Data Collection

4.2.1 Non-Standardized Semi-Structured, Open-Ended Interviews

The single data collection technique was applied (mono-method) through non-standardized semi-structured, open-ended interviews, allowing broader freedom of researchers to adapt the process of questioning to each individual interview. Researchers follow a pre-developed interview guide consisting of pre-identified topics and prepared questions, but they may omit or adjust some questions depending on the context in relation to the research topic. Semi-structured interviews may be used for exploratory studies and are suitable for obtaining data in circumstances where you want your interviewees to explain or to build on their responses, where

questions are complex or open-ended, and where the logic and order of questioning may need to be varied (Saunders et al., 2009).

4.2.2 Interview Guide

The interview guide (see Appendix A) was designed to assist in the interviewing process with the objective of asking questions which would lead to authentic and truthful answers from the interviewees, in relation to the researched topic. In its introductory part, the interview guide comprised of the interview outline, stating the classification of interviewees, estimated time for the interview, and goal and format of the interview, preliminary information that should be provided to interviewees prior to the interview, and additional comments and instructions on how the interview guide should be used. Next, the interview guide contained introductory mapping questions which were designed to provide the context about a company that the interviewee was a representative of, and the primary questions consisting of three general questions and questions based on the ten concepts derived from the literature review (see chapter 3). For each concept (topic), several questions were constructed along with a few clarifying or backup questions, and several subtopics were mentioned to help researchers maintain a good overview of the identified concepts and the researched area during the interview.

4.2.3 Data Sources, Interview Setting and Interview Process

Interviews were conducted with eight employees from four different small and medium-sized companies; two employees from each company. All the participants were decision-makers, managers or senior employees.

After initial contact with each participant, an e-mail was sent informing about the researched topic and the format of the interview. As interviewees were representatives of companies that had no or little experience with AI technology, which was intended, a five minutes long Microsoft PowerPoint document was presented prior each interview to briefly explain the theories behind AI, what is the AI technology about and what are the most common use cases of AI. The presentation also summarized the research and its objective and the process of the interview. The PowerPoint presentation can be found in Appendix B.

Smartphones Samsung Galaxy S9 and S8 and the mobile app Otter.ai were used to record audio tracks of the interviews. All the interviews were between 44 to 63 minutes long and they were conducted during the months of May and June in 2019, thus the time horizon of this research was cross-sectional. The interviews were organized in a two-to-one (two researchers, one interviewee) face-to-face setting and directed by questions of the researchers, which is an approach defined as participant (respondent) interviews (Saunders et al., 2009).

Table 7 shows summary of the interviews, including the date of the interviews, role of the interviewees and the companies that the interviewees were representatives of.

Table 7 – Summary of the interviews.

Case Company # Date Interview

Length Role of the Interviewee A 1 07/5/2019 63 minutes Managing Director

2 10/5/2019 52 minutes Sales Manager B 3 20/5/2019 46 minutes Production Manager

4 21/5/2019 48 minutes Sales & Market, Graphic Advisor C 5 23/5/2019 50 minutes Head of Business Development

6 07/6/2019 51 minutes Senior Business Developer & Project Manager

D 7 06/6/2019 46 minutes Operations Controller, System Administration & Development 8 06/6/2019 44 minutes Communications & Personnel Development Manager

Table 8 presents an overview of the four case companies with their focus area, turnover in millions of EUR and size in terms of number of employees. Each of the companies is fully presented in its respective subchapter in the results (see chapter 6).

Table 8 – Summary of the case companies.

Case Company Industry / Focus Based In Turnover Employees

A Medical supplies and consumables Norway ~ €20 mil. 16

B Graphics, Printing, Product packaging, Promotional

items, Logistics Norway ~ €12 mil. 55

C Digital solutions consulting for the financial sector Denmark < €42 mil. 92

D Property and Facility Management, Caretaking,

Gardening, Inspections, Project-based services Denmark ~ €5 mil. 100

4.2.4 Transcribing

Audio tracks of the interviews were recorded using the AI-powered mobile app Otter.ai which automatically generated speech to text with relatively high precision. The app enables to listen to the audio track of an interview while indicating in real time which part of the text is being read. It allows moving across the timeline of an audio track according to a user’s position in the transcript. The app was therefore very helpful and timesaving in subsequent process of manually revising and correcting the transcripts. Finally, each of the revised transcripts was saved in a separated Microsoft Word document in a Google Cloud repository.

4.2.5 Pilot Test

A pilot test was conducted with the first interview. The aim of the pilot test was to assess whether the interview setting, the planned length of the interviews and especially the design of the interview guide and the choice of questions is suitable in practice, and whether the interviewee’s answers, induced by the questions prepared, are of sufficient value for the research question and its objectives, i.e., if the questions in the interview guide induce answers that are beneficial for the research in the context of barriers preventing SMEs to adopt AI technology. In addition to that, it was tested whether the quality of the audio recording of the interview is good enough, so that the Otter.ai app is able to auto-generate the transcript with high precision and so that the researchers can then recognize, when revising the transcript, what does the interviewee say.

The quality of the audio recording turned out to be high enough and after the transcript was generated and revised in Otter.ai, preliminary coding of the interview was conducted by researchers using Microsoft Word and the comments feature. Answers identified as potential barriers of adopting AI technology proved to be valuable, however, it was decided to amend the interview guide (see Appendix A) as it was too rigid, it contained too many detailed questions and asking easier questions proved to induce more interesting answers.

Therefore, two mapping questions were ruled out and the primary questions were simplified, resulting in one or two questions for each topic.