Case Company C

6.3.2 Future Outlook

The company is currently in the process of moving their outdated legacy systems to a modernized platform which will allow for more flexibility and use of microservices. The organization is already planning a next AI-based project that will improve one of their products but has no plans to use AI technology internally and for supporting purposes so far, and neither intends to hire internal human resources with competence in AI or ML.

If there were an intention to use AI internally, the Head of Business Development mentioned that a good investment in an AI solution could be in something that would eventually make operations cheaper, faster, more accurate or it would reduce manual and monotonous tasks for humans. They perceive AI more as “a tool in their toolbox” rather than a complete solution. When it comes to financial resources in relation to AI-based improvements in their products, there seem to be no issues as the funds for the projects flow from their customers who ask for the new or improved products.

6.3.3 Perceived Barriers to Adopting AI Technology

The 10 most important themes representing barriers identified in both the interview with the Head of Business Development (external Appendix H) and the interview with the Senior Business Developer & Project Manager (external Appendix I) are displayed below in Table 17 in respective contexts and logical categories within which were the themes identified, as explained in chapter 5.2.

Table 17 – Most important themes for case company C.

Technological Context Organizational Context Environmental Context Category: AI Black Box

▪ Lack of AI understanding

Category: Attention to AI

▪ Lack of AI understanding

Category: AI Expertise

▪ AI talent access

▪ Dependency on external help Category: Automation of

Tasks

▪ Unclear use case

Category: Legal and Policy Constraints

▪ Ethical consequences

▪ GDPR concerns

▪ Legislation, regulation and compliance constraints

▪ Regulator concerns Category: Human Resources

▪ Lack of AI competence

Category: Risk Perception

▪ Losing human supervision

Note: Numbers in parentheses (#) used in the following paragraphs refer to coded texts labelled with respective themes in the codebook in Appendix C.

6.3.3.1 Themes of Multiple Contexts

Lack of AI understanding (technological and organizational context)

The Senior Business Developer & Project Manager perceived that a potential barrier could be if people do not have the “knowledge about what the technology can do for you” (#264). The Head of Business Development also expressed that “people’s understanding of it is a big obstacle” (#265) and mentioned as an example that

6.3.3.2 Themes of Organizational Context Lack of AI competence

In connection to AI competence, the representatives indicated that the company substitutes the competence by outsourcing the knowledge and skills that are necessary to be able to execute and implement an AI project (#6-8, #21, #22). The Head of Business Development stated that the company “had a great experience with a small company that had the knowledge” (#7, #21). He also expressed that their employees “need to be educated a bit more than they are today, in order for us (the company) not to be vulnerable” (#26). The Senior Business Developer & Project Manager mentioned that “any discussion about how to crunch the data and which model (should be used) and which estimation, the principles (which) were best, that was entirely the vendor” (#22).

He said that he thinks the product owners should know about the technology to consider its use but he thinks

“it's hard to find time to educate them to a level where they can actually have those considerations” (#11).

Unclear use case

Both interviewees expressed that it is important to explore first what the technology should be specifically used for in an organization otherwise the return on investment might be at risk. “I think it's a good approach to see, what use cases do we have? And does machine learning actually fit in here? Otherwise, we will just, we risk spending time on the wrong technology.” (#93) “And then you have to know what the result, what (is) the machine trained for, what is the end? What is the result we're trying to achieve?” (#94) “But I think it's a mistake, just to start looking without really knowing what you want to achieve. So I usually advise people to think about the use case and think about the technology after they figured out which use case they want to pursue.” (#95)

6.3.3.3 Themes of Environmental Context AI talent access

Both representatives stated that hiring new employees with AI knowledge or AI developers might be difficult.

The Senior Business Developer & Project Manager said that they “would have a problem because AI developers are very, very popular so it's hard to get good AI developers” (#44) and therefore they “would have to pay quite a high price to convince them to work for us (the company)” (#47). The Head of Business Development mentioned that “if you had to go out and hire people with machine learning skills which are pretty short demand, it would have taken four months just to get someone to hire” (#45, #46).

Dependency on external help

Both interviewees mentioned several times that they will always need external services of vendors or developer companies as the company does not employ any developers, neither has any AI or machine learning-related competencies nor hosts IT systems in-house (#146-149, #158, #159, #161, #163, #164). The Senior Business Developer & Project Manager stated that they “will always need external help for any project, more or less”

(#161). The Head of Business Development talked about the vendors they cooperated with before and if they were to expand any solution, they would ask for their services again (#148, #158).

Ethical consequences

The Senior Business Developer & Project Manager said that “there are a lot of pitfalls when you start using AI” (#196) that might lead to ethical consequences or dilemmas which they presented to and discussed with the regulator, as the Head of Business Development also stated (#194-196). “One is, of course, bias - have we created a model that's more beneficial for certain segments of the Danish population? Does it give women an advantage compared to men and older, younger and so on?” (#196)

GDPR concerns

The interviewees mentioned potential data privacy concerns in relation to the cloud hosting services used by the company (#227-232). “Privacy concerns, certainly, because we have lots of property information.” (#231)

“We had some problems about cloud hosting, because we have GDPR that limits where you can put data, you need to make sure that they are within the EU…” (#232) “…if you have a sensitive model, the FSA will probably say, you can't give that data to Google or to Microsoft or to Amazon, it's customer sensitive data…” (#229) Legislation, regulation and compliance constraints

The Head of Business Development stated that they operate in a “regulated area […] so the technology exists, it works great, but you have to get it approved. And that's one of the obstacles that we've overcome.” (#281-283) The Senior Business Developer & Project Manager also mentioned “highly regulated sector” as a potential barrier given that the company consults solutions for the financial sector (#278). Both representatives indicated that due to this restriction, the company would first have to discuss a potential solution with the regulator before implementing it (#277, #279, #280-283).

Losing human supervision

The Senior Business Developer & Project Manager mentioned as a potential barrier elimination of humans in the entire process of how one of their solutions works because the regulator was not very happy about it and preferred to keep a real person included, who would be able to “pull the emergency brake” (#286, #287). The Head of Business Development stated that it could be problematic since “there's potentially no human eyes on it” and “the machine doesn't use common sense”, therefore it would not notice things in an evaluation process that humans would notice (#289).

Regulator concerns

Both interviewees stated a number of concerns of the regulator that might act as potential barriers to adopting AI technology. Since AI is not mentioned in the law and the people working for the regulator institution did not understand how the technology works, they had to be educated and convinced about the technology and its application (#335, #341). “They thought it was some kind of black magic or voodoo that happened inside the machine.” (#336) The regulator was also concerned whether the company, before implementing it, considered other important aspects such as security, ethical consequences, backup, version history and other things, and can guarantee that it is safe when the process and people are replaced by a machine (#337-340,

#342-344).