsevier.com/business-hor izons

How intelligent is Watson? Enabling digitaltransformation through artificialintelligence

Stefano Magistretti a, Claudio Dell’Era a,*,Antonio Messeni Petruzzelli b

a School of Management Politecnico di Milano, Piazza L. da Vinci, 32 20133 Milano, Italyb Politecnico di Bari, Viale Japigia, 186, 70122 Bari, Italy

KEYWORDSGeneral-

*

Mam

ht00

purposetechnologies;Crafting digitaltechnology;Digitaltransformation;IBM Watson;Artificialintelligence;Digital productdesign

Corresponding authorE-mail addresses: stefano.mgistretti), claudio.dellera@polimi.iessenipetruzzelli@poliba.it (A. Mess

tps://doi.org/10.1016/j.bushor.20107-6813/ª 2019 Kelley School of Bu

Abstract Due to its intrinsic characteristics, artificial intelligence (AI) can beconsidered a general-purpose technology (GPT) in the digital era. Most studies inthe field focus on the ex-post recognition and classification of GPT but in thisarticle, we look at a GPT design ex-ante by reviewing the extreme and inspiringexample of IBM’s Watson. Our objective is to shed light on how companies cancreate value through AI. In particular, our longitudinal case study highlights thestrategic decisions IBM took to create value in two dimensions: internal develop-ment and external collaborations. We offer relevant implications for practitionersand academics eager to know more about AI in the digital world.ª 2019 Kelley School of Business, Indiana University. Published by Elsevier Inc. Allrights reserved.

1. AI transforms society

We live in a world in which digital solutions arewidespread, even if the distinction between digitaland nondigital technologies is somewhat complex.Technologies, and especially digital technologies,influence our behaviors, our way of living, and areadopted differently around the world (Williams &

agistretti@polimi.it (S.t (C. Dell’Era), antonio.eni Petruzzelli)

9.08.004siness, Indiana University. Pub

Edge, 1996). Scholars and practitioners definethis phenomenon as digital transformation, con-sisting of all the initiatives that leverage digitaltoolsdsuch as 3-D printing, mobile computing, andartificial intelligence (AI)dto transform processesand organizations (Nambisan, Lyytinen, Majchrzak,& Song, 2017).

Two main dimensions influence the world ofdigital technologies. The first concerns input. Dueto their constant evolution and adaptation, theprocesses to develop and sustain digital technolo-gies require the involvement of a dynamic set ofactors (Boudreau & Lakhani, 2013) whose different

lished by Elsevier Inc. All rights reserved.

820 S. Magistretti et al.

goals and capabilities yield diverse solutions. Thesecond concerns outcome. As digital technologiesincrease flexibility in creating end products andservices, the outcomes are intentionally incom-plete, allowing companies to continuously upgradetheir offering even if the solution is already on themarket (Garud, Jain, & Tuertscher, 2008).

1.1. Digital technologies redesign thecompetitive arena

The abundant scholarly debate around the defini-tion of technology (Danneels & Frattini, 2018) hasled to difficulties in properly framing digital tech-nologies. Digital technologies include big data(Kaplan & Haenlein, 2010), information knowledgemanagement systems, cloud computing, AI (e.g.,machine learning), and rapid prototyping systems(Bughin, McCarthy, & Chui, 2017; Urbinati,Chiaroni, Chiesa, & Frattini, 2018), all of whichfoster the digital transformation of companies andsociety. Big data, for example, not only reshapesthe way new products and services are designed(Erevelles, Fukawa, & Swayne, 2016) but also hasimproved the healthcare industry by providingmore insights and information than ever before(O’Donovan, Leahy, Bruton, & O’Sullivan, 2015).Furthermore, machine learning profoundlychanged the way we create and envision newproducts. An example is IBM’s Watson Chef initia-tive. Launched in 2016, this digital technology al-lows the screening and mapping of more than10,000 recipes to study the ingredients and pro-pose new combinations.

Digital technologies can change the way weinnovate and deliver value. People are not alwaysable to perceive the value in all proposed finalsolutions. Nokia Lumia is just one example of howproviding too many options is not always positive.The real value digital technologies generate is notalways found in their essence or intended functionbut it instead can be found in the actions theyenable. Netflix does not require big data to un-derstand customer behaviors and create new TVseries, but what is important is the outcome fromthe customers’ perspective (i.e., new series morealigned with their expectations). Indeed, whenfaced with too many options for the same solution,decisions are difficult to make (Verganti, 2017).

1.2. Unlocking opportunities hidden indigital technologies

In an overcrowded and fluid world in which thevalue of technology is more in the benefit gener-ated for adopters than in the technology itself,

discovering opportunities hidden in digital tech-nologies is crucial (Chesbrough, 2003). Accord-ingly, the perspective suppliers might adopt istwofold. They can develop a unique technologythat unveils a new meaning and can be adapted toexplore a unique application field (Dell’Era,Altuna, Magistretti, & Verganti, 2017; Magistretti& Dell’Era, 2018) or they can leverage theessence of the digital technology by understandingits intrinsic generalizability. Indeed, digital tech-nologies are inherently general-purpose technol-ogies (GPTs) because of their substantial andpervasive effect on society as a whole (Youtie,Iacopetta, & Graham, 2008). Researchers tend tofocus on studying and recognizing complex andrelevant technologies ex-post. Numerous studieshave been published on the underpinning elementsof GPT (Gambardella & Giarratana, 2013) but onlyfew attempt to understand and conceptualize theprocess companies should follow to bring suchtechnologies to market (Gambardella & McGahan,2010). The development of a new technologymay be scattered and not always linear. The openinnovation paradigm (Chesbrough, 2006) affirmsthat this process has evolved over the last decade,thus calling for further research to unveil itsdynamics.

1.3. Digging deeper into the discovery anddevelopment of AI

This study aims to extend current knowledge on AIas an enabler of value creation in the digital world(Kaplan & Haenlein, 2019). We pose this researchquestion: What managerial practices will helpcompanies discover the opportunities AI offersduring its development and integration in themarket? In looking for the answer, we adopted acase study methodology. We looked at Watson, thecognitive system launched by IBM in 2011. Resultsof our investigation are relevant for practitionersand researchers alike, providing some managerialimplications on how to unveil opportunities hiddenin technologies and develop AI as a GPT that canbetter support digital transformation. Our studyprovides unique insights on how to design AI as aGPT, going beyond the existing debate that is morefocused on ex-post recognition.

2. AI reshapes product development

The literature on technology management is vastand, considering our focus on AI, we pay particularattention to its intrinsic link with GPT literature.As mentioned, studies on the proliferation of

Enabling digital transformation through artificial intelligence 821

digital technologies and their link to value creationindicate that the digital era is profoundly reshap-ing the new product and new technology devel-opment fields (Dobusch & Kapeller, 2018).

2.1. AI enables digital transformation

Considering the impact of digital technologies intoday’s society, some scholars (e.g., Nambisanet al., 2017) argue that they favor more dynamicrelations between the people involved and theproduct outcomes. Moreover, research indicatesthat innovations that leverage digital technologiescan cut across industry and sector boundaries. Assuch, the digital technology and GPT research fieldscan benefit from each other. Several studies focuson digital technologies by mapping the specific fieldof adoption (e.g., Urbinati et al., 2018; Yoo,Boland, Lyytinen, & Majchrzak, 2012) while othersattempt to understand specific digital technologiesand their influence on new strategies and in-novations (Wieland, Hartmann, & Vargo, 2017).

Technologies require the involvement ofdifferent people and roles in the innovation pro-cess (Bharadwaj & Noble, 2017). The vast amountof data allows different paths for designing newsolutions and dealing with management problems(George, Haas, & Pentland, 2014). Other digitaltechnologies instead change the way decisions aremade to support and complement existing pro-cesses (e.g., 3-D printing is complementary toadditive manufacturing). Finally, digital technol-ogy researchers point out the growing role of AI asa complementary strategy and a new way ofmaking decisions. The volume of informationavailable in online repositories and fast access todatabases can increase the speed and efficiency ofthe decision-making process exponentially (VanKnippenberg, Dahlander, Haas, & George, 2015).Thus, AI is changing business models in severalindustries and reshaping markets (Nenonen &Storbacka, 2018). Examples of AI adoption in verydifferent contexts include healthcare in India orthe public library system in New York, which sug-gests new readings based on a customer’s history.

2.2. Digital technologies profoundly changenew product development processes

Our analysis of the literature related to the tech-nology field identified several studies that dealwith the creation of new products and services andvalue delivery from three different perspectives:

1. Some consider the evolution and developmentprocesses (e.g., Cuhls, 2003; Savino, Messeni

Petruzzelli, & Albino, 2017; Thomke &Reinertsen, 2012);

2. Others look at unveiling opportunities (e.g.,Verganti, 2011; Dell’Era et al., 2017); and

3. Some focus on particular technologies to shednew light on unusual aspects, such as uncer-tainty or the possibility of recombining knowl-edge to generate new technologies (e.g.,Cohen & Levinthal, 1989; Levinthal, 1998).

Given the focus of our investigation, the secondperspective (i.e., unveiling opportunities) is mostrelevant. Scholars point out that GPTs are coretechnologies that generate and spread incremen-tal or radical innovations in different fields, ac-tivities, and sectors. Some define GPTs astechnologies characterized by pervasiveness,inherent potential for technical improvements,and innovative complementarities (Bresnahan &Trajtenberg, 1995).

Numerous studies discuss the role of technologysubstitution and how firms typically develop a newversion or a new technology just to increase theperformance of an existing technology (Smith &Reinertsen, 1992) instead of searching for new op-portunities and identifying new application fields(Dell’Era et al., 2017). An initial slowdown of GPTsis caused by the rapid depreciation of skills that arespecific to older technologies (Helpman & Rangel,1999). The prolonged slowdown of GPT in theearly phases usually is followed by fast acceleration(Helpman, 1998). GPT commercialization anddiffusion is not always linear and straightforward(Ardito, Petruzzelli, & Albino, 2016) due to itssubstantial and pervasive effect on society as awhole (Youtie et al., 2008). Scholars indicate that,in the past, GPT was traditionally developed andintegrated via licensing (Gambardella & Giarratana,2013). Therefore, the ex-post recognition of howcompanies can effectively exploit AI after adoptionis no longer relevant and it becomes important toenhance ex-ante development knowledge. Little isknown about how tech suppliers can develop suchtechnologies to foster their pervasiveness andadopt them in order to create value. As scholarsand practitioners seek to understand how to createvalue from GPTs beyond licensing, managementpractices concerning knowledge search andrecombination, as well as initiatives they canimplement to foster the development andcommercialization of AI as a GPT, become vital.

This literature review shows that research ondigital technology is evolving. Following this line ofinquiry, we investigate the methods companies

822 S. Magistretti et al.

can put in place to develop and commercialize adigital GPT such as AI to create value.

3. As a GPT, IBM Watson offers infiniteopportunities

Considering the multifaceted aspect of this topic,the case study methodologydwhich is particularlysuited to answering ‘how’ questions and investi-gating complex phenomena (Yin, 2011)dis themost appropriate to explore our research problem.Considering the limitation of the methodology interms of generalizability (Harrison & Kjellberg,2010), the selection of the case is crucial.Following prior studies (Yin, 2011), we adopttheoretical and convenience sampling. In partic-ular, considering GPT and its commercialization,and relying on the theoretical definition of GPT,we wanted to study a digital technology case thatis facing an integration moment: AI. AI respects allthe previously defined features (Bresnahan &Trajtenberg, 1995) of pervasiveness and innova-tion spawning. Given the convenience samplingand the difficulties in collecting sufficient data onsuch an emerging and growing technology, weselected a globally famous case with a consider-able number of related articles and records thatproved an excellent start to data collection. Inaddition, due to our connection with the company,we were able to interview key informants in thedevelopment of this sophisticated technology.Finally, given the complexity of the phenomenon,we selected an extreme case that could simulta-neously shed light on the inputs, outcomes, and

Table 1. Data gathering

Sources Description

Interviews (semi-structured)

First wave with an exploratory persSecond wave with the aim of trianginformation to pursue a longitudinal

Bluemix Platform Bluemix developers can use IBM sercreate, manage, run, and deploy varapplications for the public cloud, aslocal or on-premise environments.

Videos Records of events regarding WatsonWorld of Watson event that summarinitiatives related to the technologypresented live by the CEO Ginny Ro

News From websites: New York Times, ThFortune, and Wired, public press resectorial studies, Gartner, Businessreports, and company websites.

organizational changes (Pettigrew, 1990) to un-derstand the longitudinal perspective of thedevelopment and commercialization of a GPT inthe digital era.

These considerations led to selecting IBM Wat-son as an eye-opening case of an emerging GPT.Watson meets the aforementioned sampling re-quirements, both theoretically and empirically.Table 1 summarizes all the main sources of datagathered.

We analyzed Watson with an inductive anditerative approach (Strauss & Corbin, 1998). Thisallowed us to shed light on the phenomenon bytheorizing the evidence emerging from the longi-tudinal case study (Eisenhardt, 1989). First, wecreated the database by tagging different data andpositioning them in a longitudinal perspective inorder to understand the technology’s developmentand commercialization. Then, each author indi-vidually read through the information to startanalyzing the data. We created different visuali-zation schemas, charts, and tables to allowcomparing the different information and datasources. Finally, we internally discussed theinterpretation of the data and how the data couldsuggest ways to commercialize and unveil the op-portunities within the AI technology.

3.1. IBM Watson is everywhere: From 0 to270 applications in 6 years

The origin of Watson dates back to May 1997 when agroup of researchers started to develop Deep Blue,an AI solution that can solve complex problems,adapt to the context, and analyze new information.

Details

pective.ulatingperspective.

� 6 in-depth interviews

� 3 managers

� 16 hours of recorded material

vices toious types ofwell as for

� 25 solutions

� 3 different models: public, dedicated,and hybrid

such as theizes theand

metty.

� > 100 videos

e Economist,leases,Insider

� 300 websites

� 38% IBM press releases

� 35% IBM web pages

� 27% external sources

Figure 1. The evolution of the IBM Watson application fields over time

Enabling digital transformation through artificial intelligence 823

Since Deep Blue, the development of the AI solutiongrew within the company and led IBM to introducethe first version of Watson on Jeopardy! in 2011,where the AI solution revealed its potential to theworld. In particular, it demonstrated how themachine-learning algorithm can search a massivedatabase to find answers to common questions. Theoutcome was Watson’s victory over human partici-pants and this event confirmed that having accessto large amounts of data, such as preloaded docu-ments, and the ability to make fast inquiries couldimprove the scouting of relevant information. Italso showed the market the initial potential of thistechnology. Wherever these two aspects are rele-vant, the technology can play a crucial role in thefuture decision-making process.

Moreover, one interviewee defined the IBMWatson technology as a combination of differenttechnologies. In particular, the two most impor-tant are natural language processing (NLP) andmachine learning. NLP is the process throughwhich deconstructing and reconstructing senten-ces allows a computer to understand and processthe information captured by the speakers. Ma-chine learning is a technique in which a machinespots connections between a particular patternand the most likely outcomes. The system re-ceives feedback from regular use, learning fromevery interaction. Every prediction it makes,whether right or wrong, is considered for the nextprediction until it can reliably spot meaningfulpatterns throughout the massive amount of data

it searches. The combination of these two tech-nologies gave rise to the creation of AI and, ulti-mately, IBM Watson.

After 2011, IBM started to significantly invest inthe research and development of this technology,which then was studied by a few employees to learnits potential. Today, more than 12 Watson researchcenters house more than 100 IBM researchers in sixcontinents. In these facilities, employees work ondeveloping and advancing the AI technology andsupporting firms in adopting this solution.

The technology evolved year after year and nowaccounts for over 270 different applications acrossmore than 30 industries. This clearly shows thepervasiveness of AI and its innovativenessdcriticalelements of any GPT. Figure 1 shows the evolutionof IBM Watson application fields since 2011.

Figure 1 illustrates that, at the beginning, thefocus was on a few application fields such ashealthcare and banks. As demonstrated in ourdatabase analysisdwhich we created bycombining more than 300 web sourcesdtheexperimentation in healthcare and banks in 2011and 2012 allowed researchers to understand thetechnology’s potential and that Watson wascapable of processing a vast amount of data andpropose accurate answers. From 2013 onward, thecompany started suggesting the technology’sadoption in diverse application fields, exploringdifferent opportunities including fast imagerecognition, sound detection, and text mining. Theprocess IBM adopted to nurture this expansion and

824 S. Magistretti et al.

exploration followed different initiatives; somewere internal while others had external partners.

Spreading the solution over different applica-tion fields allowed the company to achieve $18billion in revenues in 2017 from Watson cognitivesolutions (Source IBM Annual Report). In otherwords, Watson covered around 22% of IBM’s entirerevenues and in 2017 there were more than 1,400patents awarded for AI solutions. It is clear fromthese numbers and the description that the AI so-lution is a GPT, an extreme case that can providepowerful insights on how opportunities hiddenwithin a technology can be unveiled to createpervasive digital solutions.

4. How can you create value through AI?Elementary, my dear Watson

Our aim in studying the evolution of the Watsontechnology is to depict a path that companiesseeking to develop and create value through AI inthe digital world can follow. Our key informantsstressed that knowledge of the opportunities thetechnology offers was not explicit from thebeginning. They indicated that the technology wasable to grow and expand into different markets byleveraging two aspects: (1) its ability to processand analyze a large amount of data and (2) thecreation of a platform and an application pro-gramming interface (API) ecosystem able to sharethe technology more efficiently. Initially, the focuswas more on internal development while, in lateryears, they diversified the focus to create anetwork of external collaborations with differentplayers.

4.1. First, understand what you are able todo

Our analysis of documentary data and the in-terviews indicates that internal development wasstrategically conducted around two main initia-tives: the data and the API. In particular, the useof data is key in the digital world and big data arestarting to show their potential in terms ofenhanced decision making and better designedproducts and services. IBM understood this impor-tance and was convinced that AI would show itsgreater potential when a considerable amount ofdata was included, speeding up the search foruseful information 60-fold. As the following quotesindicate, this knowledge allowed the company tostart developing solutions for an industry in whichthe data repository is enormous and the time for

searching is not enough. Strategically, healthcarebecame the primary industry for the technology:

Watson is instrumentally useful when lots ofdata are available. So [we looked to] marketswhere big data are predominant and wherethis data can provide value to differentstakeholders. For example, in healthcarewhere many studies are published, and doc-tors do not have time to read all the articles,Watson is amazing. He can read and process 4billion pages in a minute, so then he cansupport doctors in understanding whichstudies are relevant for the disease they arefacing. (IBM Human-Centric Practice Manager)

Thus, the company leveraged some crucial aspectsof the technology to better explore the opportu-nities it offers. By understanding the most essen-tial elements of the technology, they started toadopt it in sectors in which these elementsdbigdata, different stakeholders, and time pressure toprocess a great deal of informationdare present.The process of understanding the technologyinitially was performed internally, where IBM re-searchers started to frame the technology’sessential features and components to understandthe opportunities.

The second element concerning internal devel-opment was leveraging the API. IBM employeesrealized that an exclusively internal team couldnot develop a multifaceted and powerful tech-nology such as AI. They decided to build thetechnology’s core to allow others to develop so-lutions by exploiting the API. This strategic deci-sion was taken to support the technology’s growth.The technological feature the researchers intro-duced through the API led to the protocol to buildsoftware programs by adding features and notcoding every single end solution from scratch.Indeed, the underlying assumption is that peoplecan create a software solution by adding differentcomponents developed by others and called ondemand by the software. To support this, the firmlaunched the Bluemix platform to allow differentadopters to leverage this API repository:

Bluemix can be seen as an API marketplacewhere different companies can buy differentfeatures that, through Watson, can allowthem to perform the work they need. I sharewith you just a quick example. Inside Blue-mix, we have an API for image recognition.This API is useful for Watson developers whenthe input of the information is not a voice ora written document but a set of images. Thisis valuable in the medical industry for

Enabling digital transformation through artificial intelligence 825

analyzing radiographs or in marketing forrecognizing products. If you are in a hospital,you can use the same API as a fashion com-pany, reducing the amount of work needed tohave a working solution. (IBM ResearchEcosystem Manager)

The intuition to leverage the API economy to sus-tain the technology’s development was a way torapidly and pervasively spread the use of Watson indifferent applications. In this sense, players indifferent markets could use the same piece ofcode to add the same features to their system andspeed up the technology’s adoption:

The cloud platform and the API enableeveryone to adopt a do-it-yourself approachwhile planning and developing a new appli-cation or a new concept. In this direction,IBM can help them by providing software andnot always by coding it on demand. This helpsthe growth of Watson by facilitating itsspread. (IBM Cognitive Solutions Leader)

These two strategic decisions exploiting the bigdata explosion and opening the API led to sus-taining the technology’s development andcommercialization. IBM was able to propose Wat-son as a GPT due not only to its potentiality butalso because of the strategic decisions taken.Indeed, the decision to open the API is counterin-tuitive for a core business due to complexity anddata protection issues. All the different solutionsuse the same technology and the same machine-learning algorithm, even if owned by differentplayers. Notwithstanding this, IBM considered thisas the only way to create a pervasive digitalsolution.

4.2. Second, do not be shy: Co-develop withothers

In addition to internal development that allowedWatson to become pervasive and spread its inno-vativeness, IBM made other strategic decisions tosupport its commercialization. In particular, ourdatabase analysis and interviews showed thatexternal collaboration played an essential role infostering the development and commercializationof AI technology.

Four external collaboration initiatives emergedfrom our analysis: (1) research activities withthird parties; (2) partnerships; (3) universityprograms; and (4) hackathons. In particular,research activities refer to all the initiatives IBMput in place to find new application fields for thetechnology based on researching and

experimenting the opportunities technology of-fers through strong interactions with key playersin the field. In the early stage of AI development,due to the complexity and cost, research wasundertaken with a few select players, such ashospitals or premium banking institutions, willingto support the research as they could alreadyenvision an opportunity. Driving this initiative wasthe IBM’s uncertainty regarding the effectivenessof the final solution. Thus, joint research withother players seemed the best option. Partner-ships refer to commercial collaborations withcompanies to develop Watson-based applications.In this case, the initiative was driven mainly bythe exchange of money instead of knowledge.Indeed, the partners were involved as benefi-ciaries of the final solutions but, given the natureof the technology and the need to explore itsopportunities, higher involvement was required.A partnership was formed, for example, with aninsurance company to better manage roadaccidents.

Partnerships guarantee mutual involvement inexploring opportunities and higher involvement ofadopters. This was fundamental in fostering thediscovery of more opportunities and thus envi-sioning future applications. University programsare initiatives to find new uses and deepen thesystem capabilities by collaborating with expertacademic researchers in the AI field. This initiativebrought IBM to connect with top universitiesaround the world with the intention of bringingdifferent and more theoretical knowledge to thedevelopment. Indeed, it was evident to the man-agers that the practitioners’ perspective alone wasnot enough to spread AI around the world. Finally,hackathons refer to external calls aimed at start-ups, students, and citizens to unveil potential newusers of the Watson technology. These externalcollaboration initiatives led to the explosion of theadoption of IBM AI solutions in literally every in-dustry. There were two reasons behind the deci-sion to organize hackathons: (1) to envision newopportunities and solutions (2) to communicatethe technology’s existence and power to theworld.

Figure 2 shows that the first 2 years of experi-mentation were dedicated mainly to research ac-tivities. The IBM research team, together with theresearch centers, developed AI applications for thehealth and banking industries. Between 2013 and2014, other initiatives and activities were adoptedto expand the technology’s potential. Thus, IBMlaunched partnerships, university programs, andhackathons mainly targeting external players toinspire the technology’s future development.

Figure 2. Different initiatives that supported theevolution of IBM Watson over time

Figure 3. Different strategies that can support com-panies in creating value through AI

826 S. Magistretti et al.

5. AI creates value in the digital world

This longitudinal qualitative analysis sheds light onthe way a company can adopt AI to create value intoday’s digital world. In particular, it shows therelationship between the intrinsic characteristicsof AI technology and how these influence thestrategic decisions made to sustain innovativenessand pervasive commercialization. Specifically, ourqualitative study identifies new and thus far un-explored effects a company’s commercializationof a digital technology with the aim of crafting it asa GPT. In particular, Figure 3 shows the two per-spectives companies can adopt to create valuewith AI. First, companies look inside the technol-ogy to unveil hidden opportunities and its coreessence, which is usually an internal activity.Second, they develop a strategy to create newvalue by putting in place interactions with externalcollaborators in order to increase the technology’spervasiveness of adoption and value generated.

Our study enables better understanding from anex-ante perspective on how companies canexperiment and explore AI technology to sustainits pervasiveness as a GPT. Indeed, through digitaltechnology, commercialization in the market isoverturned. It is not only a matter of developingthe technology internally and then licensing it(Gambardella & McGahan, 2010) but also devel-oping the technology with users through internalcultivation and external collaborations.

Our investigation also shows that the digitalinnovation process follows a different pathcompared to the traditional technology develop-ment view (Cooper, 2008). It is no longer a funnelprocess in which the technology is developed for asingle use (e.g., railway technology), but rather anopportunity to expand the boundaries to severaldifferent applications. This shift in perspectivefrom market to technology influences the way the

technology is designed and commercialized. Forexample, in order to create value through theadoption of AI, the Watson case shows us thatunderstanding the core essence of the technologyis crucial (Danneels & Frattini, 2018). Indeed, thestrategic decision to leverage big data and the APIto allow AI to become pervasive was key for thesuccess of the technology itself. This was onlypossible because the company started to look in-side the technology in search of its capabilities anddid not focus solely on understanding user needs.This is a perspective shared in designing mean-ingful innovations (Verganti, 2017) but it is stillunderexplored on the technological side(Magistretti & Dell’Era, 2018). IBM’s intuition tolook inside the technology first before leveragingexternal collaborations to commercialize it is oneof the most important findings of our research.

Our qualitative research aims to shed light ontechnology development, especially in the designprocess of a GPT. In this sense, experimentation indifferent application fieldsdmore than 32 in a 6-year perioddallowed IBM to create a very power-ful technology and spread it to so many countriesand markets, thereby matching all the character-istics of a GPT. This is very important for moreresearch in this field. Current GPT studies helppractitioners and academics recognize and label atechnology as a GPT (Gambardella & Giarratana,2013). This article shares insights on how com-panies in the digital era can exploit different as-pects to increase the probability of creating a GPTby analyzing GPT literature in a completelydifferent way. Indeed, by adopting differentresearch, university programs, partnerships, andhackathon initiatives, IBM was able to address 32different markets and reach a very high level of

Table 2. Managerial contributions to steer the development of globally impactful digital technology

Enabling digital transformation through artificial intelligence 827

pervasiveness, which is the first crucial element ofa GPT. Out of the 270 applications created basedon IBM AI, some were downstream (e.g., imagerecognition for a fashion player) while others weremore upstream and strategic (e.g., helping acompany better frame the machine learning algo-rithm). The use and exploitation of API also per-formed extremely well on the improvementdimension, the second crucial factor of a GPT.Indeed, IBM’s strategic decision to open the tech-nology’s development to external players by lett-ing them use the API was counterintuitive for aGPT for which firms usually target technologypatenting and licensing in an attempt to preservetheir knowledge from being exploited externally(Youtie et al., 2008). Finally, through the Bluemixsolution, the cloud platform where different ap-plications for the Watson technology are madeavailable to users, IBM also pursued theinnovation-spawning dimension of GPT.

Watson embraces the three elements of perva-siveness, improvement, and innovation spawningto sustain the growth and experimentation of thistechnology. As evident from the data, the initia-tives IBM adopteddincluding different activities toexperiment the opportunities offered by thetechnology during commercialization in differentapplicationsdallowed the company to enhancethe improvement and innovation spawning effects

(Bresnahan & Trajtenberg, 1995). Moreover, theinclusion of different APIs in the final solutionincreased the pervasiveness of the technology insociety (Youtie et al., 2008). (See Table 2)

5.1. IBM Watson teaches academics

From a theoretical standpoint, this investigationenriches academic knowledge in various per-spectives. First, the study contributes theoreticalknowledge of AI by expanding upon related liter-ature (Kaplan & Haenlein, 2019). Indeed, it showsthat aside from changes in business models(Nenonen & Storbacka, 2018) and the possibilityto adopt it to address different markets, betterand deeper knowledge of AI can create value foracademics. Indeed, by introducing the two di-mensions of internal and external exploration,companies that aim to adopt AI can create valuein the market. This is an interesting theoreticalcontribution, pointing out is even more difficultdue to the different stakeholders involved. Sec-ond, this study enhances the GPT literature byshedding light on the designing perspective of thecreation of a GPT (Gambardella & Giarratana,2013). This is a rather neglected aspect in theliterature. Third, it shows that companies cancommercialize technology differently and partic-ularly that the digital environment sustains such a

828 S. Magistretti et al.

development. The analysis shows that companiesshould approach technology development from adifferent perspective compared to the funnelapproach and the open innovation paradigm(Chesbrough, 2006). Technology is not channeledinto markets thanks to an inbound or outboundperspective (West & Bogers, 2017), but resideswithin the company, while external stakeholdersonly develop the features. This enriches the openinnovation and big data literature by showing thattechnology can be crafted to support multi-stakeholder interactions. Indeed, the big dataliterature argues that the dynamic involvement ofdifferent stakeholders can foster higher innova-tiveness (Bharadwaj & Noble, 2017). This studyalso shows the importance of stakeholders, aca-demics, partners, and citizens through hack-athons. Experimentation and development cansteer the technology toward different applicationfields, opening up commercial opportunities forthe company as well as potentially interestingmodifications that can be introduced to betterexploit the technology over different applications(Dell’Era et al., 2017).

5.2. IBM Watson teaches managers

Our investigation also has some interestingpractical implications for companies facingtechnological innovation. The first and mostimportant is that companies seeking to address awider market should leverage the hidden keyelements of the AI technology and explore themby integrating the technology in different appli-cation fields. Without exploration and integra-tion, it is difficult to unveil all the potential of AIand, as reported in the IBM Watson case, it isdifficult to predict ex-ante where the technologycan be applied. Commercializing it in one fieldallows managers to explore and understand theopportunities and particularities of the technol-ogy to create value and learn where the samekey traits are present, and even increase thevalue created through economies of scale (e.g.,big data and relevance of speed in creatingconnections between them). Second, the over-turned traditional funnel approach is relevant forpractitioners. Indeed, it suggests that theapproach toward AI development as a GPT shouldfollow a different path, not focusing on a singleuse but on broadening the application scope to alarger set of application fields (see Figure 3).Third, the involvement of different externalstakeholders, universities, hackathons, andpartners is fundamental to speeding up theintegration of AI in different markets.

6. Unlocking the opportunities hidden inAI

This study sheds light on how companies can createvalue through AI technologies. The unique andextreme case of IBM Watson offers novel insights onhow to unveil opportunities hidden in AI. First,based on the digital transformation (Nambisanet al., 2017) and GPT (Youtie et al., 2008) litera-ture streams, the investigation proposes a differentperspective on AI. Indeed, it indicates some theo-retical and managerial implications on how todesign AI solutions to become a GPT. This will helpcompanies aiming to commercialize these tech-nologies to achieve high pervasiveness in the mar-ket. In addition, to the best of our knowledge, thisinvestigation is one of the few that considers theex-ante creation rather than the ex-post study of aGPT. Moreover, it enhances current knowledge ontechnology innovation in the complex digitalecosystem. Adopting a longitudinal perspective inthe analysis, we show that technology must notonly be understood in detail (Danneels & Frattini,2018) but also crafted and designed to support itscontinuing evolution. Indeed, this allows openingopportunities while commercializing the technol-ogy without focusing on just one application andone product as is evident in the two main strategiesIBM adopted for internal development and externalcollaboration initiatives.

References

Ardito, L., Petruzzelli, A. M., & Albino, V. (2016). Investigatingthe antecedents of general purpose technologies: A patentperspective in the green energy field. Journal of Engineeringand Technology Management, 39, 81e100.

Bharadwaj, N., & Noble, C. (2017). Finding innovation in datarich environments. Journal of Product Innovation Manage-ment, 34(5), 560e564.

Boudreau, K. J., & Lakhani, K. R. (2013). Using the crowd as aninnovation partner. Harvard Business Review, 91(4), 60e69.

Bresnahan, T. F., & Trajtenberg, M. (1995). General purposetechnologies ‘engines of growth’? Journal of Econometrics,65(1), 83e108.

Bughin, J., McCarthy, B., & Chui, M. (2017, August 28). A surveyof 3,000 executives reveals how businesses succeed with AI.Harvard Business Review. Available at: https://hbr.org/2017/08/a-survey-of-3000-executives-reveals-how-businesses-succeed-with-ai

Chesbrough, H. (2003). The governance and performance ofXerox’s technology spin-off companies. Research Policy,32(3), 403e421.

Chesbrough, H. W. (2006). Open innovation: The new impera-tive for creating and profiting from technology. Brighton,MA: Harvard Business Press.

Cohen, W. M., & Levinthal, D. A. (1989). Innovation andlearning: The two faces of R&D. The Economic Journal,99(397), 569e596.

Enabling digital transformation through artificial intelligence 829

Cooper, R. G. (2008). Perspective: The Stage-Gate� idea-to-launch processdupdate, what’s new, and NexGen systems.Journal of Product Innovation Management, 25(3), 213e232.

Cuhls, K. (2003). From forecasting to foresight processesdNewparticipative foresight activities in Germany. Journal ofForecasting, 22(2/3), 93e111.

Danneels, E., & Frattini, F. (2018). Finding applications fortechnologies beyond the core business. MIT Sloan Manage-ment Review, 59(3), 73e78.

Dell’Era, C., Altuna, N., Magistretti, S., & Verganti, R. (2017).Discovering quiescent meanings in technologies: Exploringthe design management practices that support the devel-opment of technology epiphanies. Technology Analysis &Strategic Management, 29(2), 149e166.

Dobusch, L., & Kapeller, J. (2018). Open strategy-making withcrowds and communities: Comparing Wikimedia and crea-tive commons. Long Range Planning, 51(4), 561e579.

Eisenhardt, K. M. (1989). Building theories from case studyresearch. Academy of Management Review, 14(4), 532e550.

Erevelles, S., Fukawa, N., & Swayne, L. (2016). Big data con-sumer analytics and the transformation of marketing.Journal of Business Research, 69(2), 897e904.

Gambardella, A., & Giarratana, M. S. (2013). General techno-logical capabilities, product market fragmentation, andmarkets for technology. Research Policy, 42(2), 315e325.

Gambardella, A., & McGahan, A. M. (2010). Business-model inno-vation:General purpose technologies and their implications forindustry structure. Long Range Planning, 43(2/3), 262e271.

Garud, R., Jain, S., & Tuertscher, P. (2008). Incomplete bydesign and designing for incompleteness. OrganizationStudies, 29(3), 351e371.

George, G., Haas, M. R., & Pentland, A. (2014). Big data andmanagement. Academy of Management Journal, 57(2),321e326.

Harrison, D., & Kjellberg, H. (2010). Segmenting a market in themaking: Industrial market segmentation as construction.Industrial Marketing Management, 39(5), 784e792.

Helpman, E. (Ed.). (1998). General purpose technologies andeconomic growth. Cambridge, MA: MIT Press.

Helpman, E., & Rangel, A. (1999). Adjusting to a new technol-ogy: Experience and training. Journal of Economic Growth,4(4), 359e383.

Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite!the challenges and opportunities of social media. BusinessHorizons, 53(1), 59e68.

Kaplan, A. M., & Haenlein, M. (2019). Siri, Siri, in my hand:Who’s the fairest in the land? On the interpretations, illus-trations, and implications of artificial intelligence. BusinessHorizons, 62(1), 15e25.

Levinthal, D. A. (1998). The slow pace of rapid technologicalchange: Gradualism and punctuation in technologicalchange. Industrial and Corporate Change, 7(2), 217e247.

Magistretti, S., & Dell’Era, C. (2018). Unveiling opportunitiesafforded by emerging technologies: Evidences from thedrone industry. Technology Analysis & Strategic Manage-ment, 31(5), 606e623.

Nambisan, S., Lyytinen, K., Majchrzak, A., & Song, M. (2017).Digital innovation management: Reinventing innovationmanagement research in a digital world. MIS Quarterly,41(1), 223e238.

Nenonen, S., & Storbacka, K. (2018). Smash: Using marketshaping to design new strategies for innovation, value cre-ation, and growth. Bingley, UK: Emerald Publishing.

O’Donovan, P., Leahy, K., Bruton, K., & O’Sullivan, D. T. (2015).Big data in manufacturing: A systematic mapping study.Journal of Big Data, 2(20), 1e22.

Pettigrew, A. M. (1990). Longitudinal field research onchange: Theory and practice. Organization Science, 1(3),267e292.

Savino, T., Messeni Petruzzelli, A., & Albino, V. (2017). Searchand recombination process to innovate: A review of theempirical evidence and a research agenda. InternationalJournal of Management Reviews, 19(1), 54e75.

Smith, P. G., & Reinertsen, D. G. (1992). Shortening the productdevelopment cycle. Research-Technology Management,35(3), 44e49.

Strauss, A., & Corbin, J. (1998). Basics of qualitative research:Procedures and techniques for developing grounded theory.Thousand Oaks, CA: Sage.

Thomke, S., & Reinertsen, D. (2012). Six myths of productdevelopment. Harvard Business Review, 90(5), 84e94.

Urbinati, A., Chiaroni, D., Chiesa, V., & Frattini, F. (2018). Therole of digital technologies in open innovation processes: Anexploratory multiple case study analysis. R&D Management.Available at: https://onlinelibrary.wiley.com/doi/full/10.1111/radm.12313

Van Knippenberg, D., Dahlander, L., Haas, M. R., & George, G.(2015). Information, attention, and decision making. Acad-emy of Management Journal, 58(3), 649e657.

Verganti, R. (2011). Radical design and technology epiphanies:A new focus for research on design management. Journal ofProduct Innovation Management, 28(3), 384e388.

Verganti, R. (2017). Overcrowded: Designing meaningful prod-ucts in a world awash with ideas. Cambridge, MA: MIT Press.

West, J., & Bogers, M. (2017). Open innovation: Current statusand research opportunities. Innovation, 19(1), 43e50.

Wieland, H., Hartmann, N. N., & Vargo, S. L. (2017). Businessmodels as service strategy. Journal of the Academy ofMarketing Science, 45(6), 925e943.

Williams, R., & Edge, D. (1996). The social shaping of technol-ogy. Research Policy, 25(6), 865e899.

Yin, R. K. (2011). Applications of case study research. ThousandOaks, CA: Sage.

Yoo, Y., Boland, R. J., Jr., Lyytinen, K., & Majchrzak, A. (2012).Organizing for innovation in the digitized world. Organiza-tion Science, 23(5), 1398e1408.

Youtie, J., Iacopetta, M., & Graham, S. (2008). Assessing thenature of nanotechnology: Can we uncover an emerginggeneral purpose technology? The Journal of TechnologyTransfer, 33(3), 315e329.