Adoption Intention of Artificial Intelligence in the B2B Selling Process

(1)

ADOPTION INTENTION OF ARTIFICIAL

INTELLIGENCE IN THE B2B SELLING

PROCESS

Aantal woorden / Word count: 16684

Hendrik Keeris

Stamnummer / student number : 01501412

Promotor / supervisor: Prof. Dr. Steve Muylle Commissioner: Nils Van den Steen

Masterproef voorgedragen tot het bekomen van de graad van: Master’s Dissertation submitted to obtain the degree of:

Master in Business Engineering: Data Analytics

(2)

(3)

Confidentiality Agreement

PERMISSION

I declare that the content of this Master’s Dissertation may be consulted and/or reproduced, provided that the source is referenced.

Student’s name: Hendrik Keeris Signature:

(4)

(5)

Abstract

Artificial Intelligence (AI) will have a profound impact on the B2B selling process of tomorrow. It can totally transform sales processes and customer interactions, virtually changing the way business is done. AI is seen as more significant and pervasive than previous new sales technologies. To be able to benefit from its applications, sales professionals need to get hands-on experience and discover all its possibilities. Unfortunately, a lot of companies are still reluctant to adopt AI in their selling process, hindering sales professionals to start working with these AI applications. This research is an exploratory study aiming to determine by which factors and how the adoption intention of AI in the B2B selling process is impacted. To determine these factors and their impact, Partial Least Squares (PLS) analysis was used based on data collected through a cross-sectional survey that was sent to sales professionals active in a B2B selling context in Belgium. Three factors were found to have a significant, positive impact on the adoption intention of AI: the relative advantage of AI; compatibility with current sales processes and with a clear business problem; and human, enterprise and technology resources that are already in place. So, if companies want to work in the B2B selling process of tomorrow, they need to get ready for using AI. They should teach their employees about the benefits of AI, they need to find a clear business case that can be tackled by AI, and even before that they should work on the resources that are already in place. They need to collect data, implement Customer Relationship Management (CRM) and Sales Enablement Platforms, and hire specialized personnel to implement the AI applications. This research focuses on people’s perception of AI in the B2B selling process, making it unique compared to previous studies, who were mainly focused on the technicalities of AI applications in the B2B selling process.

Keywords

artificial intelligence – B2B selling process – adoption intention – cross-sectional survey – partial least squares analysis

(6)

Foreword

This master’s dissertation concludes my master’s program in Business Engineering: Data Analytics at the University of Ghent. Due to the outbreak of COVID-19, my last semester at the University of Ghent was a little bit different than expected. Luckily, the coronavirus did not cause any problems or difficulties for my master’s dissertation as my survey was conducted online. On the contrary, being stuck at home gave me more time to thoroughly finalize my master’s dissertation. I would like to thank some people who helped me realize this work.

First of all, I would like to thank Prof. Dr. Steve Muylle and Nils Van den Steen for allowing me to write my master’s dissertation about the adoption intention of artificial intelligence in the B2B selling process. I especially want to thank Nils Van den Steen for his guidance and enormous support. He was always ready to give feedback whenever I needed it, even in the difficult times of corona.

Secondly, I would like to thank all my respondents who took the time to contribute to my research by filling in the survey. It allowed me to make conclusions based on more quantitative data.

Lastly, I want to express my gratitude to my father and family for always supporting me during my studies and, in this case, during my master’s dissertation.

Enjoy reading my master’s dissertation! Hendrik Keeris

(7)

-List of abbreviations

AI Artificial intelligence

AR Augmented reality

AVE Average variance extracted

B2B Business-to-business

CRM Customer relationship management

FAQ Frequently asked questions

IS Information system

IT Information technology

OLS Ordinary least squares

PEOU Perceived ease of use

PLS Partial least squares

PU Perceived usefulness

SEM Structural equation modeling

SME Small and medium-sized enterprise

TAM Technology acceptance model

TOE Technology-organizational-environmental

(10)

List of figures

Figure 1: Prospecting Situated in the Sales Funnel ... 7

Figure 2: Conceptual Framework AI Adoption Intention ... 19

Figure 3: Conceptual Framework AI Adoption Intention with Research Hypotheses ... 20

Figure 4: Model in SmartPLS (General Model) ... 33

Figure 5: Hypothesis Testing (General Model) ... 39

Figure 6: Model in SmartPLS (General Model) ... 40

Figure 7: Hypothesis Testing (Prospecting) ... 42

Figure 8: Model in SmartPLS (Chatbots) ... 43

Figure 9: Hypothesis Testing (Chatbots) ... 45

Figure 10: Model in SmartPLS (Presentations) ... 46

Figure 11: Hypothesis Testing (Presentations) ... 48

Figure 12: Model in SmartPLS (Immersive Technologies) ... 49

Figure 13: Hypothesis Testing (Immersive Technologies) ... 51

Figure 14: Summary Confirmed Hypotheses ... 53

List of tables

Table 1: AI Applications in the Seven Steps of Selling ... 13

Table 2: Research Hypotheses ... 25

Table 3: Sample Characteristics ... 27

Table 4: Factors and their Corresponding Statements ... 30

Table 5: Performance Measures of Measurement Model (General Model) ... 35

Table 6: Fornell-Larcker Criterion (General Model) ... 37

Table 7: Performance Measures of Measurement Model (Prospecting) ... 41

Table 8: Fornell-Larcker Criterion (Prospecting) ... 41

Table 9: Performance Measures of Measurement Model (Chatbots) ... 44

Table 10: Fornell-Larcker Criterion (Chatbots) ... 44

Table 11: Performance Measures of Measurement Model (Presentations) ... 47

Table 12: Fornell-Larcker Criterion (Presentations) ... 47

Table 13: Performance Measures of Measurement Model (Immersive Technologies) ... 50

Table 14: Fornell-Larcker Criterion (Immersive Technologies) ... 50

Table 15: Confirmed Hypotheses per AI Application ... 52

(11)

1. INTRODUCTION

Nowadays, one of the hottest topics in the world of technology is without a doubt

artificial intelligence (AI). A lot of people do not understand what it means and think

of it as Terminator-like robots who will take over the world. But actually, it is already present in everyone’s day to day routine. Thanks to AI, Netflix can propose films you will probably like, virtual assistants like Siri or Alexa can answer your questions, and Waze can show you the fastest way to work. It has become commonplace in our daily lives.

In the world of business, AI has had and will continue to have a profound impact on all kinds of business aspects and processes as well. One of the processes that will be heavily impacted is the B2B selling process. AI can totally transform sales processes and customer interactions, virtually changing the way business is done. It is even seen as more significant and pervasive than previous new sales technologies (Singh et al., 2019). To be able to benefit from its applications, sales professionals need to get hands-on experience and discover all its possibilities. Unfortunately, a lot of companies are still reluctant to adopt AI in their selling process (Alsheibani et al., 2018), hindering sales professionals to start working with these AI applications. So, it is up to the managers to bridge the gap between theory and practice and make their salespeople ready for the B2B selling process of tomorrow. This poses some questions on the factors that influence that readiness to adopt AI. Knowing these factors would enable managers to take appropriate actions to increase their salespeople’s intention to adopt AI.

Previous studies were mainly focused on the technicalities of AI applications in the B2B selling process. The focus of this dissertation does not lie in the technical understanding of AI, but rather in its broader context and outcome. I want to find out

by which factors and how the adoption intention of AI in the B2B selling process is impacted. This will enable me to come up with recommendations for managers on

how to prepare their companies and salespeople for the advent of AI in the B2B selling process.

(12)

My research will talk about AI using four different AI applications: AI used in prospecting, AI-driven chatbots, AI used to prepare sales presentations and AI-driven immersive technologies. To determine potential factors that might have an impact on the adoption intention of these AI applications, I will look at existing literature on technology adoption in general and on the adoption of AI at firm level. These factors will be combined in one conceptual framework and different hypotheses will be proposed. These hypotheses will then be tested using a Partial Least Squares (PLS)

analysis based on data collected through a cross-sectional survey sent to sales

professionals active in a B2B selling context in Belgium. Eventually, the results and managerial implications will be discussed and recommendations for further research will be made.

(13)

2. LITERATURE REVIEW

2.1 What is Artificial Intelligence?

Artificial Intelligence (AI) is the field of study that aims at building machines that are

as intelligent as people. Coopeland (2019) explains AI as “the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings” and that it concerns “developing systems endowed with the intellectual process characteristics of humans, such as the ability to reason, discover meaning, generalize, or learn from past experience”. This is similar to Syam & Sharma’s (2018) explanation. They say AI refers to “the ability of machines to mimic intelligent human behavior, and specifically refers to “cognitive” functions that we associate with the human mind, including problem solving and learning” (p. 2). Simply said, AI “involves using computers to do things that traditionally require human intelligence. This means creating algorithms to classify, analyze, and draw predictions from data” (Shroff, 2019).

Artificial Intelligence is often confused with other related terms such as Machine Learning and Big Data. Before going further on how AI is applied in the B2B selling process, it is important to fully grasp this difference.

As explained before AI is concerned with building machines that are as intelligent as human beings ranging from robotics to text analysis. You can distinguish two categories: General AI and Narrow AI, also referred to as Strong AI and Weak AI (Siau & Yang, n.d.). General AI refers to building machines that are intelligent in a high range of activities in which they need to think and reason just as a human being. Narrow AI, on the other hand, can just execute a very specific task (Sales, 2017). General AI can be, for instance, a customer service chatbot that can answer any possible question, even though the anwser to that question was never programmed before by a human being. Narrow AI though, will only be able to answer a limited amount of preprogrammed questions. For now, only Narrow AI can be reached by developers and researchers. This will therefore be the focus of my thesis.

(14)

Machine learning is a subfield of Artificial Intelligence. Through algorithms it can learn

from data, come up with insights on the data and make predictions on previously unanalyzed data based on these insights (Sales, 2017). You can see it as an

input-process-output transformation (Paschen et al., 2019). The machine or computer

receives input data, processes the data through a particular algorithm and generates output in the form of insights and/or predictions. Let us get back to the customer service chatbot as an example. When you type a certain question, you provide the computer with input data in the form of text. Next, the computer will process this text through a machine learning algorithm, which will be a so-called Natural Language Processing algorithm in this case. The output of this algorithm will be a standardized question which the computer understands so it can answer in an appropriate way.

There are two forms of input data: structured and unstructured data. Structured data are “data that are standardized and organized according to predefined schema” (Paschen et al., 2019, p.3). These structured data can be internal such as customer demographics or transaction data or can be external such as social media ratings or stock exchange transactions. It can be in the form of numbers, Booleans, categories, etc. These data are indispensable in business analytics and business intelligence.

Unstructured data, on the other hand, are “data that are not standardized or

organized according to a pre-defined schema” (Paschen et al., 2019, p.3). As in the customer service chatbot, it can be written text, but it can even be spoken text, pictures or videos. As compared to traditional information systems, Machine Learning algorithms can handle the vastly increasing amount of input data that come in unstructured formats (Paschen et al., 2019).

Machine learning methods can be put into three basic categories: supervised learning, unsupervised learning and reinforcement learning (Sales, 2017). I will shortly explain the first two categories. I will not discuss reinforcement learning as it is not yet relevant in sales. More information on this topic can be found in more specialized literature. In supervised learning, an output variable corresponds to a set of input variables. The algorithm tries to find a relationship between them so it would be able to make predictions about the output value when provided with input values. If you would want to predict if a potential customer will buy or not, you will first look at characteristics of

(15)

previous potential customers and look at their outcome, if they bought or not. The algorithm will find a relationship between the characteristics and the outcomes and as a result, will be able to predict if the new potential customer will buy or not based on his or her characteristics.

In unsupervised learning, we only receive input variables. The machine or computer needs to determine the structure or find patterns in these unstructured and sometimes unlabeled input data. When you want to classify your potential customers into clusters, groups with the same characteristics, you will use unsupervised learning to find patterns in these characteristics and thus determine the clusters.

In order for the Machine Learning algorithms to work effectively, there is a need for “substantial amounts of data (big data) and high processing power that is easily accessible” (Syam & Sharma, 2018, p. 2). Big Data is data that is high in volume, high in variety and high in velocity (McAfee & Brynjolfsson, 2012) . It means that big data consists of an enormous amount of data that comes in many different forms, ranging from numbers, categories to spoken or written text, and coming in at a very fast rate. These Big Data are used as input for the Machine Learning algorithms in AI applications.

So, to conclude, what makes AI so different compared to non-AI technology? Non-AI technology works based on an explicitly defined set of rules, whereas AI works and makes decisions based on incoming data (Crowston & Bolici, 2019).

In the following section, I will elaborate more on the different AI applications that are used today in the B2B selling process.

(16)

2.2 Artificial Intelligence in the B2B selling process

When discussing Artificial Intelligence (AI) in the B2B selling process, I first need to determine a framework that properly defines the different stages in this process. In this way, I can determine in which parts of the selling process AI applications already exist. One of the oldest and most widely accepted representations of the selling process is called the “seven steps of selling” (Dubinsky, 2013). These seven steps present the typical stages a salesperson needs to go through when selling a product or service: (1) prospecting, (2) pre-approach, (3) approach, (4) presentation, (5) overcoming objections, (6) close and (7) follow-up. This model has been an important basic framework in sales training and personal selling textbooks (Hawes et al., 2004). Therefore, I believe this will serve as a good basis for my research. Below I will briefly explain the seven steps of selling and elaborate on the implications AI has on them.

2.2.1 Prospecting

Prospecting is the first stage in the selling process and is closely related to the

customer development part of sales, also referred to as customer acquisition. Here, there is a strong interaction between sales and marketing as new potential customers are often attracted through marketing efforts.

D’Haen & Van den Poel (2013) describe the customer acquisition process by the sales funnel conceptualization (Figure 1). The sales funnel starts with a list of suspects, i.e. all potential new customers available. These are in theory all other companies that are active in a B2B context except for the companies that are already in the current customer base. In practice, this list is reduced to a limited amount of companies. The B2B marketeer then needs to select companies out of the suspect list based on a set of arbitrary rules. These selected companies become prospects, i.e. “suspects who meet certain predefined characteristics” (D’Haen & Van den Poel, 2013, p.2). Next, these prospects are qualified into leads, i.e. people who show interest in your product or service and have a high propensity to buy (Syam & Sharma, 2018). These leads will then be contacted by the salespeople. Eventually, every lead that buys the product or service becomes a customer.

(17)

Figure 1: Prospecting Situated in the Sales Funnel

The prospecting step consists of qualifying prospects into qualified leads. The first contact and everything that needs to be done until a lead becomes a customer, belongs to later steps in the selling process.

Previously, prospecting was seen as the most difficult and time-consuming part of the selling job (Moncrief & Marshall, 2005). “On average, sales reps spend 80 percent of their time qualifying leads and only 20 percent in closing” (Porter, 2017). This was necessary because prospecting and, more specifically, lead qualification was and still is one of the most important parts of the selling process. When you start with better qualified leads, you will obtain a higher customer conversion rate and as a result a lower cost of customer acquisition (D’Haen & Van den Poel, 2013).

Now, thanks to emerging technologies such as Artificial Intelligence and Machine Learning, the salesforce can build a predictive model to estimate if the prospect is qualified or not (Syam & Sharma, 2018). A sales force automation tool, powered by machine learning algorithms, can for instance compare characteristics of a particular prospect to the characteristics of the current customer base in order to predict if that

(18)

Of course, the quality of your qualified leads will heavily depend on the initial prospect list. Here, AI can contribute with “its ability to target customers with highly personalized, individually tailored advertising and marketing” (Syam & Sharma, 2018). As machines or computers can run continuously, they are able to compute

personalized campaigns and target a large number of prospects. In that way more

prospects will be contacted, more leads will be generated, and sales productivity will go up. In targeting prospects for instance, Potharst, Kaymak, & Pijls (2001) have found that if prospects are targeted through mail guided by a neural network, a Machine Learning algorithm, there is a 70% response rate, compared to a 30% response rate when targeting prospects through traditional mailing.

Nowadays, the prospecting step in the B2B selling process is often no longer a responsibility of the salesperson, but is rather conducted by others in the organization. It is more often done by marketing, whose ability to help salespeople is more and more supported by database marketing and CRM technologies (Moncrief & Marshall, 2005).

2.2.2 Pre-approach

After having identified qualified leads in the prospecting step, the salesperson arrives at the pre-approach step. In this step, he first of all needs to reach out to the lead to arrange a first meeting, which can be an actual visit or just a (video) call. Secondly, he needs to gather relevant information about the lead in order to prepare for this meeting. He needs to understand the customer’s needs and problems and prepare new and relevant information that could be interesting when approaching the customer. The information gathered in the pre-approach step allows the salesperson to adapt the actual sales presentation to the particular lead (Dubinsky, 2013).

Previously, the first contact with the prospect was made through intermediaries, letters, e-mails, phone calls, etc. Over the last ten years though, the use of mobile and web-based means through which the selling organization can contact the customers has rapidly increased. In the first place, you would think of video calling, contact forms on your company’s website or contact through social media platforms such as Twitter, LinkedIn or Facebook. Gathering information, the other part of the pre-approach step, has also become much easier over the years thanks to CRM platforms. They can

(19)

gather information out of many different customer touchpoints and centralize them in one place. This makes it much easier for the salesperson to access the necessary information.

AI has its stake as well. A lot of companies are already using chatbots, for instance. These are automated conversational partners, a machine or computer that can simulate a human conversion. As explained before, these chatbots rely on AI and Machine Learning algorithms. They can be used to uncover the needs of the lead and then even process the order and payment, especially for routine purchases. If the sales is more complex, the salesperson will have to take over (Syam & Sharma, 2018). AI can also help to decide on the optimal time to approach a lead. A Machine Learning algorithm can predict when a lead is ready to buy based on his visits on the company’s website, social media feeds and other third-party data.

So, nowadays, pre-approach is still an important step in the selling process, but is no longer an individual activity. CRM and AI driven technologies help the salesperson by providing a great amount and quality of information. AI can sometimes even replace the salesperson as a whole.

2.2.3 Approach

The next step in the selling process is called ‘approach’. It is situated in the first minute or minutes of the sales meeting. Here, the salesperson wants to get and hold the

attention of the lead. He needs to open the meeting by a little small talk, introducing

himself and by generally making a good initial impression (Dubinsky, 2013).

The approach step is a very human and emotional aspect of the B2B selling process. For now – as I have found – no literature exists on how AI can impact this stage. Most of the authors discuss this step together with the pre-approach step without going into detail. It is nevertheless an interesting topic for further research as the salesperson will be more and more replaced by AI driven technologies and this human touch will disappear.

(20)

2.2.4 Presentation

After the salesperson has determined the needs of the customer, has set up a meeting and has made a first approach, he arrives at the ‘presentation’ step in the B2B selling process. This step is the main part of the sales meeting. The salesperson explains the product or service to the lead, helps him to visualize the product through a sales presentation and demonstrates the product or service to reinforce that presentation. Previously only a small part of the sales presentations was specifically tailored to each lead (Dubinsky, 2013). This has completely changed, and AI plays an important role in that change. Sales enablement platforms like Showpad can recommend, using AI and Machine Learning algorithms, which content you should use to present to a specific lead based on content that has proven to be successful in the past for the same kind of leads. In that way presentations become specifically tailored to the characteristics

and interests of the lead and have a high probability of success.

Another major contribution of Artificial Intelligence to the presentation step, lies in the use of immersive technologies, e.g. virtual reality (VR), 360-degree video and augmented reality (AR). These AI-backed, immersive technologies can improve

customer learning by helping visualizing products that customers find hard to imagine

when it is not physically present (Suh & Lee, 2005). Furthermore, meetings are often conducted remotely through videoconferencing. Syam & Sharma (2018) state that evidence has shown that “immersive technologies drive higher user engagement than ‘plain’ videoconferencing by eliminating, or at least reducing, a major drawback of videoconferencing - the sense of the presenter not being present in the room” (p.9). These technologies can now be used in the cloud, which makes it even possible for smaller companies with limited budgets to deploy them in their sales presentations. Like in previous steps, AI can also replace the salesperson in the presentation step.

AI-driven chatbots can explain complex products to potential customers and answer

frequently asked questions. Compared to a human being, it can process a lot more information found in databases. As a result these chatbots are often even better at answering questions than a salesperson (Syam & Sharma, 2018).

(21)

2.2.5 Overcoming objections

In every sales presentation the salesperson needs to overcome objections. This happens when the potential customer has questions or hesitates about the product or company. It can delay the sales process, but must be seen in a positive way. When a lead reveals his objections, you are able to discover his true needs (Dubinsky, 2013). As discussed in the previous step, chatbots can replace the salesperson by answering frequently asked questions. Sometimes though, customers or leads may not be addressed by standard FAQs, especially for more complex and individualized products or services. Although AI-based communication has seen big advances, sales representatives will stay indispensable in this stage of the selling process (Syam & Sharma, 2018).

2.2.6 Close

After all objections have been successfully overcome, the salesperson arrives at the ‘close’ step. It can be defined as “the successful completion of the sales presentation culminating in a commitment to buy the good or service” (Moncrief & Marshall, 2005). It is seen as difficult for many salespeople, especially new salespeople.

There is not much literature on AI applications in the close step. When you want to close a deal, but do not have a fixed price for your product or offering, you can rely on AI and Machine Learning algorithms. They can compute a buyer’s reservation price based on his characteristics such as industry, size and prior relationship (Singh et al., 2019).

2.2.7 Follow-up

Eventually, we arrive at the last step in the selling process: follow-up. This is a relative newer addition to the seven steps of selling. It consists of two different processes: making sure the current order is successfully delivered and following-up after the current order (Syam & Sharma, 2018).

(22)

In the first place, AI can be used to automate all workflows that are used for order

processing. In the past, processing systems could handle the more routine tasks, but

today AI-backed systems can help in more complicated areas. It can search, for instance, in different databases for data recorded in different formats or containing errors and missing values.

A very relevant part in the selling process that is often forgotten, is the follow-up after the current order has been processed. This consists of receiving feedback from customers and providing after-sale customer service. Customer feedback can be helpful to identify problem areas and to discover ways to improve your offering (Wirtz et al., 2010). Very often, feedback is given in written text. Machine Learning algorithms can help to analyze the text and deduct the true meaning. In customer service, AI-powered chatbots are increasingly used by companies (Singh et al., 2019).

In Table 1 below, the different AI applications in the seven steps of selling that I have found in literature are summarized.

(23)

Table 1: AI Applications in the Seven Steps of Selling

Step in the ‘seven steps of selling’ AI applications

Prospecting Lead qualification

Customer targeting

Pre-approach Chatbots to uncover customer needs

Chatbots to process the order and payment for routine purchases

Algorithms to decide on the optimal time to approach a lead

Approach (no AI applications found in literature)

Presentation Tailoring presentations to the characteristics and

interests of the lead

Improving customer learning with immersive technologies like virtual reality (VR), 360-degree video and augmented reality (AR)

Chatbots to explain complex products to potential customers and answer frequently asked questions

Overcoming objections (no AI applications found in literature)

Close Computing a buyer’s reservation price based on his

characteristics such as industry, size and prior relationship

Follow-up Automating complicated workflows that are used for

order processing

Chatbots to receive feedback from customers and provide after-sale customer service

(24)

2.3 The changing role of the sales professional

The previous section clearly shows that Artificial Intelligence is present in many aspects of the selling process. It has automated a large number of day-to-day sales activities and is on its way to replacing salespeople in the future. Baumgartner et al. (2016) claim that with today’s technologies, 40% of all sales activities have the potential to be automated. The fundamental shift though will not be necessarily in the automation of simple, routine activities, but rather in the area of decision-making. Computers, driven by AI and Machine Learning, will have the ability to make reliable decisions as well. This will heavily impact the role of sales professionals (Syam & Sharma, 2018). The impact of AI on the salespeople’s job will strongly depend on the type of product

or service he is selling (Syam & Sharma, 2018). First of all, salespeople who are

selling simple or standard products or services will be heavily impacted. Simple products or services are easily understood with typically low margins. It can be sold by an inside salesman or online. In these kinds of products or services AI may even replace the salesperson. A second type of product or service is the one that has a high profit margin but is still easily accessible. Because of the high profit margin, salespeople can be easily paid. Nevertheless, AI can have a big impact on this category. The salesperson will not be entirely replaced but will have to work together with AI technologies. A last type of product or service deals with complex sales in which different buyer center members with different needs come in play. Here, customer needs may not have been explicitly stated and as a result salespeople will have to uncover what their true needs are. In this rapidly changing environment, salespeople will have to build a deep customer knowledge (Verbeke et al., 2011). This category is therefore expected to face a lower or more measured impact than the other categories. Some salespeople, depending on the type of product or service they are selling, will be more heavily impacted than others, but in the end everyone will have to learn how to co-exist with AI and other technologies (Syam & Sharma, 2018). Baumgartner et al. (2016) refer to this as “machine intelligence”. The salesperson needs to know when and how he needs to let the computers do the work and when and how to take over. Sometimes, things will get too complex and the computer or machine will hand over the work to the salesperson. Conversely, sometimes, the remaining part in the selling

(25)

process will be relatively standard or routine work and the salesperson will hand over the work to the computer, who will be able to execute that work way more efficiently. So, the “machine intelligence” means that the salesperson needs to fully grasp the capability boundaries of AI in sales. In that way he will be able to interact with the AI-driven technologies in a much more efficient way, like it is one of the members of the sales team (Syam & Sharma, 2018). If salespeople do not want to be set aside by AI, they need to truly understand this. Maycotte (2016) beautifully summarizes it as “It’s those who don’t or can’t adapt that are at risk of going the way of the dinosaurs”. Luckily, today, human beings cannot be entirely replaced by computers or machines, because these computers or machines lack the human touch, the emotional qualities of a beautiful, mistake-filled human (Porter, 2017). Compared to computers, human beings are able to deal with ambiguity and uncover the unstated needs of a customer or prospect. That is why they need to focus on “managing exceptions, tolerating ambiguity, using judgment, shaping the strategies and questions that machines will help enable and answer, and managing an increasingly complex web of relationships with employees, vendors, partners, and customers” (Baumgartner et al., 2016). Moreover, AI technologies can help to free time for that relationship building (Porter, 2017).

Artificial Intelligence will not only impact the activities of the salesperson but also these of management. First of all, when hiring sales representatives, management should not only look for empathetic personalities, for people with good relationship skills, but now also for salespeople who can “understand and interpret data, work effectively with AI and move quickly on opportunities” (Baumgartner et al., 2016). Secondly, AI gives the opportunity to train employees, salespeople in this case, much more efficiently. It can understand customer behaviour, provide feedback on the effectiveness of the sales presentation and as a result improve the sales skills of the salesperson (Singh et al., 2019).

It is clear that we cannot underestimate the impact of AI on the job of sales professionals. It does not necessarily mean though that salespeople will have to worry

(26)

see that AI can bring big opportunities in enhancing their performance. They need to understand the basics of AI technologies, know what AI can do and get hands on experience with existing AI tools. In that way they will uncover what AI can mean for them and how they can benefit from it.

In order for sales professionals to learn how to co-exist with AI, it is of utmost importance that AI applications are implemented as soon as possible. Unfortunately, a lot of companies are still reluctant to adopt AI in their selling process. I will elaborate more on this in the next section and will propose my research questions.

(27)

2.4 The adoption of AI in the B2B selling process

It is clear by now that AI can have a major impact on the B2B selling process. It will take over a lot of standardized procedures and will even have the ability to make reliable decisions on its own. If organizations and salespeople start to implement AI applications in a proper way right now and learn how to benefit from it, it will certainly enhance their overall performance.

Unfortunately, a lot of companies are still reluctant to adopt Artificial Intelligence in their selling process. Most of the time, they are still at the stage of gathering information with regard to adopting AI (Alsheibani et al., 2018), hindering sales professionals to start working with these AI applications. This poses some important questions on the

enablers and barriers to implement AI in the B2B selling process. Why are

companies not yet ready to adopt AI in their selling process? Until now, to the best of my knowledge, little to no research has been done on AI readiness and AI adoption

in the B2B selling process. So, in my dissertation I aim to determine which factors

have an impact on the AI readiness and AI adoption in the B2B selling process. I will focus on small, medium and large Belgian enterprises that are selling in a B2B context. I have defined my research questions as follows:

RQ 1: Which factors have an impact on the intention to adopt Artificial Intelligence in

the B2B selling process of Belgian enterprises?

RQ 2: How do these factors impact the intention to adopt Artificial Intelligence in the

B2B selling process of Belgian enterprises?

As very little research has been done on the adoption intention of Artificial Intelligence in the B2B selling process, my research is rather seen as exploratory research. In the next section I will, based on existing literature, build my conceptual framework and look for potential factors that have an impact on the adoption intention of AI.

(28)

3. RESEARCH MODEL AND HYPOTHESES

3.1 Conceptual framework

In order to determine the factors that have an impact on the adoption intention of Artificial Intelligence in the B2B selling process, I take into consideration two different technology adoption models, i.e. Technology Acceptance Model (TAM) and Technology-organizational-environmental (TOE) framework. This follows the approach of Gangwar et al. (2015) who did research on the determinants of cloud computing adoption by SMEs.

The Technology Acceptance Model (TAM), originally proposed by Davis et al.

(1989), is well known for successfully predicting and explaining users’ intention to adopt technologies. The two main constructs of the TAM model are called Perceived Usefulness (PU) and Perceived Ease of Use (PEOU). Perceived Usefulness (PU) is defined as the extent to which someone believes that his or her performance will enhance by using a particular technology. The technology acceptance and adoption will be higher when the Perceived Usefulness is higher (Ai-Mei Chang & Kannan, 2006). Perceived Ease of Use (PEOU) is defined as “the degree to which a person believes that using a technology will be free from effort” (Ai-Mei Chang & Kannan, 2006, p.2). The easier it seems to use a technology, the higher the likelihood to adopt that technology. To the best of my knowledge, there has not been done any research on AI adoption intention in the B2B selling process, but I believe that the TAM model will serve as a good basis for my conceptual framework.

Next to the TAM model, I will also use the

Technological-Organizational-Environmental (TOE) framework, originally proposed by (Tornatzky et al., 1990). It

is a framework used at the organizational level to explain factors that influence adoption decisions. Previous research (AlSheibani et al., 2018; Pumplun et al., 2019) has already used the TOE framework in explaining the AI readiness and adoption at the firm-level. It states that the intention to adopt a technology at the firm level depends on the technological, organizational and environmental readiness. The higher the readiness in these dimensions, the higher the intention to adopt AI at the firm level.

(29)

Unfortunately, the TAM model as well as the TOE framework have their limitations. Therefore researchers suggest to integrate these models in order to improve the predictive power of the resulting model and to overcome some of their individual limitations (Gangwar et al., 2015). For more explanations on the limitations of the individual models, I refer to the specific literature.

So, as proposed, I will integrate the TAM model and TOE framework into one

conceptual framework to explain the adoption intention of AI in the B2B selling

process. Factors of technological readiness and organizational readiness are expected to affect the adoption intention of AI using PEOU and PU as mediating factors. Also, the environmental readiness is expected to directly affect the adoption intention. Figure

2 shows the conceptual framework I will be using for my research. Next, I will define

the different factors more in detail and will propose my research hypotheses.

(30)

3.2 Research hypotheses

As explained in the previous section I will integrate the TAM model and TOE framework into one conceptual framework to explain the adoption intention of AI in the B2B selling process. Now I will specify the factors of technological, organizational and

environmental readiness and will propose my research hypotheses. Figure 3

shows the conceptual framework with the specific factors and hypotheses.

Figure 3: Conceptual Framework AI Adoption Intention with Research Hypotheses

3.2.1 Technological Readiness

Technological readiness refers to the extent to which a firm is able to adopt a new technology (Richey et al., 2007). When discussing the AI adoption intention in the B2B selling process, two factors of technological readiness are considered: relative advantage, and compatibility.

3.2.1.1 Relative Advantage

The relative advantage of a new technology is explained as the extent to which that new technology is perceived as providing better performance or benefit for firms

(31)

(Rogers, 2003). Rogers (2003) also outlined that the higher the perceived benefit of a new technology is, the more likely it will be adopted.

A few examples on how AI can enhance sales performance were given in the Literature Review section. In targeting prospects for instance, Potharst, Kaymak, & Pijls (2001) have found that if prospects are targeted through mail guided by a neural network, a Machine Learning algorithm, there is a 70% response rate, compared to a 30% response rate when targeting prospects through traditional mailing. Another example was given about AI-based immersive technologies. Syam & Sharma (2018) state that evidence has shown that “immersive technologies drive higher user engagement than ‘plain’ videoconferencing” (p.9). It is therefore clear that AI can offer a relative advantage in the B2B selling process, which leads to the following hypotheses:

H1a. Relative advantage has a positive effect on PU. H1b. Relative advantage has a positive effect on PEOU.

3.2.1.2 Compatibility

Rogers (2003) explains compatibility as “the degree to which an innovation is perceived as consistent with the existing values, past experiences, and needs of potential adopters” (p.240). It is perceived that more it is possible to implement AI applications in current sales processes, more the organization will benefit from these AI applications and more it will reduce uncertainty in using these applications. Furthermore, according to Pumplun et al. (2019), further factors must also be checked for compatibility. They say that AI applications need to have a real purpose. They will only be successful when tackling a clear business problem. So, more the AI application is compatible with current sales processes and with a clear business problem, higher the PU and PEOU will be. Therefore, I posit the following hypotheses:

H2a. Compatibility with current sales processes and with a clear business problem has

(32)

H2b. Compatibility with current sales processes and with a clear business problem has

a positive effect on PEOU.

3.2.2 Organizational Readiness

In addition to technological readiness, adoption intention of new technologies such as AI is also influenced by factors that are linked to the organization’s ability to implement it. When discussing the AI adoption intention in the B2B selling process, four factors of organizational readiness are considered: top management support, organization size, resources and sales process complexity.

3.2.2.1 Top Management Support

Top management support refers to “the engagement of a top-level leader for IS/IT implementations” (AlSheibani et al., 2018, p.5). New technology adoption can be highly influenced by top management commitment (Yang et al., 2015). It is perceived that more AI adoption is supported by top management, higher the PU and the PEOU will be. This leads to the following hypotheses:

H3a. Top management support has a positive effect on PU. H3b. Top management support has a positive effect on PEOU.

3.2.2.2 Organization Size

The adoption of innovation is directly affected by the size of the organization (Rogers, 2003). Is has been found that organization size has a positive effect on adopting new technologies (Aboelmaged, 2014). This is because larger organizations have more financial and technical resources available. Therefore, it is perceived that people working in a larger organization will show a higher PU and PEOU. Thus, the following hypotheses:

H4a. Organization size has a positive effect on PU. H4b. Organization size has a positive effect on PEOU.

(33)

3.2.2.3 Resources

Human, enterprise and information technology resources are also critical to adopting new technologies (AlSheibani et al., 2018). Aboelmaged (2014) refers to technology resources as computer hardware, data, and networking that are essential to adopt new innovation. Gangwar et al. (2015) argue that “firms those have effective infrastructure, expertise in their employees, and financial support increases the usefulness of the technologies” (p.7). So, when better human, enterprise and technology resources are available, the PU and PEOU will be higher. Thus, the following hypotheses:

H5a. Human, enterprise and technology resources have a positive effect on PU. H5b. Human, enterprise and technology resources have a positive effect on PEOU.

3.2.2.4 Sales Process Complexity

As explained in the Literature Review section, the type of product sold will also have an impact on the extent to which AI is used in the selling process. It is perceived that AI will have a bigger impact on the sales of simple or standard products or services. Sales of more complicated products or selling processes which involve different buyer centers will be less impacted by AI. Therefore, I posit the following hypotheses:

H6a. Sales process complexity has a negative effect on PU. H6b. Sales process complexity has a negative effect on PEOU.

3.2.3 Environmental Readiness

When adopting new technologies, organizations are also looking at environmental conditions (AlSheibani et al., 2018). In the case of AI adoption I mainly consider competitive pressure when evaluating the environmental readiness.

(34)

3.2.3.1 Competitive Pressure

Zhu & Kraemer (2005) define competitive pressure as “the degree of pressure that the company feels from competitors within the industry”. It is perceived that competitive pressure has a direct positive influence on technology adoption. This leads to the following hypothesis:

H7. Competitive pressure has a positive effect on AI adoption intention.

3.2.4 Perceived Usefulness and Perceived Ease of Use

Factors of technological readiness and organizational readiness are expected to affect the adoption intention of AI using PEOU and PU as mediating factors.

3.2.4.1 Perceived Usefulness

Perceived Usefulness (PU) is defined as the extent to which someone believes that his or her performance will enhance by using a particular technology. The technology acceptance and technology adoption will be higher when the Perceived Usefulness is higher (Ai-Mei Chang & Kannan, 2006). This leads to the following hypothesis:

H8. PU has a positive effect on AI adoption intention.

3.2.4.2 Perceived Ease of Use

Perceived Ease of Use (PEOU) is defined as “the degree to which a person believes that using a technology will be free from effort” (Ai-Mei Chang & Kannan, 2006, p.2). The easier it seems to use a technology, the higher the likelihood to adopt that technology. The TAM models also suggests that PEOU has a positive influence on PU as technologies that are easier to use can be more useful (Schillewaert et al., 2005). So, following hypotheses are proposed:

H9a. PEOU has a positive effect on AI adoption intention. H9b. PEOU has a positive effect on PU.

(35)

Table 2 summarizes the proposed research hypotheses:

Table 2: Research Hypotheses

Factors Hypotheses

Relative advantage _{H1a. Relative advantage has a positive effect on PU.}

H1b. Relative advantage has a positive effect on PEOU.

Compatibility _{H2a. Compatibility with current sales processes and with a}

clear business problem has a positive effect on PU.

H2b. Compatibility with current sales processes and with a

clear business problem has a positive effect on PEOU.

Top management support _{H3a. Top management support has a positive effect on PU.}

H3b. Top management support has a positive effect on PEOU

Organization size _{H4a. Organization size has a positive effect on PU.}

H4b. Organization size has a positive effect on PEOU.

Resources _{H5a. Human, enterprise and technology resources have a}

positive effect on PU.

H5b. Human, enterprise and technology resources have a

positive effect on PEOU.

Sales process complexity _{H6a. Sales process complexity has a negative effect on PU.}

H6b. Sales process complexity has a negative effect on

PEOU.

Competitive pressure _{H7. Competitive pressure has a positive effect on AI adoption}

intention.

Perceived Usefulness _{H8. PU has a positive effect on AI adoption intention.}

Perceived Ease of Use _{H9a. PEOU has a positive effect on AI adoption intention.}

(36)

4. RESEARCH DESIGN

4.1 Data collection

In order to test my hypotheses, explained in the previous section, I will take a quantitative approach by using a cross-sectional survey to collect the necessary data. To find appropriate respondents, i.e. sales professionals active in a B2B selling process in Belgium, I used LinkedIn Sales Navigator in combination with LinkedIn Helper. LinkedIn helper allowed me to automatically send a maximum of 100 LinkedIn connection requests per day. Eventually, I sent out LinkedIn connection requests to 1140 people in which I briefly introduced my research and asked them to connect if they wanted to participate. 418 of them accepted my request and received additional information about my research as well as the link to the survey. Eventually 84 people filled in the survey, which gives a response rate of 20.1%. After dropping the respondents who filled in the check question in the wrong way, I ended up with 77 good responses. This means I eventually had a sample size of 77. Table 3 shows some characteristics of my sample.

(37)

Table 3: Sample Characteristics Variable Category % # Gender Female 25% 19 Male 75% 58 Age 20 - 29 45% 35 30 - 39 30% 23 40 - 49 16% 12 50 - 59 9% 7 Industry Biotechnology 1% 1 Chemicals 3% 2 Communications 1% 1 Construction 1% 1 Consulting 5% 4 Education 1% 1 Electronics 4% 3 Engineering 1% 1 Entertainment 1% 1

Financial services and insurance 3% 2

Food & Beverage 5% 4

Healthcare 5% 4 Manufacturing 1% 1 Retail 1% 1 Technology 32% 25 Telecommunications 4% 3 Transportation 5% 4 Utilities 1% 1 Other 22% 17 Position Account manager 17% 13 Business developer 19% 15

Field sales manager 5% 4

Key account manager 12% 9

Sales representative 32% 25 Other 14% 11 # employees < 10 3% 2 10 - 50 29% 22 51 - 250 25% 19 251 - 500 4% 3 > 500 40% 31 Revenue (in million euros) < 1 8% 6 1 - 10 25% 19 11 - 50 19% 15 > 50 48% 37

(38)

4.2 Measures

The survey questions are based on the research of Gangwar et al. (2015) on understanding determinants of cloud computing adoption using an integrated TAM-TOE model. As they have used more or less the same conceptual framework, their research serves as a good basis for constructing my survey questions.

Respondents needed to evaluate different statements using a 5-point Likert scale. This follows the methodology proposed by AlSheibani et al. (2018). So, the respondents needed to indicate to which extent they agreed or disagreed with the different statements. They could choose between ‘strongly disagree’, ‘disagree’, ‘neither agree nor disagree’, ‘agree’ and ‘strongly agree’. The different constructs with their corresponding statements can be found in Table 4.Organization Size is the only factor that is not evaluated through statements. It is simply asked in the control questions at the end of the survey. The whole survey can be found in Appendix 1: Survey AI Adoption Intention

As Artificial Intelligence is a very broad term that has a lot of different definitions, I needed to be more specific when asking about AI applications. Therefore, I chose 4

specific AI applications that are already frequently used in practice today. You can

locate these applications in the prospecting, pre-approach and presentation step of the B2B selling process. It concerns the following applications:

- AI used in prospecting. AI helps in qualifying prospects into qualified leads. Algorithms analyze data on your potential customers, your prospects, and predict which ones have a high chance of buying your product or service, which are called qualified leads. It helps to decide which prospects you need to target and which leads to follow up on.

- AI-driven chatbots. Algorithms can analyze written text, understand your customer’s questions and respond accordingly.

- AI to prepare sales presentations. Algorithms can analyze presentations that have been successful in the past. In that way these algorithms can

(39)

recommend which content you should use to make a new successful presentation based on the characteristics and interests of your lead.

- AI-driven immersive technologies. Immersive technologies, e.g. virtual reality (VR), 360-degree video and augmented reality (AR) are based on AI algorithms. They can help visualize products or services.

Using these four specific AI applications made it easier for the respondents to evaluate the different statements.

(40)

Table 4: Factors and their Corresponding Statements

Constructs and their corresponding statements Code

Relative advantage

Using AI, we would be able to target prospects more effectively. (*)

Using AI, we would be able to find better leads (leads with a higher conversion rate). (*) Using AI, we would need to spend less time in finding qualified leads. (*)

Using AI, we would be able to reduce costs to find qualified leads. (*) Using AI-driven chatbots, we can uncover customer needs. (*)

Using AI-driven chatbots, we can explain our products or services to customers. (*) Using AI-driven chatbots, we can answer frequently asked questions. (*)

Using AI, we can tailor presentations to the characteristics and interests of the lead. (*) Using AI, we would be able to make better presentations. (*)

Using AI, we would need to spend less time in preparing presentations. (*)

Using immersive technologies (VR, AR, 360-degree video), we can improve customer learning. (*)

Using immersive technologies (VR, AR, 360-degree video), it would be easier to present our product or service. (*)

Using immersive technologies (VR, AR, 360-degree video), we would need to spend less time in explaining our product or service. (*)

RA_1_1 RA_1_2 RA_1_3 RA_1_4 RA_2_1 RA_2_2 RA_2_3 RA_3_1 RA_3_2 RA_3_3 RA_4_1 RA_4_2 RA_4_3 Compatibility

AI applications can be easily incorporated in the prospecting phase of our selling process. (*)

AI-driven chatbots can be easily incorporated in our selling process. (*)

AI applications to tailor presentations to the characteristics and interests of the lead can be easily incorporated. (*)

Immersive technologies can be easily incorporated in our selling process. (*) AI applications could help us solve problems or issues in our selling process. (*)

C_1_1 C_2_1 C_3_1 C_4_1 C_G_1

Top management support

My top management is closely involved in the selling process. (*)

My top management is likely to consider the adoption of AI in the selling process as strategically important. (**)

My top management is willing to take risks involved in the adoption of AI in the selling process. (**)

TMS_G_1 TMS_G_2 TMS_G_3

Resources

My company hires highly specialized or knowledgeable personnel to create and/or implement AI applications. (**)

We keep a lot of data on our customers. (*)

Our salesforce is used to work with technology in the selling process. (*) We use CRM platforms in our selling process. (*)

We use sales enablement platforms in our selling process. (*)

We allocate a percentage of total revenue for AI implementation in the selling process. (**)

R_G_1 R_G_2 R_G_3 R_G_4 R_G_5 R_G_6

Sales process complexity

My company sells a product or service that needs a lot of customization. (*) Our product or service is easily understood. (*)

It takes a lot of time to go through our sales process. (*)

When selling our product or service, we have to deal with different decision makers. (*)

SPC_G_1 SPC_G_2 SPC_G_3 SPC_G_4

Competitive pressure

We are aware of AI implementation in the selling process in our competitor organizations. (**)

We understand the competitive advantages offered by AI implementation in the selling process in our industry. (**)

CP_G_1 CP_G_2

(41)

In general, my company implements innovative technologies in order to outperform our

competitors. (*) CP_G_3

Perceived Usefulness

Using AI would allow us to do prospecting in an efficient way. (**)

AI-driven chatbots would allow us to uncover customer needs in an efficient way. (**) AI-driven chatbots would allow us to answer questions of customers in an efficient way. (**)

Using AI would allow us to prepare our sales presentation in an efficient way. (**)

Immersive technologies (VR, AR, 360-degree video) would allow us to explain our product or service in an efficient way. (**)

Using AI would allow us to close more deals. (*) Using AI would advance our competitiveness. (**)

PU_1_1 PU_2_1 PU_2_2 PU_3_1 PU_4_1 PU_G_1 PU_G_2

Perceived Ease of Use

The procedure of using AI in prospecting is understandable. (**) It is easy for us to learn using AI in prospecting. (**)

The procedure of using AI-driven chatbots is understandable. (**) It is easy for us to learn using AI-driven chatbots. (**)

The procedure of using AI to prepare sales presentations is understandable. (**) It is easy for us to learn using AI to prepare sales presentations. (**)

The procedure of using immersive technologies (VR, AR, 360-degree video) is understandable. (**)

It is easy for us to learn using immersive technologies (VR, AR, 360-degree video). (**)

PEOU_1_1 PEOU_1_2 PEOU_2_1 PEOU_2_2 PEOU_3_1 PEOU_3_2 PEOU_4_1 PEOU_4_2 Adoption Intention

Overall, I think that using AI in prospecting is advantageous. (**) Overall, I am in favor of using AI in prospecting. (**)

Overall, I think that using AI-driven chatbots in the selling process is advantageous. (**) Overall, I am in favor of using AI-driven chatbots in the selling process. (**)

Overall, I think that using AI to prepare sales presentations is advantageous. (**) Overall, I am in favor of using AI to prepare sales presentations. (**)

Overall, I think that using immersive technologies (VR, AR, 360-degree video) in the selling process is advantageous. (**)

Overall, I am in favor of using immersive technologies (VR, AR, 360-degree video) in the selling process. (**) AI_1_1 AI_1_2 AI_2_1 AI_2_2 AI_3_1 AI_3_2 AI_4_1 AI_4_2

(42)

5. ANALYSIS AND RESULTS

The main purpose of my analysis was to test the proposed research hypotheses. It was not ideal to use standard techniques such as OLS regression as I was faced with two difficulties. First of all, the underlying constructs in my conceptual model, e.g. Relative Advantage, Perceived Usefulness, Adoption Intention, etc., are latent

variables. Latent variables are variables that cannot be directly observed. Perceived

Usefulness, for example, cannot be measured by one single question. To measure these kinds of variables, you should use multiple questions, also called indicators. A second difficulty was the fact that I was dealing with multiple dependent variables in my conceptual model, i.e. Perceived Usefulness, Perceived Ease of Use and Adoption Intention.

To cope with these difficulties, I decided to use Structural Equation Modeling (SEM). Specifically, I used the Partial Least Squares (PLS) technique. The PLS technique has some advantages that are very useful in my case. It has the ability to model multiple dependent variables as well as multiple independent variables. It can also handle multicollinearity among the independent variables (Garson, 2016). Furthermore it is robust when faced with non-normality and small- to medium sample sizes (Chin, 1998). Next to that, PLS can create latent variables as linear combinations of the indicator variables (Garson, 2016). So, as I was dealing with multiple independent and dependent latent variables and had a small sample size of 77 respondents, this was an ideal technique to use.

Before starting the PLS analysis I did some data cleaning in R. More specifically, I dropped the responses that showed inconsistencies in the answers on questions about a specific construct. Responses that showed too much variance in these answers compared to the mean variance were dropped. I was left with 54 responses.

After cleaning my data in R, I used SmartPLS to perform the PLS analysis. My analysis was conducted in two phases. First, I conducted an assessment of the measurement model which included assessing the reliability and validity of the measures. Then, I tested the hypotheses by using PLS to implement a path model relating the predictors to the response variables. Figure 4 shows my conceptual model created in SmartPLS.

(43)

The blue circles represent the latent variables and the yellow rectangles their corresponding indicators. In order to estimate the outer loadings, path coefficients and some performance measures (measuring reliability and validity) I used the PLS algorithm available in SmartPLS. Furthermore, I was able to estimate the significance of the outer loadings and path coefficients by employing a bootstrapping procedure (1000 subsamples, bias-corrected, accelerated bootstrap, one-tailed).

Figure 4: Model in SmartPLS (General Model)

I first conducted the analysis on the general model, which includes the four different AI applications. Afterwards, I also did the analysis for the AI applications separately.

(44)

5.1 Assessment of measurement model (General)

First of all, I assessed the measurement model, also called the outer model. In

Figure 4 it is represented by the arrows going from the latent variables (blue circles) to

their indicators (yellow rectangles). To relate the indicators to their latent variables, I used a reflective model (the arrows are going from the latent variable to the indicators). It means that the indicators are a representative set of items which all reflect the latent variable they are measuring. When dropping an indicator the latent variable will still have the same meaning, because the other indicators are also representative. This is important to know, as it allows to drop an underperforming indicator without changing the meaning of the latent variable.

I evaluated the reliability and validity of the measurement properties by applying standard decision rules to test for indicator reliability, internal consistency reliability, convergent validity and discriminant validity. Afterwards I tried to improve the assessment model by dropping some badly performing indicators.

(45)

Table 5: Performance Measures of Measurement Model (General Model)

Construct Cronbach's alpha Composite reliability AVE Outer loading P value Indicator

0,699 0,000 RA_1_1 0,625 0,000 RA_1_2 0,603 0,000 RA_1_3 0,618 0,000 RA_1_4 0,587 0,000 RA_2_1 0,679 0,000 RA_2_2 0,638 0,000 RA_2_3 0,737 0,000 RA_3_1 0,752 0,000 RA_3_2 0,740 0,000 RA_3_3 0,811 0,000 RA_4_1 0,712 0,000 RA_4_2 0,748 0,000 RA_4_3 0,601 0,000 C_1_1 0,714 0,000 C_2_1 0,822 0,000 C_3_1 0,760 0,000 C_4_1 0,548 0,000 C_G_1 0,376 0,091TMS_G_1 0,949 0,000 TMS_G_2 0,930 0,000 TMS_G_3 0,955 0,000 nr_of_employees 0,921 0,000 revenu 0,877 0,000 R_G_1 0,588 0,000 R_G_2 0,785 0,000 R_G_3 0,649 0,000 R_G_4 0,669 0,000 R_G_5 0,834 0,000 R_G_6 0,790 0,075SPC_G_1 -0,243 0,282SPC_G_2 -0,246 0,285SPC_G_3 0,046 0,439SPC_G_4 0,657 0,000 PU_1_1 0,677 0,000 PU_2_1 0,770 0,000 PU_2_2 0,788 0,000 PU_3_1 0,758 0,000 PU_4_1 0,754 0,000 PU_G_1 0,802 0,000 PU_G_2 0,773 0,000 PEOU_1_1 0,769 0,000 PEOU_1_2 0,781 0,000 PEOU_2_1 0,766 0,000 PEOU_2_2 0,762 0,000 PEOU_3_1 0,746 0,000 PEOU_3_2 0,683 0,000 PEOU_4_1 0,705 0,000 PEOU_4_2 0,778 0,000 AI_1_1 0,777 0,000 AI_1_2 0,728 0,000 AI_2_1 0,699 0,000 AI_2_2 0,774 0,000 AI_3_1 0,763 0,000 AI_3_2 0,759 0,000 AI_4_1 0,774 0,000 AI_4_2 0,598 0,014 CP_G_1 0,949 0,000 CP_G_2 0,581 0,008 CP_G_3 0,573 0,532 0,915 0,763 0,478 0,485 0,636 0,881 0,549 0,187 0,556 0,561 Competitive Pressure (CP) 0,651 0,922 0,822 0,823 0,937 0,878 0,036 0,897 0,911 Perceived Usefulness (PU) 0,865

Perceived Ease of Use (PEOU) 0,888

Adoption Intention (AI) 0,893 Organization Size (OS) 0,867

Resources (R) 0,849

Sales Process Complexity (SPC) 0,675 Relative Advantage (RA) 0,908

Compatibility (C) 0,726