AI-driven RPA implementation : the effect of knowledge asymmetry between supplier and buyer on implementation success

(1)

AI-driven RPA Implementation:

The effect of knowledge asymmetry between supplier and buyer

on implementation success

Master thesis

MSc in Business Administration: Digital Business

Name: Valentijn van den Bos Student number: 11948787 Supervisor: Dr. A. Alexiou Date: 21 June 2018

(2)

1

This document is written by Student Valentijn van den Bos, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

List of figures and tables

Figure 1: Conceptual framework ... 22

Figure 2: Conditional effects of supplier reputation on the relationship between knowledge asymmetry and trust ... 32

Figure 3: Statistical findings ... 34

Table 1: Means, standard deviations, skewness, kurtosis, and correlations between items ... 28

Table 2: Effects between knowledge asymmetry, trust and implementation success, and the mediating effect of trust. ... 30

Table 3: The effect of knowledge asymmetry on trust moderated by the reputation of the supplier ... 31

(5)

4

Table 4: The indirect effect of knowledge asymmetry on implementation success mediated by trust and moderated by the reputation of the supplier ... 33 Table 5: Representative quotations and coding scheme ... 38

(6)

5

Abstract

This thesis seeks to contribute to the literature concerning the implementation of emerging technologies through a vertical inter-organizational perspective, by studying the supplier-buyer relationship in the implementation process of AI-driven RPA. In this context, the relationship between knowledge asymmetry and implementation success is studied. The results presented are based on a mixed-method sequential explanatory study design. Surprisingly, it appears that there is a positive relationship between knowledge asymmetry and implementation success, and a positive relationship between knowledge asymmetry and trust. However, the relationship of knowledge asymmetry and implementation success was found to be fully mediated by trust. It was found, furthermore, that there is a positive relationship between trust and implementation success, and that supplier reputation moderates the relationship between knowledge asymmetry and trust. These results are further elaborated upon in the post-hoc analysis of this thesis. The study concludes with an overview of the theoretical implications, the managerial implications, and suggestions for future research.

(7)

6

1. Introduction

Over the last couple of decades, technology has had a great impact on society. Innovations keep changing the way we approach our environment and daily activities. Emerging technologies grow fast and have a big impact on multiple social-economic domains (Rotolo et al., 2015). For business, this means that organizations should constantly be on the lookout for new developments. When an organization fails to adapt to technological advancements, it risks being pushed out of the market (Ab Rahman et al., 2017). The phenomenon first described as “creative destruction” (Schumpeter, 1950), and later as “disruptive technologies” by Christensen (1995), has turned whole industries upside down.

The Disruptive Innovation theory (Christensen et al., 2015) describes how smaller companies are able to use innovations to change industries and to push big players out of an industry in a matter of years. However, emerging technologies also allow incumbents to innovate and strengthen their position in the industry. Organizations are able to develop superior products and services by complementing existing resources with new technologies. (Bergek, 2013).

Many business and IT studies have focused on determining what separates the winners from the losers in the technological battles that reshape industries. It is a very broad topic and is therefore approachable from many different angles. Moreover, as emerging technologies constantly evolve and change business landscapes, this dynamic field of literature is in constant need for studies that challenge the applicability of established theories (Christensen, 2001; Hill & Rothaermel, 2003). One way that this can be achieved, is by gaining practical insights into emerging technologies (Danneels, 2004; Markides, 2006; Rotolo, 2015).

Studies have shown that organizational factors play a big role when adopting technology (e.g. Cohen & Levinthal, 1990; Teece et al., 1997). In other words, the successful adoption of technology does not necessarily depend on the technical aspects of the specific

(8)

7

technology, but also on whether the characteristics of the organization are adequately suited to handle the technological adoption. The majority of studies on contributing organizational factors have been conducted from an intra-organizational perspective. However, factors also need to be studied from an inter-organizational perspective (Afuah 2000: Battistella et al, 2016). Especially since organizations often seek to complement their resources by collaborating with other organizations in order to keep up with technological change (Hagedoorn, 1993). This study focusses on the supplier-buyer relationship in the implementation process of an emerging technology. It thereby seeks to contribute to the scarce literature on organizational factors contributing to successful technological implementation in a vertical inter-organizational relationship. To date, research on technological relationships between organizations has mostly focused on horizontal relationships between organizations such as joint ventures and strategic alliances (e.g. Blomqvist, 2002; Ross, 2016; Vonortas & Zirulia, 2015).

Numerous authors have studied knowledge as an organizational factor contributing to the success of a technological implementation process. Perhaps unsurprisingly, such lack of knowledge of the technology is said to have a negative effect on the implementation success (Smith et al., 1998; Sridharan et al., 2010). This lack of knowledge can be compensated by involving the supplier in the implementation process (Lee, 2016). However, it is not clear to what extent such lack of knowledge can be compensated. For this thesis, knowledge is approached as a relative factor between organizations, by studying how knowledge asymmetry between supplier and buyer affects the success of the implementation process.

During the last few years, technological advances in robotics and Artificial Intelligence (“AI”) have led to innovative automation solutions that have been making an impact across multiple industries. Robotic Process Automation (“RPA”), represents software that can automate tasks across applications that were previously conducted by human workers. It is

(9)

8

software that does not require programming skills to operate it and functions through the presentation layer of the computer. Thereby it does not interfere with systems in place (Aguirre & Rodriguez, 2017). It is important to note, however, that physical robots, and software bots that assist rather than replace human activity do not fall under this category (Lacity et al., 2015).

RPA is complemented with developments in the field of AI. A device is intelligent when it perceives its surroundings, and uses these perceptions to succeed in tasks (Poole et al., 1998). An intelligent device, furthermore, possesses cognitive functions such as learning and problem solving. This gives the RPA software amongst others the ability to improvise. Without AI, the RPA software will only be able to carry out tasks that are clearly scripted. Unscripted changes to processes are therefore not in scope of RPA software without AI abilities.

With AI becoming more sophisticated, the potential job opportunities for RPA grow. It could potentially disrupt the global job market. The amount of human work that is up for robotic replacement will grow exponentially in the coming years as automation technology evolves. Research from 2013 suggested that 47% of the employment in the United States is at risk to be automated in the next five to fifteen years (Frey & Osborne, 2013). Recent research by HfS Research in conjunction with KMPG (2018) has shown, furthermore, that RPA is currently at the very top of next year’s IT investment agenda at the majority of the 250 companies that were included in the study. It is clear that RPA has considerable impact as an emerging technology, and that this impact is likely to increase over the coming years. In this thesis, AI-driven RPA will therefore be studied as an emerging technology.

Considering the above, this thesis aims to answer the following research question:

“How does knowledge asymmetry between buyer and supplier affect the success of the implementation process of AI-driven RPA?”

(10)

9

The provided research question was approached by means of a mixed-method sequential explanatory study design (Creswell et al., 2003). Besides knowledge asymmetry as the independent variable and implementation success as the dependent variable, the quantitative study also included the variables trust and supplier reputation. Trust was included to test whether trust mediates the relationship of knowledge asymmetry and implementation success. Supplier reputation was included to test moderation in the relation of knowledge asymmetry and trust. Following a literature review, it was expected that there is a negative relationship between knowledge asymmetry and implementation success, and a negative relationship between knowledge asymmetry and trust. It was furthermore hypothesized that there is a positive relationship between trust and implementation success, and that trust mediates the relationship between knowledge asymmetry and implementation success. Lastly, it was predicted that supplier reputation moderates the relationship between knowledge asymmetry and trust.

To test the hypotheses, a moderation, a mediation and a moderated mediation statistical analysis were conducted using the Hayes Bootstrapping method (Hayes, 2013). This three-step approach was used to ensure the validity of the results. It proved to be necessary to run the separate models to correctly test two of the hypotheses. The statistical study is complemented by an explanatory multi-case study (Yin, 2009) of eight organizations that recently implemented the technology, and one RPA consultant. The results are based on semi-structured interviews lasting between 35 and 60 minutes.

The current study makes multiple theoretical contributions to fields in literature regarding the implementation of emerging technologies and the relationship between buyer and supplier. The thesis concludes with offering managerial implications and directions for future research.

(11)

10

2. Literature review

This chapter provides the results of funneling through the relevant fields in literature that led to the research question that this paper seeks to answer. Firstly, the literature concerning emerging technology is discussed in a broader sense of the term. Secondly, the implementation phase of emerging technology and the supplier-buyer relationship are discussed. After that, the relevance of the studied technology, AI-driven RPA is explained. This chapter concludes by reviewing the literature on the studied variables (knowledge asymmetry, trust, supplier reputation and implementation success) and therewith building the hypotheses. Lastly, the conceptual framework summarizes the hypotheses.

2.1. Emerging technologies

Rotolo et al. (2015, p. 34) define an emerging technology as follows: “a relatively fast growing and radically novel technology characterized by a certain degree of coherence persisting over time and with the potential to exert a considerable impact on the socio-economic domain(s) which is observed in terms of the composition of actors, institutions and the patterns of interactions among those, along with the associated knowledge production processes.”

The interest in, and the impact of emerging technologies in economic and business literature dates back to the previous century. The literature on emerging technologies builds on the literature on economic innovation, which was first popularized by the work of Schumpeter (1950). He explained the phenomenon of “creative destruction”, with which he described how an innovation process can revolutionize an economic structure from within and thereby creating a new structure that destroys the old one.

Later, Christensen (1995) introduced the term “disruptive innovation”, which is closely related to the concept of creative destruction. An innovation is disruptive when it creates a new

(12)

11

market or value network and thereby disrupts the markets and value networks in place. Disruptive innovations are a threat for the established firms in the market because these established firms tend to focus on the mainstream market while smaller firms are focusing on developing their business model for the future market (Bowler & Christensen, 1995). Consequentially, it proves to be difficult for incumbents to adopt emerging technology in comparison to smaller firms (Chang et al., 2012), which inability leads to disruptive innovation. This topic has not seized to grow in relevance during the last few decades. Technological advances are following each other at a rapid pace, forcing incumbents within existing markets to adapt and innovate (Ab Rahman et al., 2017). It is therefore important to continuously challenge and improve the existing theory and assess its applicability in the business landscape of today (Christensen, 2001; Hill & Rothaermel, 2003). Moreover, the literature on emerging technologies has been in need of more specific and practical insights on the matter (Danneels, 2004; Markides, 2006; Rotolo, 2015). Where Schumpeter and Christensen were putting more emphasis on describing the general phenomenon, later literature has also focused on studying specific technologies in practice. For example, by studying the implementation phase of an emerging technology (e.g. Angeles, 2005). This paper aims to contribute by studying the implementation of AI-driven RPA, an emerging technology that is growing in significance throughout multiple industries.

2.2. Implementation of emerging technologies

Implementation is about bringing a decision or a plan into effect (Simpson et al., 1989). In the context of this study, it is defined by the period between the moment that the decision has been made to adopt an emerging technology, and the moment when the technology is operational within the company. The technical aspects of technology are not decisive for the level of success of the implementation. Emerging technologies often disrupt processes and

(13)

12

relationships within the organization, the interaction between the organization and the technology is therefore of crucial importance for the result of the implementation process (Edmondson, 2003). Influential authors have stressed the significant impact of multiple organizational factors that affect the ability of an organization to successfully implement technology, such as the capability of the organization to adapt its resource base (e.g. Teece et al., 1997) and the ability of firms to process external knowledge (e.g. Cohen & Levinthal, 1990).

However, most authors took an intra-organizational approach when addressing the factors leading to successful implementation of emerging technologies. Factors outside the organization also need to be studied to understand technological change (Afuah, 2000; Battistella et al., 2016). The need for inter-organizational perspectives has been growing as organizations increasingly depend on relationships with external organizations to be able to keep up with the environmental changes caused by the rapid evolvements in technology and innovation (Hagedoorn, 1993). In this paper, the results of the study of factors of influence in the relationship between organizations are listed.

Another distinction is apparent within the literature concerning inter-organizational factors. Within this category, most literature has focused on horizontal inter-organizational relationships, such as strategic alliances or other forms of technological partnerships (e.g. Blomqvist, 2002; Ross, 2016; Vonortas & Zirulia, 2015). In general, vertical relationships between supplier and buyer have received less attention than horizontal relationships between organizations (Squire et al., 2009). The relationship with a supplier is important because most organizations do not develop their technology in-house. Consequentially, the supplier has more experience concerning the technology that is about to be implemented. The relationship between supplier and buyer can therefore positively influence the outcome of the implementation process (Ragatz et al., 2002). Research needs to determine how factors in the

(14)

13

relationship between buyer and supplier influence this outcome. This paper seeks to contribute to filling this gap by reporting studied factors in the supplier-buyer relationship during the implementation of AI-driven RPA. As mentioned, AI-driven RPA is an emerging technology that is widely being implemented by organizations to increase efficiency and reduce costs (Lacity & Willcocks, 2015). However, the practical implications of RPA have not received as much academic attention when compared to its managerial popularity.

2.3. The supplier-buyer relationship

Although the literature on the supplier-buyer relationship is quite scarce in an emerging technological context, extensive research in other industries has shown that indeed suppliers are able to play an important role in the success of their buyers (Clark, 1989; Cusumano & Takeshi, 1991; Dyer, 1996; Dyer & Nobeoka, 2000). Authors have reasoned that involving suppliers in design and development processes can lead to competitive advantage for the buyer (Ragatz et al., 1997). The potential of a supplier-buyer relationship is that a buyer may lack some capabilities or resources that can be complemented by the capabilities and resources of the supplier. A buyer thereby strengthens its position in comparison to their competition. Thus, the relationship may lead to “Co-Opetitor-based competitive advantage” (Afuah, 2000). This difference in resources and capabilities between supplier and buyer are certainly not rare in the context of emerging technology implementation. Many organizations desire to innovate their products and processes while lacking the technical skills to do so. The solution for these organizations may be to hire external consultants (Ko et al., 2005), and/or to integrate the supplier in the implementation process (Hagedoorn, 1993). Suppliers can thus contribute to bridging the lacking resources and capabilities of buyers when implementing emerging technologies. Specifically, involving a supplier in setting the technical metrics and objectives of the implementation process is said to contribute to the development of a better design and

(15)

14

financial performance of the implemented technology. Supplier involvement in business metrics is of lesser importance (Petersen et al., 2005).

The literature mentioned in the paragraph above discusses proof that a supplier-buyer relationship can be beneficial for the buyer, by comparing cases with varying levels of supplier-involvement. However, such research rarely addresses which, and how specific factors in the relationship between supplier and buyer affect the outcome of the implementation (Ragatz, 2002), especially in a technical environment (Afuah, 2000; De Ruyter et al., 2001). From a review of recent literature does not follow that this gap has been filled since these observations were made. This paper addresses this gap by studying the relationship between knowledge asymmetry and the implementation success in the supplier-buyer relationship, in the context of emerging technology implementation.

2.4. Robotic Process Automation

To answer the research question, the implementation of AI-driven RPA is being studied. RPA represents software that can automate tasks across applications that were previously conducted by human workers. The software does not require programming skills to operate it, and functions through the presentation layer of the computer. Thereby it does not interfere with systems in place (Aguirre & Rodriguez, 2017). AI enhances the automation possibilities of RPA as it increases the independent capabilities of the technology. For example, the ability to improvise in case an automated process has slightly changed in-between tasks.

RPA fits the definition of an emerging technology that was given in the in paragraph 2.1. because it develops quickly with future applicability that is not yet fully understood, especially in combination with AI applications. The technology is furthermore being implemented across different industries such as law, energy and finance. The automation possibilities may have significant socio-economic impact on these industries (Lacity &

(16)

15

Willcocks, 2015). As mentioned, in 2013 research suggested that 47% of the employment in the United States might be at risk because of automation technologies such as RPA (Frey & Osborne, 2013).

2.5. Hypotheses

2.5.1. Implementation success

There has been an ongoing debate on how to effectively measure the success of the implementation of an IT project (Al-Mashari et al., 2003; Bradford & Florin, 2003; Delone & McLean, 1992; Zhe Zhang et al., 2005). The reason for this debate is that there is no obvious measurement for success, which problem also applies to discussions concerning the implementation of an emerging technology. One could for example wonder from which perspective the level of success is determined. The idea of success of a buyer might be very different than that of a supplier. It gets even more complicated when determining the variables that can be measured to assess implementation success. For example, should the level of satisfaction of the user after the implementation weigh heavier than the calculated monetary results that can be linked to the implementation after a certain period?

The different theories on IT implementation success originate from two main theoretical streams, although the two streams are not completely separated as derived theories overlap (Bradford & Florin, 2003). The first stream originates from the Diffusion of Innovations theory by Rogers (1962). The Diffusion of Innovations theory is best known for the curve that describes the process of how innovations are entering the market through different groups of adopters, from early adopters to laggards. However, Rogers (1962) describes organizational and individual characteristics that an organization should possess to be able to adopt an innovation. In this context, the theory originally focused on the question whether an organization has the right characteristics to adopt an innovation or not.

(17)

16

Subsequently, derived theories also focused on how the characteristics of an organization and its environment can lead to the success of IT projects (e.g. Fichman 2000).

The second main theoretical stream in IT project success originates from the Information Success theory by DeLone & McLean (1992). The title of their paper is strikingly called “Information Systems Succes: The Quest for the Dependent Variable” (1992). Their descriptive model contains the following dimensions which determine the level of implementation success: system quality, information quality, use, user satisfaction, individual impact and organizational impact. By focussing on these dimensions, this theory puts more emphasis on the characteristics and the outcome of the IT implementation rather than focussing on the organizational capabilities needed for successful implementation.

For this thesis, an approach has been chosen that is line with the Information Success implementation theory to assess the implementation success. Implementation success is determined by taking four dimensions into consideration: correspondence success, process success, interaction success, and expectation success (Al-Mashari et al., 2003). Correspondence success determines whether there is a match between the implemented technology and the objectives of the implementer. Process success is achieved when the implementation has been completed within time and budget. Interaction success occurs when the user’s attitudes towards the implementation is positive. Lastly, the expectation success depends on whether the implementation of the technology has matched the user’s expectations. Together these dimensions determine the level of implementation success, which serves as the dependent variable in the conceptual model of this thesis.

2.5.2. Knowledge asymmetry

Knowledge is a reoccurring factor when reviewing literature on the implementation of emerging technologies. Studies have shown that knowledge is a determining factor for the

(18)

17

performance of the technology, the potential of an innovation cannot be realized without knowledge (Lee, 2016). A lack of in-house understanding of the technology may therefore have a negative effect when implementing technology (Edmondson et al., 2003; Smith et al., 1998; Sridharan et al., 2010). However, as previously discussed this lack in resources can be compensated in the supplier-buyer relationship as the buyers can use the knowledge of their supplier by integrating him in the implementation process (Carlucci et al., 2004; Ragatz, 2005). When studying this relationship, it is therefore interesting to address the knowledge factor by looking at how the levels of knowledge in the supplier-buyer relationship relate to one another. This study will therefore focus on knowledge asymmetries between supplier and buyer in the implementation process. Knowledge asymmetry serves as the independent variable for the conceptual framework that is presented in the end of this chapter.

Blomqvist (2002, p. 5) defines asymmetry as “difference in resource, capabilities and power as well as management and culture of actors”. The height of the asymmetry is determined by how much difference there is in resources between the actors; as mentioned, knowledge is the measured resource in this study. In this study, a higher level of knowledge asymmetry means that the supplier of the technology possesses more expertise regarding the technology, and that the supplier has better personnel and resources to support the technology than the buyer, prior and during the implementation process. Asymmetries can be a source of issues in the relationship between supplier and buyer. Suppliers may for example face difficulties with the development of trust in the relationship with their buyer (Johnsen & Ford, 2008). Studies on the effect of knowledge asymmetry have furthermore stressed that this specific asymmetry can cause problems as well. Knowledge asymmetry causes knowledge barriers between parties that negatively influence the performance of an implemented technology (Attewel, 1992). However, more research is needed to capture the influence of these characteristics in the supplier-buyer relationship (Johnsen & Ford, 2008). The first two

(19)

18

hypotheses are established following the reasoning of the discussed literature on implementation success and knowledge asymmetry:

H1: There is a negative relationship between knowledge asymmetry between supplier and buyer and implementation success.

H2: There is a negative relationship between knowledge asymmetry between supplier and buyer and the perceived trustworthiness of the supplier.

2.5.3. Trust

Trust is a variable that has been studied in many fields of literature such as social psychology, sociology, economics and marketing (Doney & Cannon, 1997). Doney & Cannon (1997, p. 36) define trust as “the perceived credibility and benevolence of a target of trust”. Trust therefore depends on the perception of one party of the intentions of the other. In a supplier-buyer context, a trustworthy party has also been defined as: “a trustworthy buyer or supplier is one who: does not act in a purely self-serving manner, accurately discloses relevant information when requested, does not change supply specifications, and generally acts in an ethical manner” (Smeltzer, 1997, p. 41). Trust in the context of this study is on an organizational level; this means that the level of trust was measured by the perception of the buyer towards the supplier as an organization. As opposed to the perceived trustworthiness of individuals within the company, such as the sales representatives of a company. These individuals may damage or contribute to the trustworthiness of the supplier but do not define its trustworthiness as a whole (Doney & Cannon, 1997).

Prior research has shown that trust is an important factor in partnerships. It has been described as the factor that can either make or break the relationship (Ariño et al., 2001, George

(20)

19

& Farris 1999). In supplier-buyer relationships, trust has also proved to be one of the most important relationship characteristics (Doney & Cannon, 1997; Dwyer et al., 1987; Squire et al., 2009). Although organizational trust has proven to be less influential for parties than the individual-based trust (Blomqvist, 2002), especially in high-technology industries both types of trust play a crucial role in the relationship between organizations. High-technology industries are defined by innovations and fast disruption, parties are therefore faced with the complexity and the perceived risk of the products and services that suppliers deliver (Ruyter, 2001). This causes asymmetric relationships. Parties who engage in these relationships are therefore positioning themselves vulnerably when they seek a partner to complement the resources and capabilities that they themselves lack (Blomqvist, 2002). Therefore, it becomes more evident that parties trust one another in their intentions when, e.g. in this study, buyers are implementing a technology that they might not technically understand. Blomqvist (2002) did extensive research on trust in asymmetric technological relationships between organizations, which showed the importance of trust on multiple levels within the relationship. She for example concluded that trust of more value in asymmetric relationships than formal contracts. However, her research focused on horizontal relationships between companies. This study aims to study the importance of trust in asymmetric technological supplier-buyer relationships as well. Trust is therefore studied as a potential mediator between knowledge asymmetries and the success of the implementation phase. Research in comparable fields, e.g. a study on effective knowledge transfers by Levin & Cross (2004), has shown that trust often has a mediating role in these sorts of relationships. Based on the arguments discussed in this paragraph, hypotheses 3 and 4 are defined:

H3: There is positive relationship between the perceived trustworthiness of the supplier by the buyer, and the level of implementation success.

(21)

20

H4: Trust mediates the relationship between knowledge asymmetries between supplier and buyer and the level of implementation success.

2.5.4. Supplier reputation

The reputation of the supplier is the last variable that is included in the conceptual framework of this study. Doney & Cannon (1997, p. 37) define a supplier’s reputation as “the extent to which firms and people in the industry believe a supplier is honest and concerned about its customers”.

In a business context, both competence and goodwill are needed to develop trust (Blomqvist, 2002). Goodwill entails the moral responsibility and the positive intention held towards the other party. To put this in the perspective of a buyer: a buyer needs to believe that the intentions of the supplier are aligned with his own, before he will trust him as a partner. On the other hand, trust in the professional relationship between two parties is largely determined by the perceived level of competence. In asymmetric relationships, this is even more evident because the incentive for the partnership relies largely on the believe that the intended partner is able to fill the gap in resources and knowledge held by the other party (Blomqvist, 2002). However, it is harder for buyers to assess competence when knowledge asymmetry is high because they lack knowledge of the technology to judge the performance of the supplier. During an implementation process of an emerging technology such as AI-driven RPA, asymmetric relationships often occur. As mentioned before, companies want to implement a technology because they learn of the potential benefits that it holds, which does not mean that they have the technical knowledge to implement it independently. Trust therefore becomes important. This kind of trust originates partially from the reputation of the organization (Doney & Cannon; 1997; Smeltzer, 1997; Wagner & Coley, 2011). Suh & Houston (2010) even go as far as stating that supplier reputation is such a big part of trust that it is more important than

(22)

21

trust in a supplier-buyer relationship. It is expected that the reputation of the supplier weighs heavier on the trustworthiness of a supplier when knowledge asymmetries are higher between supplier and buyer. The reputation of a company is established through the judgements of multiple external sources. When a buyer does not have the technical knowledge to individually assess the resources and capabilities of the supplier, he will depend more on such external judgements. Moreover, when buyers lack the knowledge to judge the competence of the supplier, they have to rely more on the external judgements of goodwill of the supplier. This part is covered by supplier reputation as measured in this study. To test this claim, the following hypothesis is included in the model:

H5: The reputation of the supplier moderates the relationship between knowledge asymmetry and trust between the supplier and buyer, so that trust is higher when reputation is higher.

2.6. Conceptual framework

The defined hypotheses are summarized in the conceptual model that is displayed on the following page in figure 1. The next chapter describes how these hypotheses were tested.

(23)

22

Figure 1: Conceptual framework

H1: There is a negative relationship between knowledge asymmetry between supplier and buyer and implementation success.

H2: There is a negative relationship between knowledge asymmetry between supplier and buyer and the perceived trustworthiness of the supplier.

H3: There is positive relationship between the perceived trustworthiness of the supplier by the buyer and the level of implementation success.

H4: Trust mediates the relationship between knowledge asymmetries between supplier and buyer and the level of implementation success.

H5: The reputation of the supplier moderates the relationship between knowledge asymmetry and trust between the supplier and buyer, so that trust is higher when reputation is higher.

(24)

23

3. Methodology

3.1. Setting & data collection

The hypotheses were tested following a mixed-method sequential explanatory study design (Creswell et al., 2003). This method entails that the research consists of a quantitative study, which is complimented by a qualitative study. The qualitative study aims to triangulate and further investigate the findings of the quantitative study. This method was chosen because the literature on the studied relationships is scarce, and because the subject of this thesis is a part of the dynamic landscape of emerging technologies. The complementary qualitative study gives more insights in why knowledge asymmetry affects implementation success in the found way. This makes the contribution of this thesis more tangible.

The quantitative study is based on data that was originally collected by Dr. A. Alexiou, Prof I. Oshri, and Dr. S. Khanagha. The data was gathered by means of an online questionnaire (Appendix 1). The scales used are all adopted versions of previously validated scales. The sample was furthermore defined by multiple criteria to ensure a high level of reliability and validity. The survey checked whether: the respondent worked for an organization that had contracted a third-party RPA supplier, if their automation solution has AI abilities, and if the respondent was involved in the implementation process. Respondents that did not meet these qualifications and smaller companies with ten or fewer employees were screened out. The original dataset consisted of 153 remaining respondents that completely filled in the survey. Prior to any analysis, unengaged respondents were deleted from the data set. The questionnaire contained plural reverse coded questions to test the attentiveness of respondents. Respondents were excluded from the analysis if they missed the reverse coded character of two or more questions. 107 respondents remained after deleting the unengaged respondents. The results of the quantitative study are based on this remaining population. The remaining respondents work

(25)

24

for companies from different sectors and have varying positions within their organizations. To analyze the data, the statistical software SPSS was used.

The following qualitative study further investigated the findings that were revealed by the quantitative analysis. The analysis is based on a multiple case study (Yin, 2009) on companies that recently implemented AI-driven RPA, originating from nine relevant companies from different sectors. The cases were identified through purposeful sampling using similar criteria as the quantitative study. As a result, ten semi-structured formal interviews were conducted that lasted between 35 and 60 minutes. The interviews were conducted in collaboration with two other students who also studied the supplier-buyer relationship in AI-driven RPA implementation. One of these interviews was conducted during a telephone conversation. The other interviews were conducted in person. The semi-structured approach was chosen to ensure that all relevant findings were discussed, while leaving room to further discuss additional phenomena that were brought to light during the interview. The first interview served as pilot for the remaining interviews and was therefore not included in the analysis. It tested whether the questions were clearly formulated and whether the structure of the interview contributed to the fluency of the conversation. The questions were improved accordingly following the pilot (Appendix 2). Eight out of the nine remaining interviewees work for companies with 35 to 80,000 employees, active in different sectors such as finance, logistics and insurance. Their position within the hierarchy of the companies differed, but all of them were directly involved in the implementation process of RPA on the side of the buyer. A consultant was furthermore interviewed who had supported multiple organizations in their implementation phase. He was included to offer bystander insights of the studied relationships. The collected data was transcribed and coded following a selective coding procedure (Corbin & Strauss, 1990). This coding procedure was completed in the qualitative data analysis tool NVivo. All interviews were conducted in Dutch, transcribed in Dutch, and coded in Dutch.

(26)

25

This approach was chosen to prevent any language barriers from affecting the data. All interviewees spoke Dutch as their first language. The quotes in table 5 were translated to English.

3.2. Measures

Unless indicated otherwise, the measurements are based on a five-point Likert scale ranging from 1 = “totally disagree” to 5 = “totally agree”. The Cronbach’s alpha for all scales that were included in the analysis, are shown in table 1.

3.2.1. Implementation success

Following Al-Mashari et al. (2003), implementation success was determined by considering four dimensions: correspondence success, process success, interaction success, and expectation success. These dimensions were represented in the items that were used to determine the scale of implementation success. Example items include: “The automation project was completed with low implementation and service costs”, “The automation service attained our desired level of quality”, “The automation implementation was delivered within the intended time line”, and “Overall, the objectives of the automation project were completely achieved”. By considering these dimensions, the model of this thesis has measured the implementation success of the automation solution, perceived by the buyer.

3.2.2. Knowledge asymmetry

To determine the scale of knowledge asymmetry between buyer and supplier, four dimensions of knowledge asymmetry were included. Respondents we asked to indicate the difference in understanding of the automation solution, resources to support the automation solution, relevant skilled personnel, and the difference in in-depth knowledge of the automation

(27)

26

solution between them and the supplier. Example items include: “It was clear that the automation solution supplier had a better understanding of the automation solution than my organization does”, and “It was clear that the automation solution supplier was more able to lend skilled personnel to resolve unexpected problems than my organization can”.

3.2.3. Trust

Trust in this study represents the perceived credibility and benevolence of the supplier by the relevant buyer, on an organizational level. The items that were used in the questionnaire to determine the scale of trust are based on the measurements that were generated by Doney & Cannon (1997). They measured trust based on eight items. Example items include: “We believe the information that our automation solution supplier provides us with”, “When making important decisions, our automation solution supplier considers our welfare as well as its own”, and “Our automation solution supplier is trustworthy”.

3.2.4. Supplier reputation

In order to measure supplier reputation, another scale from Doney & Cannon (1997) was used. The items assess whether the supplier is known for being honest and concerned about its customers. The items include: “Our automation solution supplier has a reputation for being honest”, “Our automation solution supplier is known to be concerned about customers”, and “Our automation solution supplier has a bad reputation in its market”.

3.2.5. Control variables

Two variables were controlled throughout the statistical analysis: the size of the firm and the respondent’s years of experience in outsourcing. The firm size was held constant as it might influence the available resources of the company and the amount of external knowledge

(28)

27

that can be acquired during the implementation process. The size of the organization was determined by the number of employees. Respondents could indicate the global number of employees of their organization by choosing one of nine options. The lowest option indicated ten or fewer employees, and the highest option indicated 5,000 employees or more. Respondents where asked to specify the number of employees if the highest option applied to their organization.

The years of experience in outsourcing was held constant because the implementation of RPA shows many similarities with Business Process Outsourcing (“BPO”) (Lacity et al., 2015). Broadly speaking RPA is in some ways also a form of outsourcing. Instead of outsourcing processes to another organization, the processes are outsourced to a software program. Experience in BPO may therefore serve as an advantage for an organization when implementing RPA. Knowledge asymmetry, as measured in this study, does not take this sort of knowledge into account. Respondents were given six options to indicate their level of experience. The lowest alternative indicated less than one year, and the highest alternative indicated more than ten years of experience in outsourcing. Respondent where asked to specify the number of employees if the highest option applied to their organization.

(29)

28

4. Results

This chapter discusses the results from the statistical analysis and the post-hoc qualitative analysis. The Bootstrapping method by Hayes (2013) was used to carry out the statistical analysis in SPSS. Rather than only testing the full moderated mediation model, a simple mediation and moderation analysis were run as well. Although seemingly superfluous, this proved helpful to ensure a higher validity of the output, and to prevent direct relationships from being distorted by additional variables in the full model. The latter proved to be necessary for testing the relationship between knowledge asymmetry and trust (H3). Moreover, the separate moderation analysis provides more insights into H5 than the full moderated mediation analysis.

The post-hoc analysis discusses insights on the statistical findings based on the presented quotes in table 5. These quotes are non-exhaustive but serve as examples of the found relationships.

Table 1: Means, standard deviations, skewness, kurtosis, and correlations between items

4.1. Normality, reliability and correlation

Table 1 shows the results used to assess the normality of variables and reliability of the used scale, of the data, and the (simple) correlations between the variables. When looking at the measures of skewness and kurtosis, the results show that almost all values comply with the

(30)

29

rule of thumb that values should be within the range of -1 and +1. The skewness of employee numbers and the kurtosis of knowledge asymmetry exceed this range. However, the values are still well between the acceptable range of -2 and 2 (Field, 2009). It can therefore be concluded that no strong deviations from normality were present.

Reliability was judged by calculating Cronbach’s Alpha (Cronbach, 1951). The scales for Performance (a = .81), Trust (a = .83) and Supplier Reputation (a = .73) all score high on reliability. The corrected item-total score furthermore showed that the items have a good correlation with a score above .30. Also, excluding one of the items would not substantially improve the reliability of one of the scales. Following the reliability test on knowledge asymmetry, one item was deleted. The scale improved by more than .10 following the exclusion, this led to Cronbach’s Alpha = .69. The deleted item concerned the following reverse coded question: “It was clear that my organization possessed more in-depth knowledge of the automation solution than the supplier”. It is probable that the structure of this sentence was wrongly interpreted by unengaged and engaged participants alike, which resulted in low reliability of the scale. However, the scale for knowledge asymmetry still falls short (α < .70), even after the exclusion of the reverse coded item. Despite failing this reliability test, the variable was excluded from the remaining analysis. One of the limitations of Cronbach’s Alpha is that it assumes identical covariance between items (Dunn et al., 2014). As discussed in the previous chapter, knowledge asymmetry was determined by multiple factors such as understanding of the technology by the respondents, resources to support the solution and knowledge possessed by personnel. As these different factors can independently contribute to knowledge asymmetry, the covariance between the items is not self-evident. This negatively affects Cronbach’s Alpha for knowledge asymmetry.

(31)

30

The correlation between variables was determined using the Pearson correlation coefficient. The results in Table 1 show that all variables are highly correlated with a value close to 0.5 or higher (Cohen, 1988).

4.2. Mediation analysis

Table 2: Effects between knowledge asymmetry, trust and implementation success, and the mediating effect of trust.

Consequent

M (TrustTot) Y (SuccessTot)

Antecedent Coeff. SE p Coeff. SE p

X (KnowledgeAsymTot) a 0.578 0.768 < .001 c' 0.179 0.096 .064 M (TrustTot) --- --- --- b 0.475 0.984 < 0.01 Constant i1 i2 R2 = 0.364, p < .001 F (3, 103) = 19.608 R2 = 0.391, p < .001 F (4, 102) = 16.364

Effect SE p LLCI ULCI

Direct effect c' 0.179 0.096 .064 -.011 .368

Total effect c 0.453 0.085 <.001 .286 .621

Effect Boot SE Boot LLCI Boot UCLI

Indirect effect ab 0.275 0.090 0.129 0.474

The results presented in table 2 follow from a simple mediation analysis conducted using ordinary least path analysis via the PROCESS tool (Hayes, 2013). It shows that knowledge asymmetry between buyer and supplier, indirectly and positively influences the level of implementation success through the trustworthiness of the supplier, as perceived by the buyer. There is a positive relationship between knowledge asymmetry and trust (a = 0.58, p < .001), which shows that a buyer with less knowledge (more asymmetry) is more likely to trust the supplier in the relationship. There is also a positive relationship between trust and implementation success (b = 0.48, p < .001), this shows that the implementation success is

(32)

31

expected to be higher when the buyer relatively trusts the supplier more. A bias-corrected bootstrap confidence interval for the indirect effect (ab = 0.275) based on 5,000 bootstrap samples was above zero (0.129 and 0.474). Finally, the analysis shows that there is no evidence to support the assumption that a direct relationship exists between knowledge asymmetry and implementation success independent of the effect of trust (c’= 0.179, p = .064). It can therefore be concluded that the level of trust fully mediates the relationship between knowledge asymmetry and implementation success.

4.3. Moderation analysis

Table 3: The effect of knowledge asymmetry on trust moderated by the reputation of the supplier

TrustTot (M) R2 = 0.653, MSE = 0.176 Coeff. SE t p Intercept i1 3.292 1.136 2.987 .005 KnowledgeAsymTot (X) c1 -0.445 0.305 -1.458 .149 SupplierReputationTot (W) c2 -0.100 0.304 -0.329 .743 c1*c2 (XW) c3 0.169 0.077 2.210 .029 SupplierReputationTot Effect SE t p Low 0.096 0.088 1.088 .279 Moderate 0.231 0.071 3.257 .002 High 0.367 0.099 3.691 < .001

(33)

32

Figure 2: Conditional effects of supplier reputation on the relationship between knowledge asymmetry and trust

A simple moderation analysis via the PROCESS tool (Hayes, 2013) resulted in the numbers presented in table 3 and figure 2. The analysis examines under which levels of supplier reputation, knowledge asymmetry between supplier and buyer affects the level of trust. There is a significant interaction between knowledge asymmetry and supplier reputation on trust (c3 = .169, p = .029). Thus, the effect of knowledge asymmetry on the level of trust between the supplier and buyer depends on the reputation of the supplier. Moreover, this model accounts for 65% of variance in the level of trust between supplier and buyer. A closer look at the conditional effects in table 3 indicates that the relationship between knowledge asymmetry between parties and the trust between them, is only significant for moderate (effect = .231, p = .002), and high-level supplier reputation (effect = 0.367, p < .001), in contrast to low level supplier reputation (effect = .096, p = .279). These results, which are visualized in figure 2, show that higher levels of supplier reputation increase the positive effect that knowledge asymmetry between suppliers and buyers has on the perceived trustworthiness of the supplier.

(34)

33

4.4. Moderated mediation analysis

Table 4: The indirect effect of knowledge asymmetry on implementation success mediated by trust and moderated by the reputation of the supplier

Consequent

TrustTot (M) SuccessTot (Y)

Antecedent Coeff. SE p Coeff. SE p

X a1 -0.445 0.305 .149 c' 0.179 0.096 .064 M --- --- --- b 0.475 0.098 < .001 W a2 -0.100 -0.304 .743 --- --- --- W*X (WX) a3 0.169 0.077 .029 --- --- --- Constant i1 3.292 1.136 .005 i2 1.374 0.388 .001 R2 = 0.653, p < .001 F (5, 101) = 38.021 R2 = .391, p < .001 F (4, 102) = 16.364

The table above results from a moderated mediation analysis via the PROCESS tool (Hayes, 2015). Table 4 summarizes the findings of this analysis, which combines the models of the simple moderation and the simple mediation discussed in this chapter. It shows that even when the models are combined, the identified effects still hold except one. The exception concerns the relationship between knowledge asymmetry and trust, which has a different result in the models that include supplier reputation. However, the identified effect resulting from the mediation analysis leads as the relationship was not distorted by an additional variable in that model.

the findings in table 4 furthermore compliment the results from table 2 and 3 by reporting the conditional indirect effects between knowledge asymmetry and implementation success, for the levels of Supplier Reputation. They show that moderated mediation has only

SupplierReputationTot Unstandardized Boot Effects

Boot SE Boot LLCI Boot ULCI

Low 0.046 0.050 - .044 .158

Mid 0.110 0.046 .045 .232

(35)

34

occurred for moderate and high-level supplier reputation, as the bias-corrected bootstrap confidence interval for the conditional indirect effect, based on 5,000 bootstrap samples were entirely above zero (respectively .045 and .232, .073 and .346).

Figure 3: Statistical findings

*p < 0.05, **p < 0.01 Supported hypotheses:

H3: There is positive relationship between the perceived trustworthiness of the supplier by the buyer and the level of implementation success.

H4: Trust mediates the relationship between knowledge asymmetries between supplier and buyer and the level of implementation success.

H5: The reputation of the supplier moderates the relationship between knowledge asymmetry and trust between the supplier and buyer, so that trust is higher when reputation is higher. Rejected hypotheses:

H1: There is a negative relationship between knowledge asymmetry between supplier and buyer and implementation success.

H2: There is a negative relationship between knowledge asymmetry between supplier and buyer and the perceived trustworthiness of the supplier.

(36)

35

Figure 3 summarizes the identified effects, and lists the tested hypotheses. H1 and H2 were surprisingly rejected, H3, H4 and H5 were supported by the quantitative study. The next paragraph aims to triangulate these findings by analyzing qualitative data. This data was gathered by ten conducted interviews, one of which interview served as a pilot. These interviews were conducted to obtain more understanding, obtain more context, and to obtain insights into the findings discussed in the last paragraphs.

(37)

36

4.5. Post-hoc analysis

Table 5 summarizes the findings that followed from a selective coding procedure (Corbin & Strauss, 1990) of the data gathered through a multi-case study (Yin, 2009) of nine buyers and one RPA consultant (10 formal interviews ranging between 35 and 60 minutes). One these interviews served as a pilot for the remaining interviews and was therefore not included in the coding procedure. The purpose of this study is to triangulate and explain the findings that resulted from the quantitative findings discussed earlier in this chapter. The quotes in table 5 serve to support the general findings of the coding procedure which will be discussed and explained hereafter. The qualitative findings confirm the effects of the variables that were found in the quantitative study.

To start with knowledge asymmetry, the conducted interviews provided valuable insights into how these asymmetries affect the perception of buyers of the implementation success resulting from the implementation phase. Many buyers that had limited knowledge of RPA did not see having technical knowledge as a necessity (Table 5, Row 1a). These buyers often learned of the potential benefits of the technology and looked for a way to realize such benefits within their own companies. As a result, these buyers leave it up to the suppliers of the technology to use their knowledge to prove the worth of their product (Table 5, Row 1b), this gives the supplier a certain amount of freedom while handling the implementation process. One can imagine that this positively influences the timeframe and costs of the implementation because it can limit the adjustments of the process as a result of discussions with buyers. The knowledge asymmetry may, furthermore, limit the expectations of a buyer because it forces them to focus on the possibilities of the specific product of the supplier rather than on the technology in a broader sense (table 5, row 1c). Knowledge of RPA allows a buyer to be more critical towards the possibilities that the technology offers and the steps that are taken by the supplier during the implementation phase (Table 5, row 1d). This shows that the perceived

(38)

37

success of the implementation, judged by a buyer with more knowledge is subject to more critical considerations. Moreover, buyers that possess more knowledge may be more daring in their ambitions with the technology. This results in more complicated challenges for the supplier, which are harder to successfully complete in terms of implementation success (Table 5, row 1e). This is easier with buyers with less knowledge that take the implemented technology and its features as presented by the supplier.

The explanation of the positive relationship between knowledge asymmetry and trust follows the reasoning of the findings on the relationship between knowledge asymmetry and implementation success. This connection in reasoning was expected because of the statistical findings on the mediation of trust in this relationship. In table 5, row 2a and 2b suggest that buyers are inclined to trust their supplier because of the knowledge asymmetry, mainly because it is harder for buyers to judge the actions that may give reason to distrust the supplier during the implementation process. This is due to the buyer being less involved in the technical side of the implementation. In this line of reasoning it is apparent that trust is a fundamental factor between buyers and suppliers in order to reach high implementation success when knowledge asymmetries exist (Table 5, row 4). As a result, the supplier plays a bigger role in the implementation process in such a situation. This explains the full mediation effect of trust that was found in the relationship between knowledge asymmetry and implementation success.

Finally, the moderating effect of the reputation of the supplier in the relationship between moderate and high levels of knowledge asymmetry and trust was also confirmed by the respondents. The expected positive effect of the moderator results from the fact that buyers look at how their supplier is perceived by external parties. Buyers with less knowledge of RPA have to rely more on these types of sources. The interviews with respondents that experienced high levels of knowledge asymmetry (e.g. table 5, row 5b) showed that the moderation effect can be explained by the fact that these buyers have less criteria to base their trust upon. Supplier

(39)

38

reputation therefore plays a bigger role in the trust assessment of the buyer when knowledge asymmetries are higher.

Table 5: Representative quotations and coding scheme

Coding scheme: [A]: Knowledge Asymmetry; [B] Implementation success; [C] Trust; [D] Supplier Reputation

Relationship Effect Representative quotations

1a H1 +/- “Look, there will always be an asymmetry in knowledge [A]. At the moment that a supplier meets it buyer, there is always a difference in what the buyer knows of the product that the supplier sells. How has that fact affected the success [of the implementation] [B]? Not at all, in my opinion, because we are also hosting software we do not know anything about. And the supplier does its thing on them, but we make sure that it operates in the right environment.”

1b 1c 1d 1e + + + +

“In the beginning [of implementing RPA] it was more important that things [RPA] just worked [B]. We were not critical because we did not have the knowledge internally. We were not very critical about the internal aspects of it [RPA] or that sort of things [A].”

“Exactly, we just thought, we have a [software] package and we operate between its boundaries. (…) So, do we have certain wishes? Yes of course, but really, and I apologize for repeating myself, I just see it as Excel. I bought advanced Excel, to put it like that [A].”

“But sometimes you are more capable to think about alternatives [to actions of the supplier] because you know more [A]. While someone without knowledge [of the technology] could think, that sounds about right. [B]” “So [RPA] is a good tool to begin with, you do not need any programming skills [to work with RPA]. If you are a little IT savvy, and you follow a training for a few days then you can already start scripting. Only when you really want to achieve something, in terms of scalability, you always get in trouble.” 2a H2 + “Knowledge asymmetry naturally influences the

implementation process, because in that case you let yourself be leaded by their [supplier] knowledge [C]. And then it is up to me to say: OK, you are saying that this is the best direction, but is it really the best direction? That is hard to control or to

(40)

39

2b +

check like OK, the direction that they chose was the right direction [A].”

“No, In the beginning we said: we want robots and they have to work [B]. And [supplier], you guys say that you are capable of pulling it off, start working and show us [C]. Along the way you notice that they make certain discussions of which you might think: is that really a smart decision? A good decision? But then you do not have the knowledge to really get involved [A].”

3a

3b

H3 +

+

“Yes, I believe that [the importance of trust for the success of the implementation]. I need to think whether it caused delays, that distrust [of the respondent’s supplier] [C]. Look, I can imagine that you would want to investigate [the actions of the supplier] more if there is a lack of trust, and that the process takes longer as a result [B].”

“There always needs to be a minimal basis of internal support and trust [between buyer and supplier] [C], otherwise it does not make sense [the relationship between buyer and supplier] [B].”

4 H4 + “Trust, that is the most importing thing there is. When you reach a point where you think that there are trust issues [between supplier and buyer] or that they [supplier] have a hidden agenda, then it becomes very hard because of knowledge asymmetry [A]… To accept that they chose the right direction. When the trust is right, then you can blindly believe [C] that things are often going to fall into place [B].” 5a

5b

H5 +

+

“Yes, it is a tiring relationship. But not a relationship we want to end [C]. At the end of the day they are one of the best three suppliers [D]… [of RPA in the market].”

“[Supplier A] is positioned in Gartner’s Magic Quadrant. They are technological pioneers [D]. What also helps is that they attracted a couple of big investors in the last year and a half. That shows that there is trust in the software. [Interviewer: And do you base your trust on that?] Yes, yes? Yes. You know, it generates trust [C], especially the partnerships [between supplier and other technological companies]." (Note: high knowledge asymmetry between respondent and supplier [A].)

(41)

40

5. Discussion

5.1. Theoretical contributions

First, on a general level, this study contributes to the literature on emerging technologies. New technologies continuously emerge and change society and organizations alike (Ab Rahman et al., 2017). It is therefore important to challenge existing theories on emerging technologies and test them in new contexts (Christensen, 2001; Hill & Rothaermel, 2003). This thesis has answered this call by studying RPA with AI abilities, a technology that has been widely implemented in recent times but has yet to receive more academic attention. The positive role of trust in the relationship between a buyer and a supplier, is a topic that already has been studied in many comparable settings (Blomqvist, 2002; Doney & Cannon, 1997; Dwyer et al., 1987; Smeltzer, 1997; Squire et al., 1997). It was therefore no surprise that a positive relationship was found between trust and implementation success. However, this finding contributes to the existing literature by portraying the relevance of trust once more in the modern context of AI-driven RPA implementation. The quantitative and qualitative findings of this thesis show that there is a strong positive relationship between trust and implementation success.

Within the scope of emerging technology implementation, the findings of this study contribute to the existing literature by taking an inter-organizational perspective between supplier and buyer. Knowledge as a factor in the implementation process of technologies has already been studied from different perspectives, often intra-organizational perspectives (e.g. Lee, 2016) or horizontal inter-organizational perspectives (e.g. Blomqvist, 2002; Ross, 2016; Vonortas & Zirulia, 2015). The findings on the effect of knowledge asymmetry on implementation success in a supplier-buyer relationship contributes to those studies by studying a vertical inter-organizational relationship.

AI-driven RPA implementation : the effect of knowledge asymmetry between supplier and buyer on implementation success