Abstract

The emergence of new technologies, such as artificial intelligence (AI), is transforming

society and changing fundamental beliefs about money, value judgments, and enterprise
implications. AI is becoming increasingly prevalent, serving as an exceptional advisor
and an omniscient informant. While AI offers great potential to tackle serious global
issues, such as climate change, food security, and healthcare, there are also potential
hazards and ethical concerns that require careful consideration.

The manufacturing industry is grappling with the complexities of improving efficiency,

but there is hope in the form of AI. The industry's intricate nature, with variations and
interactions among system members, presents challenges in streamlining processes.
However, the rapid transformation brought about by new technologies like AI offers
opportunities to enhance competitiveness and efficiency. AI can automate operations,
optimise production processes, and provide valuable insights that humans may struggle
to generate alone. By implementing AI in supply chain management (SCM), companies
can mitigate risks and reduce errors, delays, and wastage. AI can also contribute to
predictive maintenance, minimising downtime and costly repairs, while process
optimization can streamline operations and maximise productivity. Embracing AI in the
manufacturing sector holds immense potential for increased productivity, profitability,
and overall success. It is crucial to develop specific knowledge and understanding about
AI to ensure effective implementation and high-quality decision-making in the future.

Providing guidance for the future of AI is crucial, as it holds the promise of solving
significant problems and driving innovation. However, its implementation can also
present challenges for companies. Accurate forecasting and understanding the
implications of AI are essential for navigating this landscape. The successful usage of
AI hinges on effective implementation, particularly in complex environments like the
supply chain (SC). Therefore, the aim of this thesis is to provide insights and deeper
knowledge on how integrating AI into SCM can enhance its operations. By using a
qualitative method and the grounded theory in order to analyse the data collection, we
have discovered how the implementation of AI can improve SCM as well as its
drawbacks. The implementation of AI has the potential to revolutionise production
processes, streamline operations, and improve decision-making and forecasting, leading
to increased prosperity and cost savings for companies. However, it is important to
acknowledge and address several challenges and considerations to ensure the successful
implementation of AI. Mainly, the implementation of AI comes at the cost of complex
integration, time consuming strategizing and costly investments into the system.

Our results highlight the importance for companies to not only implement AI, but
integrate it, in order for successful utilisation. Unlike prior research, this study
highlights the dynamic nature and variations in SCM operations and the challenges
practitioners encounter. In addition, the study confirms previous findings on the positive
impacts of AI, such as enhanced productivity, cost reduction, and improved decision-
making. However, it emphasises the significant costs and time commitments involved
in implementing AI, creating decision-making obstacles for companies. This
underscores the importance of thoroughly evaluating the anticipated benefits of AI in
relation to the initial investment and time constraints.
1.0 Introduction

To commence the degree project, the problem background concerning the implementation of
AI and its potential impact on the manufacturing industry will be presented. This will be
followed by a discussion on the emergence of the research problem, which will lead to the
formulation of the research question. Lastly, the purpose of the degree project will be
presented, along with the focus, delimitations, and limitations that the thesis will entail.

1.1 Problem background

The widespread implementation of Artificial Intelligence (AI) in society has become

increasingly prevalent. The digitalization of currencies and the workforce further complicates
the future civilization, combined with other force such as the aging population and climate
change, making it difficult to predict with certainty. Nonetheless, it is clear that technology's
significance in our daily lives is increasing (Colbert et al., 2016, p. 731). The socio-materiality
of money, value judgments, and their implications for businesses are being transformed by
technology, resulting in a shift in our fundamental beliefs (Dodgson et al., 2015). Harari
(2017,
p. 354) labels this phenomenon "Dataism," which he characterises as a new religion that
extols the rising importance of technology. This highlights the extent to which technology has
replaced outmoded faiths, values, and worldviews. AI plays a vital role in this context,
serving as both an exceptional advisor and an all-knowing informant. AI has the capacity to
supplant human intuition, which has traditionally informed our decision-making processes
(Harari, 2017, p. 354; Glikson & Woolley, 2020, p. 627). According to several authors
(Goralski & Keong Tan, 2020; Vinuesa et al., 2020, p.6), AI is expected to play a crucial role
in addressing some of the world's most pressing issues, including climate change, food
security, and healthcare. However, as emphasised by Jobin et al. (2019), the future of AI also
presents potential risks and ethical concerns, such as biases, privacy, and accountability, that
must be considered to ensure ethical development and implementation. Therefore, despite
AI's enormous potential, it is essential to approach it with care and thoughtfulness to assess
and mitigate any potential consequences.

The potential of AI is significant, as noted by Cheatham et al. (2019, p.1). Intelligent systems
are increasingly assuming various aspects of organisational decision-making, including using
AI to improve strategic decisions (Rimol, 2022). In the manufacturing industry, significant
investments have been made towards AI and automation of processes, presenting an
opportunity to increase efficiency (Schaeffer et al., 2018, p.4-5; Bughin et al., 2017, p. 13).
The primary goal of AI is to develop intelligent machines and establish a definition of
intelligence (Schank, 1987, p.60). In practice, AI is a scientific discipline that enables
programs to improve by learning from their experiences and behaving intentionally,
intellectually, and adaptively (Schank, 1987, p. 64; Shubhendu & Vijay, 2013, p.28).

Intentionally, AI gathers information from multiple sources using sensors, digital data, or
remote inputs, analyses the data instantly, and acts on the insights obtained (Shubhendu &
Vijay, 2013, p.29; West & Allen, 2018, p.2). Due to significant advances in storage systems,
1
computing power, and analytic approaches, AI systems can perform complex analyses and
make informed decisions (West & Allen, 2018, p.2). From an intellectual perspective, AI
including machine learning (ML), examines data for underlying trends and if it detects
relevant patterns, software designers can use that information to evaluate specific issues
(Brynjolfsson & McAfee, 2017). The key requirement for ML is the availability of robust data
that algorithms can use to detect valuable patterns. Finally, AI systems can learn and
adapt as they make

2
decisions (West & Allen, 2018, p.3; Brynjolfsson & McAfee, 2017). Semi-autonomous
vehicles, for instance, have features that alert drivers and vehicles of approaching traffic
congestion, potholes, highway construction, or other potential traffic obstacles. Vehicles can
leverage the experience of other vehicles on the road without requiring human involvement,
and the entirety of their gained "experience" is instantly and fully transferable to other
similarly constructed vehicles. Therefore, the manufacturing industry widely believes that AI
will benefit both individual corporations and the industry as a whole (Schaeffer et al., 2018,
p.4).

Moreover, the widespread adoption of AI raises concerns about the displacement of human
workers, particularly in industries that are easily automated (Cheatham et al., 2019, p. 2;
Howard, 2019, p.921). As AI technology advances, it becomes increasingly capable of
performing complex tasks that previously required human expertise (Shubhendu & Vijay,
2013,
p. 28). For example, Merantix, a German company, uses deep learning to solve medical
problems and can identify lymph nodes in the human body in Computer Tomography pictures
(Rothe, 2017). Furthermore, AI systems can carry out complex tasks like driving a car and
creating an investment portfolio without active human intervention or supervision (Scherer,
2015, p. 363). Some researchers argue that AI's decision-making accuracy and logic are
designed to mimic or even surpass those of humans (Wirtz & Müller, 2018, p. 1085-1086;
Grace et al., 2018, p. 729), which has led experts to warn that AI could result in significant
job losses and increased economic inequality (Sharma et al, 2021, p.3). Moreover, some
believe that AI will outperform various fields in the future decades (Grace et al., 2018, p.
731), while others claim that AI will create new job opportunities, but the number of jobs lost
will outweigh the new job creation. These developments raise questions about accountability
and transparency, as well as the potential for unintended consequences (Scherer, 2015, p.
364).

In order to implement AI, a sizable network of linked computing units must be able to process
massive amounts of data using ML techniques (Wirtz & Müller, 2018, p. 1085). The
enforcement of AI within a corporation, according to Scherer (2015, p. 364 et seq.), results in
a loss of control on a number of fronts. A loss of control can occur through a variety of
methods, including: a malfunction, such as a corrupted file or physical damage to input
equipment; a security breach; the superior response time of computers over humans; or
defective programming (Scherer, 2015, p. 366). The fourth possibility poses the most
significant difficulties since it implies that a loss of control could be a direct but unexpected
consequence of a conscious design choice (Scherer, 2015, p. 366). If the AI is built with
capabilities that allow it to learn, adapt and act autonomous, it could be challenging to
maintain control if it is lost (Scherer, 2015, p. 366). Implementing AI may therefore pose a
risk to the general public and society (Scherer, 2015, p. 366). Due to the requirement for
nearly flawless modelling of highly nonlinear processes in a fast environment, applications of
AI in the manufacturing industries have proven to be particularly difficult (Kim et al., 2021,
p.125).

Despite being in its initial stages, AI has enormous potential as a modelling, analysis, and
3
automation technique that can shift the manufacturing paradigm in the near future, according
to the vast amount of recent literature investigating AI in related manufacturing industries
(Kim et al., 2021, p.125). AI has received extensive research attention in the fields of medical
image analysis, bioinformatics, drug discovery, recommendation systems, financial fraud
detection, processing of works of art, and military (Kim et al., 2021, p.125). Moreover (Pannu,
2015, p.79) claim that AI has the potential to revolutionise a number of industries, including
health care, finance, and transportation, by increasing productivity, lowering costs, and
enhancing decision- making. Which might explain why financial services, information and
communication

4
technology, healthcare and life sciences are all sectors that are ahead of the race towards AI in
comparison to the manufacturing industry (Schaeffer et al., 2018, p.6).

1.2 Arriving at the research problem

The manufacturing industry has emerging challenges with increasing its efficiency due to
complexity. Complexity refers to the variations and interactions of system members.
Industrial manufacturing can have complexity in both products and production, with the level
varying based on industry, product type, and operational strategies (Park & Okudan Kremer,
2015, p.216). Additionally, the manufacturing sector is currently undergoing a rapid
transformation as new technologies such as AI are being developed and used to add efficiency
and competitiveness in the business. In this instance, efficiency is economic and is determined
by comparing actual costs, revenues, and profits to what the production unit would ideally
want to achieve, subject, of course, to the relevant quantity and price restrictions. According
to Sharma et al. (2021, p.3) efficiency in businesses with AI is important for the
manufacturing sector because it can lead to certain benefits such as reducing operational
costs, increasing production speed, improving product quality, and lastly enhancing decision-
making processes. Liu et al. (2022, p.1) continues that efficiency in the manufacturing
industry is critical due to the fact that it directly affects the productivity and the profitability in
companies. By productivity we mean the ratio of input to output. Assuming the unit produces
a single output via a single input, thus this ratio is simple to determine (Knox Lovell, 1993, p.
3). Even though the industry is heavily dependent on efficiency, they still have trouble
increasing it, due to its very complex process to streamline because the process involves
multiple stages (Park & Okudan Kremer, 2015, p.215).

The manufacturing industry relies on a complex network of suppliers and distributors, which
can contribute to inefficient supply chain management (SCM). Based on Coyle’s et al., 2017)
book, SCM aims to examine and manage supply chain (SC) networks. An essential goal is to
increase a company's competitiveness in the global marketplace in the face of fierce
competition and rapidly changing client requirements (Coyle et al., 2017). Some also argue
that the goal is to increase throughput while simultaneously reducing both inventory and
operating expenses (Hugos, 2018, p. 7). Additional scholars contribute to the definition of
SCM, saying that SCM is essentially an ensemble of procedures designed to coordinate and
manage the whole SC, from the suppliers of raw materials to the final consumer (Heikkilä
2002, p. 749; Melo et al., 2009,
p. 401). SCM focuses on improving processes as a whole rather than local optimization of
specific business units (Heikkilä, 2002, 749). However, due to the complexity within SCM,
there is a higher likelihood of errors, delays, and wastage, which can all impede productivity.
Reducing complexity can help to integrate the production processes, minimise errors and lead
to lower costs and increased efficiency (Liu et al., 2022, p.2). Naturally, when SCM is
inefficient, it can lead to setbacks, increased costs, and reduced customer satisfaction (Clarke-
Potter, 2019).

SCM can mitigate these possible risks through five areas: production, inventory, location,
transportation, and information (Hugos, 2018, p. 7). Production refers to a SC ability to

5
produce and store goods. Factories and warehouses serve as production facilities and
companies can therefore work with a lean production approach in order to increase efficiency
in the production (Fairris & Tohyama, 2002, p. 529). The same goes for inventory efficiency,
the lean inventory philosophy (Mishra et al., 2013, p. 300) can increase efficiency, since it
views excess inventories as waste and focuses on fostering inventory efficiency in companies.
Companies' inventories reflect their locked-up capital. As a result, increased inventory
efficiency can minimise the working capital required by companies to continue their
operations, resulting in

6
stronger cash flow situations (Mishra et al., 2013, p. 301). Furthermore, companies with low
inventories, i.e., high inventory efficiency, are less susceptible to inventory write-offs due to
obsolescence (Hendricks & Singhal, 2009). For Location a common facility limited location
problem involves a collection of geographically scattered clients and a collection of facilities
to meet customer requests (Melo et al., 2009, p. 401). Furthermore, companies can use a given
metric to measure the distances, timeframes, or expenses between consumers and facilities
and therefore promote a better view of efficiency. The following questions should be
addressed by companies in order to improve their efficiency regarding location: (i) Which
facilities should be utilised? (ii) Which consumers should be served by which facility (or
facilities) in order to reduce total costs? (Melo et al., 2009, p. 401). Regarding company
transportation, it may play a critical integrative role in the structure of SC when companies
strategically compete on the basis of cost, service, or time (Morash & Clinton, 1997, p. 5).
Here companies work mainly with being reliable and just-in-time principle (JIT) for the
purpose to increase efficiency (Morash & Clinton, 1997, p. 5; Aityassine et al., 2021).

Lastly, information within SCM serves as an essential approach for the survival of companies
as well as an enabler for them to integrate (Lotfi et al., 2013, p. 298). Sharing information in a
SC can be advantageous for a company in several ways (Lotfi et al., 2013, p. 300). For
example, the products more closely fit customer demand, and market shifts may be predicted.
Currently, companies share information through the following categories: (1) Inventory
Information; (2) Sales Data; (3) Sales Forecasting; (4) Order Information; (5) Product Ability
Information; (6) Exploitation Information of New Products (Lotfi et al., 2013, p. 300).
However, further research is needed to determine how and what information should be
provided, as well as the advantages of improvements in quality (Lotfi et al., 2013, p. 300;
Boussehaib & Belhsen, 2023). Thus, companies struggle with the underlying areas that
companies work with to mitigate inefficiency. Within the areas SCM can improve the
efficiency can technology be implemented to enhance their capability (Fairris & Tohyama,
2002; Mishra et al., 2013, p. 298; Melo et al., 2009).

The manufacturing sector is one of the key industries that can benefit significantly from AI,
particularly in areas such as SCM (Sharma et al, 2021, p.5). Despite AI being a widely
accepted tool to help in decision-making, SCM has applied AI very limited (Min, 2010, p.
20). Additionally, relevant SC literature is also limited, despite some recent efforts to include
current AI methodologies into its research, in comparison to other organisational fields
(Pournader et al., 2021, p. 1). Often, SC Managers are interested in assessing the practicality
and the applicability of the suggested AI technology (Min, 2010, p. 19). However,
approximately 37 % of companies, looking at all industries, are searching to outline their
strategies with AI, and 35
% are struggling to find the right application of AI (Pournader et al., 2021, p. 2). For SCM
this plays a huge role, mainly because a single item of inventory might cost between 15 % to
35 % of its product value annually for a company to store (Timme & Williams-Timme, 2003).
As a result, a company's performance in a competitive market frequently depends on its
capacity to manage and organise inventory at the lowest possible cost while having it
constantly accessible to customers (Min, 2010, p. 20). This capacity can be strengthened by
having precise, real-time information about predicted client requests, inventory quantity and
7
type, and order cycle time to fulfil the customer order (Min, 2010, p. 20). Traditional decision
criteria based on mathematical models, such as economic order amount, cannot, however,
capture the core of inventory management since this type of information is frequently difficult
to assess, predict, and get (Min, 2010, p. 20). That is, a tool like AI, which can substitute
experienced inventory managers' strong judgement and intelligence and deal with the
unexpected, is more suited to handle inventory control and planning choices. Hence, AI
should be implemented within SCM

8
in order for the manufacturing industry to increase their margins and become even more
competitive.

Various AI techniques can be applied to reduce complexity and increase efficiency, such as
process optimization (Waltersmann et al., 2021, p.22). AI has the potential to simplify
complex processes and provide insights that would be difficult or impossible for humans to
generate alone (Azzam & Beckmann, 2022, p.4). If companies have a well-planned AI and is
implemented accurately it can develop and advance ways to manage daily operations and
improve the efficiency in the company (Sharma et al, 2021, p.12). Therefore, effective
implementation of AI in the manufacturing sector can bring significant improvements in
productivity and profitability and contribute to the industry's overall growth and success.
Further, with the significant attention AI is receiving and the significant investment being
made in it, its implementation becomes crucial (Strohm et al., 2020, p. 5530; Schaeffer et al.,
2018, p.4-5; Mishra et al, 2013, p. 298). Some applications of AI have failed to achieve
widespread adoption. Previous literature claims this is due to low technical performance,
whereas other authors (Strohm et al., 2020, p. 5526; Frank et al., 2019, p. 80) say that we
should look at additional implementation barriers, such as organisational or social
characteristics, to better understand the reason behind failure implementations. If
underutilisation would appear, Boucher (2020) argues that it may stem from public and
business mistrust of AI, poor infrastructure, lack of initiative and low investment, or, as AI's
ML relies on data, from a fragmented digital market. The factors behind successful
implementation as well as failed implementation are therefore not evident.

To optimise AI within the manufacturing industry, more specific knowledge about the subject
needs to be developed. AI is a general term for machines that have the ability to perceive,
logic and learn. ML is a branch of AI and computer science which focuses on the use of data
and algorithms to imitate the way that humans learn, gradually improving its accuracy. It
addresses the question of how to build computers that improve automatically through its
experience (Jordan & Mitchell, 2015, p. 255). In the concept of ML, the machine, after
processing the data, intelligently extracts the patterns in them, learns them and turns them into
knowledge (Ghahremani Nahr et al., 2021, p. 38; Glikson & Woolley, 2020, p. 628). To
conclude, ML is the study of how a computer may learn directly from data and so learn to
solve problems (Ratner, 2000). Because of the sheer volume of data generated by both
humans and machines, it is impossible for humans to comprehend, understand, or make
complicated judgments using this data (Reim et al., 2020, p. 183). ML is therefore one of the
most essential parts of AI, since it enables machines to ease the burden humans currently have
with processing large volumes of data.

However, the range of AI research is very wide and not that specific (Ghahremani Nahr et al.,
2021, p. 38). AI is the facilitator to making difficult decisions in the future (Ghahremani Nahr
et al., 2021, p. 45). Therefore, the evolution in AI, ML and new knowledge of automatisation
is essential for business decisions. ML can improve efficiency through predictive maintenance
which means that ML algorithms can analyse data from sensors and machines to predict when
equipment is likely to fail, allowing maintenance to be performed proactively before a
breakdown occurs (Pech et al., 2021, p.1). This reduces downtime and prevents costly repairs.

9
It can also be through process optimization where AI can be used to optimise production
processes and minimise downtime (Weichert et al., 2019, p.1893). AI (and indirectly ML) has
the potential to automate numerous commercial operations, drastically reduce the need for
human labour, boost an organisation's productivity, and save both time and money
(Ghahremani Nahr et al., 2021, p. 45; Pech et al., 2021, p.1; Rimol, 2022). One must therefore
develop more

10
specific, applicable comprehension about AI in order to ensure high quality decisions will be
made by machine in the future.

1.3 Research question

Given the identified research gaps and the background information on the topic, the primary
research question that this degree project seeks to address is:

- How can implementation of Artificial Intelligence improve Supply Chain

Management?

1.4 Purpose

The purpose of this study is to contribute with guidance regarding the future of AI. AI is
expected to solve large issues and facilitate new innovation, but will also result in challenges
for companies, therefore accurate forecasting regarding AI can be highly beneficial (Scherer,
2015, p. 358; Grace et al., 2018, p. 729). The implementation is the catalyst for successful
usage of AI, therefore the goal is to provide insights and deeper knowledge of how
implementation of AI within a complex environment like the SC, can improve its SCM. As
already mentioned, organisations usually lack the skills and knowledge necessary to deploy
and use AI and underutilisation can hamper a company largely (Grace et al., 2018, p. 733).
Our essay therefore also aims to help businesses become more knowledgeable so they can
handle any problems that come up when implementing AI. As a result, we aim to increase
successful implementations of AI within the manufacturing industry and focus our research
on minimising the potential challenges AI might bring during implementation. In addition, we
also aim to establish a contribution to the long-term benefit implementation of AI has, as well
as challenges. Alternatively, debate potential solutions and shed light on the societal
difficulties that may come from this.

1.5 Focus, delimitation, and limitation

This thesis will exclusively focus on analysing and researching the manufacturing industry
due to its potential for significant growth and development with the implementation of AI.
This choice was made based on the fact that according to Accenture, AI will boost the
industrial sector by over US$3.7 trillion by 2035 (Schaeffer et al., 2018, p.5). Further this
sector was selected in order to capture an industry that is still in the experimental stage
regarding leveraging AI (Schaeffer et al., 2018, p.6) compared to other sectors like the
healthcare sector and the transportation sector that already has wider research and also are the
leaders in the AI race.

Furthermore, we have delimited this degree project to analyse only the implementation of AI
and not the maintenance or reconstructing part that might follow after one has implemented
AI. This is mainly because the implementation part of AI is crucial and complex for
companies to do (Strohm et al., 2020, p. 5526) and we wanted to gain deeper knowledge and
understand how companies in the manufacturing industry are facing this challenge when
implementing AI. Numerous articles and authors explore the manufacturing process and
11
mention the implementation part of AI as an important step to take to stay competitive
(Sharma et al, 2021, p.3; Strohm et al., 2020, p. 5526; Thun et al., 2021, p.727). Due to the
considerable attention and investment AI is receiving, its adoption has become increasingly
significant (Strohm et al., 2020, p. 5530; Schaeffer et al., 2018, pp. 4-5). However, these
analyses usually emphasise that implementation is crucial and an important future tool and
rather than delving into the specifics of how the current adoption of AI can enhance SCM
and the methods utilised to achieve it.

12
 Furthermore, there is an absence of scientific evidence supporting the impact that the
implementation of AI has regarding SCM. Our aim is to bridge this gap by offering insights
that can be extended to other industries beyond our chosen sector, thereby contributing to the
broader understanding of the subject.

Lastly we limit our study to a certain geographical area, Sweden. This is mainly because the
manufacturing industry in Sweden is extensive and contributes significantly to the national
economy. According to Armelius (2022), this sector alone accounts for nearly 20 percent of
the business Sweden’s value added, which is included in the gross domestic product (GDP).
We also chose Sweden in order to maximise the number of shared factors among participants
to be able to analyse and evaluate their response based on their common background.

Disadvantages of AI

1. High Costs

The ability to create a machine that can simulate human intelligence is no small feat. It requires
plenty of time and resources and can cost a huge deal of money. AI also needs to operate on the
latest hardware and software to stay updated and meet the latest requirements, thus making it
quite costly.

2. No Creativity

A big disadvantage of AI is that it cannot learn to think outside the box. AI is capable of
learning over time with pre-fed data and past experiences, but cannot be creative in its approach.
A classic example is the bot Quill who can write Forbes earning reports. These reports only
contain data and facts already provided to the bot. Although it is impressive that a bot can write
an article on its own, it lacks the human touch present in other Forbes articles.

3. Unemployment

One application of artificial intelligence is a robot, which is displacing occupations and

increasing unemployment (in a few cases). Therefore, some claim that there is always a chance
of unemployment as a result of chatbots and robots replacing humans.

For instance, robots are frequently utilized to replace human resources in manufacturing
businesses in some more technologically advanced nations like Japan. This is not always the
case, though, as it creates additional opportunities for humans to work while also replacing
humans in order to increase efficiency.

13
4. Make Humans Lazy

AI applications automate the majority of tedious and repetitive tasks. Since we do not have to
memorize things or solve puzzles to get the job done, we tend to use our brains less and less.
This addiction to AI can cause problems to future generations.

5. No Ethics

Ethics and morality are important human features that can be difficult to incorporate into an AI.
The rapid progress of AI has raised a number of concerns that one day, AI will grow
uncontrollably, and eventually wipe out humanity. This moment is referred to as the AI
singularity.

6. Emotionless

Since early childhood, we have been taught that neither computers nor other machines have
feelings. Humans function as a team, and team management is essential for achieving goals.
However, there is no denying that robots are superior to humans when functioning effectively,
but it is also true that human connections, which form the basis of teams, cannot be replaced by
computers.

7. No Improvement

Humans cannot develop artificial intelligence because it is a technology based on pre-loaded

facts and experience. AI is proficient at repeatedly carrying out the same task, but if we want
any adjustments or improvements, we must manually alter the codes. AI cannot be accessed and
utilized akin to human intelligence, but it can store infinite data.

Machines can only complete tasks they have been developed or programmed for; if they are
asked to complete anything else, they frequently fail or provide useless results, which can have
significant negative effects. Thus, we are unable to make anything conventional.

2.0 Scientific methodology

In this chapter, we will begin by discussing how we chose our topic and our prior knowledge
on the subject. Then, we will delve into research philosophy and explore our philosophical
beliefs. Finally, we will examine our research approach and design to provide a better
understanding of how our degree projects have taken shape.

2.1 Choice of subject

The selection of this subject mainly comes from management courses at

14
civilekonomprogrammet at Umeå University. From these courses, we sparked interest in how
management works in different industries to make companies better and felt the need to dive
into the subject more. We started to search about this subject in order to find a gap in the
literature. We found an interest in AI and that it is an emerging trend that has developed as a
transformative technology in recent years, with the potential to revolutionise many industries
and management in industries, including manufacturing. Therefore, the selection of AI as a
subject in our thesis reflects the importance and the potential benefits it can offer to the
manufacturing industry. By examining how the implementation of AI can improve SCM in
the manufacturing industry, we aim to contribute to a better understanding of this technology
and provide valuable insights for businesses looking to adopt it. Furthermore, as we delved
deeper into the subject of AI and its potential applications in manufacturing, we realised the
need for more research in the SC. The SC is characterised with complexity and touch upon
several business areas within a company. By improving SCM, hopefully companies will be
able to improve the organisation at large. Finally, it is interesting and significant to contribute
to the advancement of this field at both an individual and societal level.

3.0 Theoretical framework

The theoretical framework created for the thesis focuses on understanding and analysing the
impact of Artificial Intelligence (AI). It includes four key areas: (1) AI technology, (2) AI
implementation processes, (3) potential effects on SCM, and (4) potential effects on the
manufacturing industry. The literature was thoroughly evaluated, and gaps were identified.
The chapter was structured to align with research objectives and a connection was established
between AI and SCM. The context of the manufacturing industry was also considered.
Overall, the theoretical framework provided a comprehensive understanding for the thesis.

3.1 Breaking down the concept of Artificial Intelligence

In 1955, John McCarthy introduced the term "artificial intelligence" to explore whether
machines could solve problems and use language as effectively as humans (Helm et al., 2020,
p.69; Min, 2010, p.14; Pournader et al., 2021, p.2). AI involves developing computer systems
that can perform tasks that typically require human intelligence, including learning, reasoning,
problem-solving, and decision-making. To achieve this, AI is equipped with the ability to
learn and understand new concepts, learn from experience, engage in reasoning, draw
conclusions, and interpret symbols in context (Pournader et al., 2021, p.2-3). These
capabilities have enabled AI's successful use in various domains such as game playing, human
performance modelling, machine learning, data mining, genetic algorithms, and expert
systems (Min, 2010, p.14). Additionally, AI has found extensive use in robotics, where it
helps create machines capable of performing dangerous or difficult tasks beyond human
capacity. For instance, AI-powered robots can be used in manufacturing, healthcare, and
research environments for tasks such as data collection, surveying, and patient care (Min,
2010, p.14).

By harnessing the power of AI, organisations can automate complex tasks, improve
efficiency, and gain valuable insights into their data. As AI technology continues to develop
and improve, we can expect to see even greater benefits and applications in the years to come.
15
In fact, the manufacturing industry is already experiencing the advantages of AI. It is evident
that AI is contributing to increased productivity, reduced costs, and improved decision-
making processes within the industry (Pannu, 2015, p.79). However, the adoption of AI is still
limited, compared to other industries. The use of AI technologies such as expert systems (ES),
machine learning (ML), and genetic algorithms (GA) is becoming increasingly prevalent in
various applications in the manufacturing industry (Zeba et al., 2021, p.2; Min 2010, p.20;
Wuest et al., 2016, p.24).

3.1.1 Expert System

One of the most important techniques that are used when implementing AI in the
manufacturing industry is ES (Zeba et al., 2021, p.2). ES is referring to computer programs
that have their own decision-making abilities to address a specific problem. ES is a powerful
tool for maintaining the vital knowledge necessary for manufacturing competitiveness. They
are commonly used in organisations for knowledge dissemination and training purposes. ES
enables a simplified transfer of knowledge, which results in minimal costs (Leo Kumar, 2018,
p.4767). In the manufacturing industry, ES can be used to make decisions related to quality
control, scheduling, and process control. For example, an ES can be used to detect defects in
products on the production line and suggest corrective actions to operators. ES can also help
in optimising production schedules, by analysing production data and providing suggestions
for process improvements (Min, 2010, p.16-17).

16
Additionally, ES can aid in predictive maintenance, by analysing data from sensors to predict
when maintenance is required. According to research, ES demonstrates high performance in
domains where human intelligence can be effectively formalised and structured (Pournader et
al., 2021, p.4). An ES is usually composed of four main components. The first one is
knowledge base, which is a collection of rules, facts, and knowledge that has been acquired
from a human expert. This repository of information is essential for the ES to make decisions
(Min, 2010, p.17). The second component is the inference engine and is considered the
"brain" of an ES as it is responsible for coordinating the search, reasoning, and inference
processes based on the rules and information in the knowledge base. The third one is the
scheduler and is the component that is responsible for coordinating and controlling the
sequencing rules within the ES. The final component of an ES is the user interface, which
allows for communication and interaction between the system and its user through a course of
user queries (Min, 2010, p.17). ES are particularly useful in practical SC problems because
they operate using concepts and terminology that are familiar to the user. This makes the
system more easily accepted by practitioners and therefore more applicable to real-world
situations (Min, 2010, p.17). Incorporating ES into the manufacturing processes can offer
significant support to operational workers when executing and managing crucial and
significant tasks.

3.1.2 Genetic Algorithms

The use of genetic selection principles in optimising search tools for complex problems is
referred to as a GA (Min, 2010, p.17; Katoch et al., 2020, p.8094). GA not only serves the
purpose of optimization, but also aids in ML, research, and development (Lambora et al.,
2019, p.380). GA is a type of AI method that uses a search process to find a solution and is
useful in finding the shortest path in a graph by encoding the path as chromosomes. For the
algorithm to work, Lima et al. (2020, p.2) states that two things should be defined. The first is
a fitness function that calculates the best solution to the given problem. The second definition
is a function that can represent "DNA" for the selected solutions. GA is a method of solving
complex problems by mimicking the process of natural selection, where the "fittest"
individuals are selected and their traits are combined and mutated to create new, potentially
better solutions. Using genetic algorithms can provide an optimal solution with high accuracy
and in a very short time (Lambora et al., 2019, p.380). The GA generally consists of five
components, (1) A genetic representation of potential solutions to the problem, (2) A way to
create a population, (3) An evaluation function measuring the fitness of solutions to see
whether they will survive, (4) Genetic operators that alter the genetic composition of offspring
and (5) parameter values that determine population size (Min, 2010, p.17-18). The five
components of a GA work together to optimise solutions to a given problem. Overall, using a
GA can provide highly accurate solutions in a short amount of time, making it a valuable tool
in research, development, and ML.

3.1.3 Machine learning

ML has proved to be effective in optimising processes, monitoring and controlling operations,
and enabling predictive maintenance across various industries, particularly in the
manufacturing sector and it is one of the most promising improvements in the industry
(Dogan & Birant, 2021, p.2; Wuest et al., 2016, p.27). Successful appliance of ML is

17
contingent on a change of task characteristics and contextual factors of work activities
(Brynjolfsson et al., 2018, p.45). The majority of advancements and applications in AI are
attributed to a type of algorithms referred to as ML, which also includes deep learning
(Brynjolfsson et al., 2018, p.43). Considered a subset of AI, ML exhibits the experiential
“learning” associated with human intelligence, while also having the capacity to learn and
improve its analyses through the use of computational algorithms (Helm et al., 2020,
p.69). ML algorithms can analyse data on

18
inventory levels, demand forecasts, and other factors to ensure that the right materials are
available when they are needed, minimising waste and reducing costs (Papageorgiou, 2009,
p.1931-1932; Pournader et al., 2021, p.7). For example, based on data, the algorithm predicts
that there will be a surge in demand for the product in the next month. The algorithm also
analyses the availability of raw materials needed for production and predicts that there may be
a shortage in the coming weeks. Using this information, the company takes proactive
measures to order more raw materials and adjust their production schedule to ensure that they
have enough stock to meet the demand without overstocking and wasting materials. By
leveraging ML, manufacturers can identify patterns and anomalies in data that humans may
miss, leading to more efficient and effective decision-making (Pournader et al., 2021, p.7). As
such, it is no surprise that the market for ML in manufacturing is expected to grow
significantly in the coming years (Pournader et al., 2021, p.7). In addition, deep neural
networks, also known as deep learning systems, have been the driving force behind much of
the recent progress in ML performance (Brynjolfsson et al., 2018, p.44). Deep learning
techniques have made it possible for machines to surpass humans in different tasks, such as
image and speech recognition, natural language processing, and predictive analytics.

Furthermore, ML techniques can be divided into several categories based on their learning
method (Min, 2010, p.16). These categories include (1) concept learning, which aims to
identify relevant concepts for future decision-making through inductive learning, (2) decision
tree learning, which involves testing object properties and composing a decision tree for
classification, (3) perceptron learning, which minimising errors and solves decision obstacles
using a single perceptron layer, (4) bayesian learning, which educate computers to learn
probabilistic functions and lastly (5) reinforcement learning, which uses feedback in the form
of rewards to educate computers to perform at high levels. Although these techniques differ in
their approach, they all seek to replicate natural learning based on human knowledge and
experience accumulated over time (Min, 2010, p.16). The various categories of ML
techniques provide a wide range of approaches for solving problems and achieving goals.
Each method has its own strengths and weaknesses, and selecting the appropriate technique
depends on the specific problem at hand. For example, decision tree learning may be more
suitable for problems with discrete variables, while bayesian learning may be better suited for
problems with probabilistic data. The process of machine learning can aid in comprehending
the driving force behind collaborative behaviour among SC partners, leading to the sharing of
critical information and enhancing the partnership through the organisational learning process
(Min, 2010, p.16).

In summary, while the scope of tasks that can be achieved through ML may be more
constrained and definable, it is important to recognize that this is just one aspect of the
broader field of AI, and that the capabilities of ML are constantly expanding and improving
and is needing more research (Brynjolfsson et al., 2018, p.47). One promising area of research
is the application of ML techniques for SCM. As SC becomes more complex and global, the
need for effective management strategies is becoming increasingly important. ML algorithms
can be used to analyse large amounts of data to identify potential risks and help managers
make informed decisions. Therefore, there is a need for further research in this area
(Pournader et al., 2021, p.11). Although there are lots of different techniques and solutions

19
through AI, the implementation process and how to introduce this to a business is a challenge.

20
3.2 The Implementation Process

3.2.1 Implementation Challenges

Large changes are anticipated as a result of AI, but according to Strohm et al. (2020, p. 5526)
the implementation of AI has quite the complex character. It is a fact that implementing AI in
the manufacturing sector can be challenging due to issues such as data availability, data
quality, cybersecurity concerns, and protection to change from human workers (Sharma et al,
2021, p.3; Strohm et al., 2020, p. 5526). For example, AI and automation can be seen as a
threat to job security, and workers may be resistant to changes in their work processes or the
introduction of new technologies (Sharma et al, 2021, p.3). Despite these challenges, it is still
important to adopt AI in this sector to remain competitive in the global market (Sharma et al,
2021, p.7-9). Further, with the significant attention AI is receiving and the significant
investment being made in it, its implementation becomes crucial (Strohm et al., 2020, p. 5530;
Schaeffer et al., 2018, p.4-5). As mentioned, some AI applications have failed to gain popular
acceptance and there is a debate whether the failed implementation is caused by poor
technical performance or organisation and/or social characteristics (Strohm et al., 2020, p.
5526; Frank et al., 2019, p. 80). If underutilization occurs, according to Boucher (2020), it
might be due to public and commercial fear in AI, weak infrastructure, a lack of initiative, and
low investment, or, because AI's ML relies on data, a fragmented digital market. As a result,
the causes underlying successful and unsuccessful implementation are hidden.
Underutilization of AI is viewed as a serious concern since it could, for instance, result in the
EU's failure to implement important initiatives like the EU Green Deal, loss of comparative
advantages over other regions, economic stagnation, and reduced possibilities for citizens
(Boucher, 2020). One must therefore ensure the catalysts behind a successful implementation
in order to secure the large investment that a large part of the society has carried out.

This shows that implementing AI in the manufacturing industry can be difficult. However, to
meet the industry and market standards, digitalization is crucial (Thun et al., 2021, p. 727).
Some other challenges that can be faced for companies during the implementation is trust in
the systems, transparency, analogue processes, and the misunderstandings people can have of
AI (Reim et al, 2020, p. 182). Trusting the system is not easy to solve and to do so companies
must ensure compatibility with existing systems, deal with network speed and stability issues,
and address data security concerns (Thun et al., 2021, p. 736). Put simply, people are less
likely to trust an AI application if they don't comprehend how it functions. Trust can be
influenced not only by the technology itself, but also by the company behind it and its ability
to convey information (Reim et al., 2020, p. 182).

Further, ensuring transparency in AI systems involves providing clarity on their operations

and decision-making processes. However, as AI comprises a dimension of technologies, it
becomes challenging to understand its decision-making mechanisms. Establishing
transparency in intelligent systems poses an important challenge (Reim et al., 2020, p. 182).
The concept of explainability is often mentioned in the literature when discussing
transparency in the context of AI. Explainability encompasses both interpretability and the
trustworthiness of the systems. For instance, recent studies on users' trust in applied AI
assume that transparency needs to be assessed in a way that considers how the average person
comprehends explanations and evaluates their relationship with a service, product, or
21
company (Larsson & Heintz, 2020, p. 7). Another challenge that is mentioned is analogue
processes. Analog processes refer to manual or physical methods of collecting and managing
data that are not digital. For example, filling out paper forms or using spreadsheets to track
information. In order to effectively implement AI, it is important to have digital processes in
place, which allow for the efficient collection and

22
storage of data (Reim et al., 2020, p. 182). The last challenge that Reim et al. (2020, p. 183)
mentions is the misunderstanding of AI. Developing a better understanding of AI and its
potential to enhance current operations is likely to lead to a more positive attitude towards the
changes it brings (Reim et al., 2020, p. 185). Understanding the benefits with the
implementation and highlighting barriers to implementing new digital tools is a challenge for
companies. The old tools can become a hindrance to implementation, and new tools need to
have the entire focus. However, due to problems with reliability, response time or login
challenges, coexistence between new and old tools can make the application more powerful
and less exposed to flaws in the digital technology (Thun et al., 2021, p. 737).

3.2.2 The Roadmap for possible Implementation Success

Even though literature regarding AI implementation within SCM is very limited, SCM and
the manufacturing industry might be able to apply established theories. Reim et al. (2020, p.
186) created a roadmap with four key insights that can be good to know when implementing
AI in a business to best succeed. The roadmap is connected to the challenges that have been
identified earlier, such as transparency issues, lack of trust in AI among employees, use of
analogue processes, and misunderstandings about AI. The main lessons learned can be
summarised as follows, (1) gaining an understanding of both AI and the organisational
capabilities required for digital transformation; (2) comprehending the current business
model, potential for business model innovation, and the role of the business ecosystem; (3)
acquiring and improving the necessary capabilities for AI implementation; and (4) achieving
organisational acceptance and building internal competencies (Reim et al., 2021, p. 186).

The first initial step in the plan is gaining an understanding of both AI and the organisational
capabilities required for digital transformation. It is a plan to produce a conceptual framework
for the management of AI and to assess the firm’s potential (Reim et al., 2020, p. 187). The
authors suggest that successful implementation of AI is dependent on data acquisition and
infrastructure. It is important to use AI in the right way, with the right language and to
understand it before the implementation process begins. The other step in the roadmap is
comprehending the current business model, potential for business model innovation, and the
role of the business ecosystem. This step suggests that before making any changes, it is
essential to understand how the business currently creates, captures, and delivers value to its
customers and how technology can be utilised to exceed their expectations (Reim et al., 2020,
p. 187). The third step is acquiring and improving the necessary capabilities for AI
implementation which address that to initiate further development of these capabilities, it is
essential to have a proper understanding of the current business model, internal and external
capabilities, and customer needs (Reim et al., 2020, p. 188). Introducing AI into a business
often involves significant changes to core operations and capabilities, leading to uncertainty
and risk. Firms have two options when undergoing these transformations which is being the
first developer or first follower. To inspire the development of both technical and strategic
solutions, firms can perform benchmarking activities and evaluate surrounding companies.
The last step in the roadmap that Reim et al. (2020) is mentioned is achieving organisational
acceptance and building internal competencies. In this context it is important to explore
collaboration with partners in order to establish a better understanding for AI applications. It
is also highlighted that feedback and evaluation during the process is extremely important

23
when implementing AI (Reim et al., 2020, p. 188).

Furthermore, to ensure the success of the application of AI, organisations should enhance
their planning and monitoring processes. Currently, some organisations take little action to
keep an eye on present practices and/or the effects of adding new technologies. Strohm et al.,
(2020, p.

24
5528) believes the unstructured character of implementation processes can be explained by
the absence of established regulations or best practices. Having an unstructured planning and
monitoring regarding AI implementation tends to restrain the implementation (Strohm et al.,
2020, p. 5528; Sun & Medaglia, 2019, p. 378). There are two main reasons why planning and
monitoring the process should be enhanced. First off, if the benefits or organisational goals
that using AI might achieve, are not clearly articulated, it becomes challenging to assess the
implementation's performance afterward (Strohm et al., 2020, p. 5528). Secondly, if
implementation plans do not outline how AI should be included into the workflow, it causes
substantial variances in how the technology is used between departments (Strohm et al., 2020,
p. 5528). Additionally, having an unplanned implementation facilitates risk appearing, which
has proven to implicate in exceeding budget and postpone schedule (Hoon Kwak & Dixon,
2008, p. 553). Hence, by planning and monitoring the implementation process more
thoroughly, companies can assess the success of the implementation further and save both
time and costs.

3.3 The outcome of AI

3.3.1 Outcomes of AI in Supply Chain Management

3.3.1.1 Potential advantages from implementing AI in the Supply chain

The implementation of AI in SCM has been advocated by several researchers even though
theories to support it are narrow (Dash et al., 2019; Bughin et al., 2017; Reim et al.,2020, p.
186). One of the primary objectives of supply chain optimisation is to ensure the efficient
operation of manufacturing and distribution activities, which involves the optimal positioning
of inventory, and the minimization of manufacturing, transportation, and distribution costs
(Sithole et al., 2016, p. 18). Sithole et al. (2016) highlighted the potential for AI to improve
SCM. The study found that ongoing cross-functional optimizations that make use of new
technologies, such as AI and Internet-of-Things (IoT), can lead to significant improvements in
supply chain optimization. This can lead to increased productivity and competitiveness for
companies. Given the common driving forces of reducing costs, accurate forecasting, and
gaining a competitive edge within the SC, there is a growing appreciation for implementing
technologies that can help achieve these objectives.

Several studies have highlighted the potential advantages of implementing AI in the SC.
Bughin et al. (2017, p. 22) found that AI-based approaches are expected to reduce forecasting
errors by 30 to 50 percent compared to conventional approaches. This can lead to significant
cost savings, as it enables more accurate inventory management and reduces waste.
Additionally, the study found that AI can reduce inventory costs by 20 to 50 percent and costs
associated with storage, transportation, and SCM by 5 to 10 percent and 25 to 40 percent,
respectively. Another potential advantage of implementing AI is real-time visibility into the
SC. According to a study by Kohli et al. (2021a), AI-based systems can provide real-time
tracking of inventory levels, monitor supplier performance, and identify potential issues
before they become problems. This enhanced visibility can help improve customer
satisfaction and reduce waste. Overall, the implementation of AI in the SC can bring
numerous advantages, including cost savings, increased productivity, improved forecasting
25
accuracy, personalised customer offerings, and enhanced visibility. These benefits make AI
an attractive option for companies looking to improve their SC operations.

Despite the potential benefits that AI can offer and the driving forces behind it, its
implementation in SCM remains uncommon due to the high complexity of SCM and the
limited exploration of AI's potential applications in this field (Min, 2010, p. 20). SCM
involves

26
managing relationships across various industries such as marketing, logistics, and production,
and requires a comprehensive understanding of interconnected decision-making processes and
the development of intelligent knowledge bases for collaborative problem-solving (Dash et
al., 2019, p. 44). The control base of SCM is characterised by networking and process
integration across functional, regional, and organisational interfaces (Van Hoek, 1998).
Hence, although many are aware of the potential benefits that AI can provide along with the
driving forces mentioned above, its implementation remains rare due to the high level of
complexity.

In conclusion, while the potential advantages of AI in SCM are numerous, its implementation
remains uncommon due to the high complexity of SCM and the limited exploration of AI's
potential applications in this field. However, as companies strive to reduce costs, improve
forecasting accuracy, and gain a competitive edge, there is a growing appreciation for the
potential of AI in the SC. With ongoing cross-functional optimizations and the use of new
technologies such as AI and IoT, significant improvements in SC optimization can be
achieved. As AI continues to evolve, it is likely to play an increasingly important role in
SCM, helping companies to achieve greater efficiency, productivity, and competitiveness.
Therefore it is essential for companies to also evaluate the disadvantages with it.

3.3.1.2 Potential disadvantages from implement AI in the Supply chain

Since the SCM is such an important part of a company, it is crucial to be able to rely on it.
Many experts argue that SCM is the most critical aspect of management for a company to
succeed (Dash et al., 2019, p. 43; Hult et al., 2007, p. 1036). Implementing AI in the SC can
also have potential disadvantages. Developing and implementing AI systems require a
significant investment of time and money, which can be a barrier for some companies,
particularly smaller ones with limited resources (Huang et al., 2018). There is also a risk of
overreliance on AI systems, which can lead to complacency and a lack of human oversight
(Kohli et al., 2021b). Additionally, AI systems may not always perform as intended, and
errors can occur. In some cases, these errors can have serious consequences, such as SC
disruptions or incorrect inventory management (Kumar et al., 2019a). Finally, implementing
AI systems may require significant changes to existing processes and systems, which can
cause disruption and resistance from employees (Huang et al., 2018). Therefore, companies
need to carefully weigh the potential benefits and drawbacks of implementing AI systems in
the SC and be prepared to address any issues that may arise.

Implementation cost is one of the disadvantages with implementing AI within the SC. AI
systems require a significant investment of time and money, and while the potential benefits
of AI are significant, the cost of it and maintaining it can be a major challenge, particularly for
small- and medium-sized enterprises (SMEs) (Schmidt et al., 2021, p. 170). Additionally,
SMEs also have more limited resources and access to financing, which can make it even
further difficult for them to invest in AI technology (Schmidt et al., 2021). The
implementation cost is especially high within SCM, since the integration of AI requires a new
level of technical complexity (Kohli et al., 2021b). The cost of implementing AI includes
purchasing AI technology, hiring experts to implement and maintain it, and training
employees to use it, and was found to be one of the top barriers for companies.
27
Another disadvantage of implementing AI in the SC is the potential for errors and
malfunctions. Although AI systems can enhance SC operations, they may not always function
as intended, leading to overreliance on these systems by humans, as noted by Kohli et al.
(2021b). This overreliance can lead to serious consequences such as incorrect inventory
management or SC disruptions. Kumar et al. (2019a) emphasise the importance of
companies being vigilant in

28
monitoring and addressing any issues that arise with their AI systems to mitigate potential
negative impacts. Here, Gharehgozli et al. (2021) suggested that companies should develop
contingency plans to address such risks and ensure that their employees have the necessary
skills and knowledge to intervene if necessary.

Lastly, AI technology is still relatively new, and many individuals may not understand how it
works or how it can benefit the SC. Additionally, there may be a lack of trust in the accuracy
and reliability of AI systems. Wang et al. (2021) found out that individuals who lack
understanding of AI may be sceptical of its capabilities and reluctant to adopt it in their SC
operations and therefore highlighted the importance of education and training to increase
understanding and trust in AI systems. Actually, lack of trust in AI systems is a significant
barrier to adoption, therefore education employees should prioritise in order to increase
implementation of AI (Hasija & Esper, 2022, p. 389; Nagurney et al., 2021, p. 400). Nagurney
et al. (2021) suggested that companies should work to build trust in AI systems by ensuring
their accuracy and reliability and communicating their benefits to stakeholders. Moreover,
implementing AI systems may lead to significant changes in current processes and systems,
resulting in both resistance and disruptions from employees (Huang et al., 2018).

In conclusion, implementing AI in the SC can bring numerous benefits, such as increased

efficiency and improved decision-making. However, it also comes with potential drawbacks,
including high implementation costs, the risk of errors and malfunctions, and the need for
employee education and trust-building. To successfully implement AI in the SC, companies
need to carefully weigh the potential benefits and drawbacks and be prepared to address any
issues that may arise. They should develop contingency plans, ensure the accuracy and
reliability of AI systems, communicate their benefits to stakeholders, and prioritise employee
education and training. By doing so, companies can maximise the potential benefits of AI
while minimising the potential drawbacks.

3.3.2 Outcomes of AI in the Manufacturing industry

3.3.2.1 Potential advantages when implement AI in the Manufacturing industry

The use of AI might improve the manufacturing sector. According to Elomri et al. (2020),
artificial intelligence may increase efficiency by streamlining production times, preventing
interruptions and identifying equipment faults prior to they arise. Furthermore, as mentioned
by Kumar et al. (2019b), AI may be utilised for predictive maintenance, assisting
manufacturers in reducing maintenance costs and avoiding unexpected downtime. Kumar et
al. (2019b), continues and argues that AI may help improve quality control by evaluating data
from sensors, cameras, and other sources to discover faults and quality concerns. Moreover,
AI can enhance safety in the manufacturing industry by monitoring working conditions and
alerting workers to potential hazards (Kumar et al., 2019b). Finally, Li et al. (2020) claim that
artificial intelligence (AI) may assist manufacturers in identifying new product opportunities
and improving existing goods by studying consumer input, market trends, and other data
sources.

Increasing efficiency is one of the key benefits of using AI in the manufacturing industry. AI-
29
controlled machines and robots can perform repetitive and labour-intensive tasks faster and
more accurately than humans, allowing workers to focus on higher-value tasks that require
human expertise (Li et al., 2020). AI can also optimise manufacturing processes by reducing
waste and minimising downtime, leading to higher productivity and efficiency. It is noticeable
that the benefit of AI within SCM will also be beneficial for the manufacturing industry, since
SCM is an essential aspect of it (Piplani et al., 2021). AI algorithms can analyse data from

30
multiple sources, such as weather forecasts, customer demand and supplier availability, to
make real-time decisions that minimise costs and lead times. For example, a study published
in the International Journal of Production Research examined the effectiveness of
implementing AI algorithms in a SC network. The researchers found that AI reduced overall
logistics costs by 9.3% and improved delivery performance by 21.7%, resulting in a
significant improvement in overall SC efficiency (Piplani et al., 2021).

Predictive maintenance is another advantage of implementing AI within the manufacturing

industry. By using machine learning algorithms to analyse sensor data from devices, AI can
predict when machines are likely to fail, enabling proactive maintenance and reducing the risk
of costly outages (Kumar et al., 2019b). Lin et al., (2021, p. 218) investigated the influence of
predictive maintenance on a semiconductor manufacturing process using machine learning
techniques. They discovered that implementing predictive maintenance reduced downtime by
30% and boosted overall equipment effectiveness by 5%, resulting in a 5% gain in output and
a 10% decrease in maintenance expenses. Additionally, Yang et al. (2019, p. 1338)
investigated the effectiveness of an artificial intelligence-based predictive maintenance system
in a plastic injection moulding machine. Results showed how the predictive maintenance
system minimises downtime by 30% and increases productivity by 6%, leading to a 12%
increase in machine utilisation Yang et al. (2019, p.1340). Lastly, predictive maintenance can
also improve inventory management, since managers can, in a more precise way, plan
maintenance activities and reduce the need for expensive spare parts (Cai et al., 2021).
Specifically, it can reduce inventory costs by 13 % and improve inventory turnover by 25%,
leading to significant cost savings (Cai et al., 2021, p. 1135). All together, AI-powered
predictive maintenance can improve maintenance efficiency, minimise downtime and
optimise inventory management. Manufacturers can perform proactive maintenance by
predicting machine failures before they occur, resulting in higher asset utilisation, production
output and cost savings.

AI can detect faults and abnormalities in real-time by analysing data from sensors, cameras,
and other sources with machine learning algorithms, allowing for early detection and remedial
action. Therefore, improved quality control is an additional area where the manufacturing
industry would benefit from implementing AI (Kumar et al., 2019b). Gao et al. (2020) found
out how the implementation of an AI-based quality control system reduced the defect rate by
60% and increased the production yield by 4%, leading to significant cost savings. Another
study investigated the effectiveness of using AI for surface defect detection in the
manufacturing of steel products (Wu et al., 2020) and found out that the AI-based system
achieved an accuracy rate of 95% in detecting surface defects, outperforming traditional
detection methods. Lastly, AI can streamline the inspection process by reducing the need for
manual inspections and increasing accuracy. One study found that the implementation of an
AI- based inspection system reduced inspection time by 60% and improved defect detection
accuracy by 20%, resulting in a significant improvement in overall product quality (Wang et
al., 2019, p. 1475). In summary, quality control based on artificial intelligence has the
potential to dramatically improve product quality, reduce defect rates and increase production
output. Manufacturers can save money and increase customer satisfaction by detecting defects
and anomalies in real time.

31
AI can enhance safety in the manufacturing industry by identifying potential safety hazards
and predicting when accidents are likely to occur through the use of predictive analytics. For
instance, AI can also be used to monitor worker behaviour and ensure that safety protocols are
being followed. Then, computer vision systems can be used to monitor workers and identify
unsafe practices, such as workers not wearing required safety equipment or standing too close

32
to dangerous machinery (Huang et al., 2020). Additionally, implementation of AI-based
safety management systems have shown a reduction in the number of safety incidents and
increased worker safety awareness in manufacturing plants (Huang et al., 2020, p. 73). In
summary, AI can be used to enhance safety in the manufacturing industry by identifying
potential safety hazards, preventing accidents, and monitoring worker behaviour to ensure
safe practices are being followed.

Lastly, innovation and product development is another area that can benefit companies after
implementing AI. By analysing data from multiple sources, including customer feedback and
product usage patterns, AI can provide insights that enable manufacturers to develop and
improve products more effectively (Li et al., 2020, p. 258). AI-implementation can reduce the
time required for product development by 30 % and increase the success rate of new product
launches by 20 % (Kwak et al., 2019). Hence, AI can provide insights that enable
manufacturers to develop and improve products more effectively, leading to more efficient
product design, faster time-to-market, and increased customer satisfaction.

In conclusion, the manufacturing industry can benefit greatly from the implementation of AI.
By improving efficiency, predictive maintenance, quality control, safety, and innovation, AI
can increase productivity, reduce costs, and improve customer satisfaction. With the
continued advancement of AI technology, it is likely that more benefits will emerge, making
AI an increasingly important tool for manufacturers looking to stay competitive in today's
fast-paced business environment. However, companies should also evaluate the potential
drawbacks of incorporating AI systems in their operation. Therefore, in order to implement AI
successfully, one must require careful consideration and planning to ensure that the benefits
of AI are balanced against its potential costs and risks.

3.3.2.2 Potential disadvantages when implement AI in the Manufacturing industry

Numerous scientific publications have emphasised the possible drawbacks of using AI to the
manufacturing industry. According to Li et al. (2020), the expenses and complexity of
creating and integrating AI systems might be a substantial barrier for many manufacturers.
There is also a possibility of employment losses as a result of some traditionally performed
functions by humans possibly becoming automated (Li et al. 2020). Furthermore, security
risks are another concern, since the use of AI in manufacturing increases the risk of
cyberattacks (Delshad & Safaei, 2021). Fourthly, Kumar et al. (2019a) examines how the lack
of transparency in AI systems might be problematic in the manufacturing industry, as
decision-making processes can have a substantial influence on production processes and
outcomes. Elomri et al. (2020) explore how the usage of AI might result in a dependency on
technology and open the door to SC risks. Therefore, before incorporating AI into their
processes, manufacturers need to thoroughly assess any potential drawbacks.

Similar to SCM, the implementation of AI in the manufacturing industry requires high

investments and is associated with complexity. However, to bring in a new perspective, the
following drawbacks are more significant for manufacturing than the previously mentioned
drawbacks. First, the use of AI in manufacturing can lead to employment losses as some tasks
traditionally performed by humans are automated (Li et al., 2020, p. 349; Kohli et al., 2021a).
33
Companies introducing AI systems need to carefully consider the impact on their employees
and develop strategies to mitigate any negative consequences (Kohli et al., 2021a). This raises
ethical concerns. For example, there are concerns about the use of AI to monitor and track
employees, which could be seen as an invasion of privacy.

34
Security risks are another drawback of using AI in manufacturing. The use of AI in
manufacturing can also pose security risks as these systems can be vulnerable to cyber-attacks
(Delshad & Safaei, 2021, p. 131; Gunasekaran et al., 2019). With AI comes the collection and
analysis of large amounts of data. Ensuring the security and privacy of this data is critical, and
companies need to take the necessary measures to protect their data (Gunasekaran et al.,
2019). The large amounts of data required for AI systems to function effectively must be
collected, stored and processed in a secure manner. If this data is compromised, it can have
serious consequences, such as data breaches or loss of confidential information. Therefore,
Delshad and Safaei, (2021) emphasise that manufacturers need to ensure that their AI systems
are secure and protected from potential cyber threats.

Furthermore, the lack of transparency is a significant drawback to the adoption of AI in the

manufacturing industry. While AI has the potential to improve production efficiency, quality
control and predictive maintenance, its lack of transparency can introduce risks such as errors,
bias and safety concerns. Lack of transparency in the development and training of AI
algorithms can lead to biassed decisions (Kang et al., 2020). This can be particularly
problematic in the manufacturing industry, where decisions made by AI systems can have a
significant impact on worker safety, production efficiency and product quality. Furthermore, a
lack of transparency in the decision-making process of AI systems can make it difficult for
manufacturers to understand how the system arrived at a particular decision (Guo et al., 2020,
p. 152). This can be a problem when manufacturers need to fix problems with the AI system
or make changes to improve its performance. In addition, the lack of transparency in the data
used to train AI systems can lead to inaccurate predictions and decisions. According to Song
et al. (2020, p. 3447), this can be particularly problematic in the manufacturing industry,
where AI systems are often used to predict equipment failures or detect quality defects in
products. Researchers suggest that addressing the lack of transparency requires a
multidisciplinary approach that includes incorporating transparency into the design and
development of AI systems, providing transparency to end users, and setting regulations and
standards for AI (Guo et al., 2020; Song et al., 2020; Kang et al., 2020). Failure to address
transparency issues can lead to mistrust, legal and ethical problems, and unintended
consequences.

Finally, technology dependency is a potential drawback to implementing AI in the

manufacturing industry (Elomri et al., 2020). According to Jäger et al. (2021), heavy use of
AI can lead to a decline in creativity and innovation among workers who become overly
dependent on technology to perform tasks. This can lead to a lack of employee engagement
and lower job satisfaction, which in turn can lead to higher turnover rates. In addition, this
reliance on technology can lead to significant production disruptions when malfunctions,
system failures or hacking occur, and raise ethical concerns. According to Brown and Wilson
(2019), increasing reliance on AI and automation may reduce the capabilities of the human
workforce, leading to a situation where workforce diversity is limited and workers cannot
respond to unforeseen events. This could also lead to job losses and potential economic
inequality. In addition, the inability of the workforce to respond appropriately to new
situations can have a negative impact on innovation and growth. Therefore, manufacturers
need to carefully manage their integration of AI to mitigate these risks and ensure that they

35
are not overly dependent on the technology.

In conclusion, while AI has the potential to revolutionise the manufacturing industry, there are
several drawbacks that need to be carefully considered before integrating AI systems into
manufacturing processes. These include the potential for job losses, security risks, lack of
transparency, and technology dependency. To overcome these challenges, manufacturers need
to develop strategies to mitigate the negative impact of AI on their employees, ensure the

36
security and privacy of their data, increase transparency in the development and training of AI
algorithms, and manage their integration of AI to avoid over-reliance on technology. By
doing so, manufacturers can unlock the benefits of AI while minimising its potential
drawbacks and creating a sustainable future for their business and workforce.

37
4.0 Practical methodology

Our practical methodology outlines the methods used when retrieving our data for our thesis.
Specifically, it includes our data collection, data analysis and ethical consideration. Initially
our data collection comprises our sampling techniques, description of how the interviews
were conducted and the transcription of data. Thereafter, our data analysis is described step by
step to provide transparency. Lastly, the ethical consideration we took into account when
collecting our data is displayed.

4.1 Data collection methods

In order for us to ensure that the research question is answered, and our purpose is being
fulfilled, we aimed to recognize the data collection method and choose a method that is
consistent with our philosophical assumptions (Howard-Grenville et al., 2021, p. 1315). In our
study, our research question is focused on how the implementation of AI can improve SCM
within the manufacturing industry, and we believe that existing data is insufficient to answer
this question. Therefore, we have gathered primary data to increase our chances of answering
the research question.

For this study, we used semi-structured interviews to collect primary data. This approach is
aligned with our interpretivist research design, which aims to gain an in-depth understanding
of the subjective experiences and perceptions of respondents regarding how implementing and
managing AI within the manufacturing industry can improve SCM. Semi-structured
interviews were the most suitable and is a commonly used method for obtaining qualitative
data in an interpretivist study. It offered flexibility in terms of question order, while
maintaining a natural flow and direction to the conversation (Kallio et al., 2016, p. 2960).
Open-ended questions were used in our semi-structured interviews to encourage the
respondents to provide detailed and personal answers beyond a simple 'yes' or 'no' response,
providing us with extensive insights based on respondents' experiences (Bearman, 2019, p. 4).
Probing questions, such as 'why', 'how', and 'can you provide examples', has also been used to
further develop respondents' answers. Additionally, the semi-structured format eliminated the
need for follow-up interviews, enabling us to continuously confirm or clarify responses during
the interview without deviating from the interview guide (Kallio et al., 2016, p. 2960-2961).

Structured and unstructured interviews are common options for data collection, but neither
were suitable for our study. Structured interviews ask the same questions in the same order,
leaving little room for deviation from the planned interview guide and possibly missing
important information (Blouin et al, 2011, p. 517). This type of interview is also associated
with the positivist paradigm, which does not align with our interpretivist approach. On the
other hand, unstructured interviews do not have pre-planned questions, making it difficult to
take notes and capture all the information provided by the respondent. This approach may
also fail to elicit crucial information from the respondent unless the interviewer asks the right
questions (Blouin et al., 2011, p. 517).

4.1.1 Sampling technique

As our research focuses on a specific topic, it was not feasible for us to study the entire
38
population. Therefore, we opted to use a sample of participants with relevant experience and
knowledge, which is a common practice in analytical qualitative studies (Gill, 2020, p. 579).
To ensure that our sample was relevant to our research question, we employed purposeful
sampling as our primary sampling method. This method involves using our judgement to
select

39
participants who we believe can provide insightful answers to our research question.
Purposeful sampling allowed us to actively include participants who met our selection criteria
(Gill, 2020,
p. 580). These criteria were based on three factors that we deemed essential to uphold.

To begin with, we limited our sample to companies operating in Sweden, as the

manufacturing industry is a significant contributor to the Swedish GDP. This allowed us to
collect comparable data from companies operating in the same market. Secondly, we only
included participants who were knowledgeable about AI, meaning that they had been part of
organisational discussions or were proficient regarding the implementation of this technology.
Furthermore, since we aimed to investigate activities concerning how the implementation of
AI can benefit SCM, it was essential that our respondents were actively involved in SCM.
This approach enabled us to gather information from an environment that is highly engaged
with AI. All the interviewees were either SC managers or members of manufacturing
companies, providing us with a holistic view of how they operate and enabling us to develop
the theory presented in subsequent chapters.

We obtained information about organisations that met our criteria by initiating research
through searching websites. We contacted suitable companies and individuals via email and
LinkedIn, and provided them with an information form (see appendix one) outlining our
research project and the criteria they needed to fulfil. Through this contact, we confirmed
their eligibility and ability to participate in an interview. If we did not initially contact the
appropriate manager, we requested that the manager forward our request to individuals they
believed had knowledge and experience regarding risk identification in pharmaceutical
projects. We obtained a total of two interviews, thanks to forwarding. Individuals who agreed
to participate in our study, contacted us to arrange a time and place for the interview, and
signed a consent form (see appendix three) agreeing to the terms of participation.

We chose purposeful sampling, which is a type of non-random sampling, for several reasons.
First, since we did not aim to generalise our findings to the whole population, it was more
appropriate to directly contact participants who could provide relevant data (Benoot et al.,
2016,
p. 2). Random sampling is typically used in positivist studies, where the goal is to generalise
from the population (Taherdoost, 2016, p. 20). Second, purposeful sampling allowed us to
identify participants who met our criteria. Lastly, non-random sampling techniques are
particularly relevant to the interpretive paradigm, which focuses on participants' experiences
and knowledge in depth of the phenomenon under study (Benoot et al., 2016, p. 3). Therefore,
this sampling technique was best suited to our research purpose of providing insights on how
the implementation of AI can benefit SCM in the manufacturing industry.

Snowball sampling has been used in conjunction with purposeful sampling to ensure inclusion
of individuals with relevant experience in the study (Taherdoost, 2016, p. 20). This involves
asking participants if they know anyone with similar knowledge or experience who could be
included in the study. Given the extensive networks that managers or employees within SC
typically have, snowball sampling was deemed appropriate for our study (Larson & Gray,
40
2021, p. 11).

Determining the appropriate sample size for a qualitative research study involves various
factors. Saturation, where new interviews yield minimal or no new information, is one such
factor that may guide sample size (Aldiabat & Le Navenec, 2018, p. 247). However, Gill
(2020,
p. 581) suggests that it cannot be predetermined, and therefore, specifying a sample size may
not be necessary. He mentions that to justify the findings in qualitative studies, the authors
need

41
to demonstrate to the reader that the sample size and selection were appropriate and adequate
for addressing the research questions (Gill, 2020, p. 581). Conversely, not achieving
saturation may indicate further research opportunities. Given our time constraints for the
degree project, achieving saturation was our ideal aim, but we were aware that it may not
have been feasible. Instead, we focused on obtaining a comprehensive understanding of the
topic, which was realistic given our limitations. Furthermore, Hagaman and Wutich’s (2016,
p. 205) study suggests that conducting six to sixteen interviews is a reasonable approach. Our
study included a sample of six interviews, which according to Magnusson and Marecek
(2015, pp. 36-37) should be fine, since in an interpretivist study, the amount of interviews is
less relevant than the quality of them.

4.1.2 Conducting the interviews

In a qualitative study like ours, face-to-face interviews are the most suitable method as they
allow participants to see each other and build trust, leading to higher participation rates
(Krouwel et al., 2019, p. 7). However, since our participants were located throughout Sweden,
we conducted videoconference interviews via platforms such as Zoom to avoid time-
consuming travel (Basch et al., 2020, p. 926). Remote interviews have their benefits,
including increased participation from geographically distant areas (Solarino & Aguinis,
2021, p. 663). Videoconference interviews are preferable over phone calls, as they allow for
non-verbal expressions to be observed and analysed (Basch et al., 2020, p. 922). However,
phone calls may reduce biases stemming from the lack of visual contact, which can be a relief
for participants. In light of these factors, we determined that videoconference interviews were
the most suitable option for our thesis.

Effective planning and time management were crucial for our thesis due to the limited
timeframe we had. To respect the participants' time and energy, we scheduled predetermined
45-minute interviews. Unstructured interviews can be very time-consuming, whereas
structured interviews with closed questions are the most time-effective option, with semi-
structured interviews falling somewhere in the middle (Mueller & Segal, 2015, p. 1). It is
common to underestimate the time needed for an interview. We found 45 minutes to be a
suitable time frame for the interviews, as well as for the transcription and analysis of the data,
within our timeframe for the thesis. While most interviews were within the predetermined
time frame, some were shorter or longer, for more detailed duration of each interview, see
table three. Semi- structured interview has the tendency to exceed time frames, however some
interviews of ours where quite short. The reason behind this was due to the participants
differences in knowledge and elaboration skills. For instance, our shortest interview was
conducted within 28 minutes, this participant had long elaborations which resulted in several
questions being answer at once. Meanwhile we did not want to interrupt our participants.
Overall, the data collected through our interviews aimed to seek saturation which we believe
it has.

We followed a semi-structured approach in our interviews by using an interview guide that

contained prepared questions and themes (Mueller & Segal, 2015, p. 2). All interviews
followed the same structure, and we shared the questions with the participants beforehand to
allow them to prepare for more thoughtful and developed answers (Collis & Hussey, 2014, p.

42
133). Our open-ended questions encouraged participants to provide detailed answers (Gill et
al., 2008, p. 292). However, it was important for the interviewer to actively engage and
encourage the participants during the interview to become more integrated with them and
encourage elaboration (Kallio et al., 2016, p. 2955). Asking follow-up questions was crucial
to gain insights, but it was essential to do so gently to avoid interrupting the participants
(Kallio et al.,

43
2016, p. 2960). We aimed to strike a balance between the questions and themes while also
receiving advanced questions that contributed to insightful data analysis.

To ensure ethical research practices, we began each interview by informing the participants of
the confidentiality of their information during and after the study. We also received explicit
consent from them to record the interview for transcription purposes. Moreover, all
participants had previously signed a consent form that detailed the research ethics and their
right to withdraw or decline to answer any questions, see appendix three. The interview guide
provided structure to the conversation and helped us stay on track while adhering to the
allocated time frame. Following Solarino and Aguinis' (2021) recommendation, at the end of
the interview, we invited participants to share any additional comments or information they
deemed relevant, as this can provide valuable insights to the research.

4.1.3 Interview guide

The semi-structured interviews that were conducted were guided by our interview guide. The
guide is composed of five main themes, all relevant to our research question (refer to table
two). Once the themes were established, we proceeded to formulate questions for each theme,
taking into account the research question, theoretical framework, and purpose. As researchers,
we strived to strike a balance in formulating questions that are brief and uncomplicated, non-
leading, but still capable of eliciting sufficient data to answer the research question (Kallio et
al., 2016, p. 2960).

Table 2: Themes for our interview guide

Introduction
Background
Artificial Intelligence
SCM Efficiency
Concluding remarks - Ending thoughts

The interview guide is provided in two versions, one in English and one in Swedish, as shown
in appendices five and six. This is to ensure that the interviews are conducted in a language
that both parties are proficient in, reducing the risk of translation issues or other language-
related barriers (Squires, 2009, p. 277). If any participant preferred to conduct the interview in
English, we were flexible and willing to accommodate their preference. This is because we
aimed to use a language that was understandable and relevant to the participants (Squires,
2009, p. 285). However, the interviews were conducted in Swedish since both the researchers
and all participants are native Swedish speakers, allowing for the collection of more insightful
and in- depth data.

Each interview had a standardised structure, starting with an introduction that explained the
study's purpose, introduced us, and informed the participants of their rights. The researchers
aimed to establish a positive atmosphere and build trust by creating a comfortable setting, as
44
the beginning of the interview is critical (Gill et al., 2008, p. 292). The interview followed a
specific sequence of themes and began with background questions to provide context. All
questions were asked neutrally to avoid any biases or confusion. This approach enabled

45
participants to express their perspectives on the phenomenon being studied and describe
environments or processes as they experienced them, which is crucial for conducting
successful semi-structured interviews (Gill et al., 2008, p. 292-293). Short follow-up
questions were used to clarify answers and gather more in-depth data. Towards the end of the
interview, probing and specific questions were asked to encourage exploration in particular
areas (Gill et al., 2008,
p. 293). The interviews concluded with a summarising question, followed by an opportunity
for participants to provide additional data or express their opinions on implementation of AI
within SCM in the manufacturing industry.

4.1.4 Recording and transcription

In order to improve the quality of the interviews, we made the decision to record the audio of
each interview. This allowed us to focus on the conversation and follow-up on any unclear
points, without the distraction of taking notes (Harvey, 2011, p. 436). Of course, we made
sure to inform participants of our intentions to record, and we obtained their consent
beforehand, in line with our research ethics. However, if a participant was uncomfortable with
being recorded, we respected their wishes and was ready to take detailed notes instead.
Luckily, all our participants allowed recording. To ensure we had a backup recording, we
used two devices (our mobile phones and Zoom/Teams) throughout all the interviews,
mitigating the risk of technical difficulties leading to lost audio recordings.

We took measures to ensure the respondents were comfortable with the recording process,
such as reminding them of their anonymity and company name confidentiality and
emphasising it throughout our communication with them. However, some respondents may
still have hesitations despite these efforts (Harvey, 2011, p. 436). To transcribe the audio
recordings, we first used a transcription tool, Microsoft Word's own transcription tool, to save
time and then reviewed and made any necessary amendments to ensure accuracy. The purpose
of transcription was to use grounded theory to analyse the data we collected, which is
essential for answering our research question.

4.2 Data analysis

We have selected grounded theory as our data analysis method because our project aims to
generate theories based on observations in reality rather than testing pre-existing theories.
Grounded theory involves breaking down collected data into its key components through a
series of coding cycles, resulting in a theory that is highly rooted in the original data (Saldaña,
2013, p. 51).

The two alternative approaches for content analysis are inductive and deductive, as described
by Elo and Kyngäs (2008, p. 109). In our study, we have chosen the inductive approach
because we aim to generate new theoretical insights and gain a deeper understanding of how
the manufacturing industry implements AI within SCM. Therefore, we will be going from the
specific to the general in our content analysis.

The coding process is rarely accurate on the first try, as noted by Saldaña (2013, p. 10).

46
Therefore, we have implemented the coding process multiple times and conducted it
separately between researchers to gain diverse perspectives on the collected data. This, in
combination with a repetitive coding process, aims to increase the reliability of the data.
Throughout the coding and recoding process, we have aimed to refine our codes and
categories to become more conceptual and abstract, thereby creating a more robust foundation
for theory building (Saldaña, 2013, p. 11). Additionally, we compared our coding and
discussed any discrepancies to

47
comprehend each other's interpretations of the data. Furthermore, the data analysis format is
based on a four-step process, which is presented in the following section.

Step 1. Developing initial codes

To develop concepts and formulate a theory, we begin by analysing the transcribed interviews
and creating the initial codes. These codes represent our first impressions of the data, as
described by Saldaña (2013, p. 5). We use questioning as a valuable tool to search for answers
within the interviews, as recommended by Grodal et al. (2020, p. 593), specifically we broke
down our research question that aims to answer ‘how’ something is conducted into more
detailed questions when identifying initial codes with ‘what’, ‘when’ and ‘where’ the
implementation of AI can improve the SCM. The objective here was to gain a general
understanding of the data by reading through it. This process yielded in 314 initial codes from
the six interviews, which are listed in table three below.

Table 3: Overview of initial codes

Case Role of Duration of Pages Number of

respondent interview transcribed initial codes
Intervie Digitalization 37 min 9 pages 73
w1 manager codes
Intervie Supply 42 min 10 81
w2 chain pages codes
planner
Intervie Logistic 34 min 6 Pages 28
w3 manager codes
Intervie Logistic 28 min 8 Pages 42
w4 manager codes
Intervie Digitalization 45 min 8 Pages 58
w5 manager codes
Intervie Production 30 min 7 Pages 32
w6 planning codes
manager
Total 48 314
pages codes

Step 2. Developing first-order codes

Secondly, we gathered the initial codes from the previous session. The process here included
an analysis of all interviews, aiming to establish which initial codes were of similar
characteristics and of interest to our topic, AI implementation to improve SCM. Generating
categories that could serve as the basis for new theoretical insights was the essence of our
analysis process (Grodal et al., 2020, p. 594). By coding, we could categorise our gathered
data on shared beliefs and opinion (Saldaña, 2013, p. 9). To create categories, we grouped
similar components, elements, or quotes into a first-order code. To preserve the authenticity
48
of the data collected from the interviews and avoid misinterpretation, we always used the data
within its proper context and quoted it verbatim from the respondents' words. Our first-order
code was the initial step for us to depart from a specific and detailed perspective and
approach a more

49
general finding (Grodal et al., 2020, p. 594). In total, we identified 15 first-order codes, which
are presented in table four below.

Table 4: Overview of first-order codes

First-order codes

1.a.1 Creating scenario analyses to facilitate business decisions

1.a.2 Generating feedback and evaluation to move forward

1.b.1 The current situation
1.b.2 Improving administrative work to gain time
1.b.3 Improving companies’ competitiveness

1.b.4 Preventing disruption to maintain production

2.a.1 Decreasing contingencies to improve data quality
2.a.2 Creating an ‘even flow’
2.b.1 Keeping inventory to an appropriate level
2.b.2 Saving the most out of resources

3.a.1 Planning and restructuring

3.a.2 Creating similarity in work to diminish deviation
3.b.1 Growing the accuracy and live up to expectations
3.b.2 Increasing the transparency in order to reduce suspicion
3.b.3 Simplifying the road-map in order to see the end-goal

Step 3. Developing second-order codes

Thereafter, we compared the first-order codes with our theoretical framework and the existing
literature that we had previously noticed. We found that comparing emerging concepts,
theories, or hypotheses with existing literature was a critical aspect for us and the theory-
building we aim to contribute with (Eisenhardt, 1989, p. 544). We leveraged literature from
the theoretical framework to interpret the collected data. Therefore, the second-order codes
are the results of combining the first-order codes and representative literature, which can be
found in table five below.

Table 5: Overview of the development of second-order codes

First-order codes Second-order codes

1.a.1 Creating scenario analyses 1A. Increasing accuracy of forecasts to
to facilitate business decisions enable success
50
 1.a.2 Generating feedback and evaluation
to move forward
1.b.1 The current situation 1B: Refining methods and processes in
order to sustain

1.b.2 Improving administrative work

to gain time
1.b.3 Improving companies’
competitiveness
1.b.4 Preventing disruption to
maintain production
2.a.1 Decreasing contingencies to 2A: Enhancing the allocation of
improve data quality resources
2.a.2 Creating an ‘even flow’
2.b.1 Keeping inventory to an appropriate 2B: Optimising the supply chain
level
2.b.2 Saving the most out of
resources
3.a.1 Planning and restructuring 3A: Reducing complicated
transformation
3.a.2 Creating similarity in work
to diminish deviation
3.b.1 Growing the accuracy and live up 3B. Working on increasing trust to
to expectations the AI-system
3.b.2 Increasing the transparency in
order to reduce suspicion
3.b.3 Simplifying the road-map in
order to see the end-goal

Step 4. Developing aggregate dimensions

Lastly, we reviewed our second-order codes and analysed how they concurred with each
other. Especially during this stage, we took into account the research question we have, to
ensure that the emerging concepts and theories would address the research question. In
addition, alongside with the process of scrutinising how the second-order codes co-occur, we
kept relevant literature in our mind, highlighting the significance of comparing emerging
theories with existing literature (Eisenhardt, 1989, p. 544). We then organised and structured
the concepts and theories, into a more condensed and abstracted form, creating three
aggregate dimensions, which can be viewed in table six.

51
Table 6: Overview of the development of aggregated dimensions

Second-order codes Aggregated dimensions

1A. Increasing accuracy of forecasts 1. Automating the production for more
to enable success prosperity
1B. Refining and automating in order
to sustain
2A. Allocating resources to 2. Conserving financial funds
improve contingency management
2B: Optimising the supply chain
3A. Reducing complicated 3. Integrating previous approaches
transformation to mitigate emerging challenges

3B. Working on increasing trust to

the AI-system

4.3 Research ethics

The adherence to ethical principles ensures equitable treatment of all research participants in
accordance with widely recognized standards upheld by the research community (Bryman &
Bell, 2007; Saunders et al, 2019; Collis & Hussey, 2014).

Throughout our research, we have made a conscious effort to uphold academic ethics both in
conducting our research and in handling the data collected from respondents. To demonstrate
our commitment to ethical research practices, we have followed Bryman and Bell (2007, p.
71) recommendations, which outline how we have treated our respondents and their
participation, as shown in the table seven below.

Table 7: Ethical research principles

Ethical principles (Bryman & The researchers’ ethical approach

Bell, 2007, p. 71) to research
during this study:
Harm to participants We maintained an open-minded attitude
It is essential to safeguard the mental and during the interviews to create a
physical health of both the researchers welcoming atmosphere for the
and participants to avoid causing any participants. Additionally, we allowed
harm to any of the individuals involved. them to choose between conducting the
interview in the comfort of their own
home or at their workplace, a familiar and
comfortable environment for them.
52
 Dignity During the interviews, we made sure to
Respect for the dignity of both the demonstrate respect to our respondents by
respondents and researchers is necessary acknowledging and affirming their
to prevent any participant from statements. This reassured them that we
experiencing discomfort or anxiety. were actively listening and that their
insights were highly valued.

53
 Privacy We ensured that the questions asked were
The confidentiality of the strictly related to the respondents'
respondents must be protected by professional roles and did not include any
the researchers. personal questions that may have made
them uncomfortable.
Informed consent The respondents were given an
The researchers have the responsibility to information form and a consent form to
ensure that the respondents fully ensure they were fully informed about the
comprehend the ethical research interview process and to clarify the
principles that are being followed during expectations of both the researchers and
the study before obtaining their consent. the participants.
Anonymity The respondents were repeatedly
The researchers assured the respondents informed about the guarantee of their
that their participation in the study would anonymity throughout the study, using
not result in their identity being disclosed various methods such as the consent
or made identifiable to anyone. form, information form, emails, and a
statement made before recording the
interview.
Confidentiality To ensure the confidentiality of research
The confidentiality of research data for data for the respondents and the
the participants and the organisations they organisations they represent, we provided
represent was ensured by the researchers. them with both an information form and a
consent form that they had to sign. These
forms reassured them that their data
would be kept confidential as promised.
Deception, Honesty and We made every effort to encourage our
transparency respondents to provide truthful
The research process can be disrupted by information during the interviews.
dishonest or misleading responses from However, we acknowledge that there is a
the respondents, as well as possibility of deception or
misrepresentations of data by the misinformation, which can undermine the
researchers. Therefore, it is essential for research process. We remain hopeful that
all participants involved in the research to our respondents have shared their genuine
demonstrate honesty, trustworthiness, and perceptions and experiences to the best of
openness. their knowledge and memory.
Affiliation We disclosed our professional affiliation
The researchers have an ethical and provided our personal information
responsibility to disclose any sources of and our supervisor's contact details to the
funding, sponsorship or professional respondents. Although our research was
affiliations related to the research. not sponsored, we made sure to be
transparent about our university
affiliation.
Reciprocity We ensured mutual benefits for both
54
It is important to ensure mutual benefits researchers and respondents throughout
for both researchers and respondents the research process by having active
during the research process, and active participation and collaboration from all
participation and collaboration from all participants. Our respondents were
parties involved is necessary. willing to answer our questions, and if
they didn't know the answer or felt they
didn't have a good answer, they were
honest about it. Additionally, we offered
to send the thesis to our respondents once
it was finished, if they wished to receive
it.
Misrepresentation It is important to report the findings in a
To ensure accurate reporting of research clear and concise manner, using
findings, researchers must avoid any appropriate language and avoiding any
presentation of data that is misleading or manipulation of data to support
misrepresentative, and that may lead to preconceived notions or biases. The goal
misunderstandings is to present the findings in a way that is
honest, transparent, and informative,
while avoiding any potential harm or
confusion to the readers. To ensure that
the research findings were reported
accurately, we transcribed the audio
from the

55
 interviews and conducted separate coding
of the data multiple times. This allowed
us to determine the most appropriate
coding and minimise the risk of
presenting data outside of its relevant
context. Both authors reviewed and
compared their coding to ensure accuracy
and appropriate representation of the data.

4.4 Quality criteria

Maintaining high standards of quality and veracity is imperative in the realm of research. To
uphold these standards, we will define two criteria that we have incorporated into our degree
project. These two criteria are trustworthiness and authenticity and are normally used in
research that is qualitative.

4.4.1 Trustworthiness
The trustworthiness criterion is subdivided into Credibility, Transferability, Dependability,
and Confirmability (Amin, 2020, p. 1473; Cope, 2014, p. 89).

Credibility
Credibility pertains to the veracity of the data or the viewpoints of the participants, as well as
how the researcher has interpreted and presented them (Cope, 2014, p. 89). We have exerted
considerable efforts to attain transparency and establish credibility. To guarantee that the
phenomenon truly reflected reality, we employed triangulation by incorporating multiple
sources of information (Amin et al., 2020, p. 1475). Multiple researchers employing the same
methodology should ideally yield similar outcomes. Nonetheless, in reality, this is not always
the case. If any disparities arose regarding the interpretation of the data, we deliberated the
discrepancies and subsequently arrived at a mutual understanding for a novel interpretation
that was more appropriate. Furthermore, we have ensured that the respondents were
adequately informed prior to the interview by sending out an information letter. During the
interviews, we made sure to establish a comfortable and secure atmosphere, which helped the
respondents feel at ease and facilitated a more natural and productive conversation. Our
efforts led to the collection of rigorous data. Additionally, we guaranteed anonymity, which
allowed the participants to delve deeper into their experiences. This aligns with the idea of
prolonged engagement (Amin et al., 2020, 1473). Lastly, we followed a practical
methodology that involved a thorough and systematic coding process. Through persistent
observation, we combined theories with the data to develop a model that answered our
research question. To create authentic codes, we repeatedly reviewed the data and processed it
in a systematic manner. By following these rigorous procedures, we aimed to enhance the
credibility of our study and ensure the accuracy of our findings.

Transferability
The transferability criterion in a study is about how general conclusions can be drawn
56
the results in other social environments and situations (Korstjens & Moser, 2018, p. 121).
Incorporating transferability is crucial, as it is ultimately up to the reader to determine whether
the findings are applicable beyond the specific context of the study (Cope, 2014, p. 89). We
acknowledge that our sample size was relatively small, consisting of only six participants.
However, it was not our intention to make generalised claims. Nonetheless, we have
confidence that data saturation has been achieved and therefore our findings can be applicable
to comparable settings. We believe that our result can be applicable to other
manufacturing

57
companies around the world as well as other industries in Sweden. Schaeffer et al. (2016, p.
6) mention that industries that are behind in the implementation of AI are for example the
agriculture industry, metals & mining industry and the food & beverages industry. These
industries can benefit from this result and incorporating AI into their SCM could potentially
lead to similar improvements in efficiency. It is important to note that the transferability
criterion in a study is about how general conclusions can be drawn from the results in other
social environments and situations (Korstjens & Moser, 2018, p. 121).

Dependability
The dependability criterion refers to the extent to which the research findings are consistent
and stable over time and across conditions (Cope, 2014, p. 89). Dependability is attained
when a different researcher conducts a study with comparable features and arrives at the same
conclusions, or at least concurs with the conclusions drawn by the original researchers (Cope,
2014, p. 89). In our thesis, we aim to ensure that the research process is defined and presented
clearly, enabling the reader to easily follow the decisions made by us. Moreover, to ensure
dependability, we have regularly presented our work in progress to a supervisor and peers,
receiving feedback and making changes accordingly. Based on these measures, we assert that
our study fulfils the requirements for the dependability criterion.

Confirmability
The last criterion of trustworthiness is confirmability. The concept of confirmability pertains
to the researcher's capacity to demonstrate that the collected data accurately reflect the
responses of the participants and are not influenced by the researcher's personal biases or
perspectives (Cope, 2014, p. 89). In order to attain a large degree of objectivity and
confirmability, we have opted to provide a comprehensive display of our coding process,
culminating in the aggregate dimension. However, in business research, it can be challenging
to completely detach oneself from personal values (Cope, 2014, p. 89). To mitigate this
potential interference, we ensured that both researchers were involved in the data collection
and analysis stages, allowing for a dialogue when interpreting the responses of the
participants. In addition, we have provided quotes from respondents that portray every new
and developing theme to ensure that we fulfil this criterion as well (Cope, 2014, p. 89).

4.4.2 Authenticity
Authenticity is divided into five components, Fairness, Ontological authenticity, Educative
authenticity, Catalytic authenticity, and Tactical authenticity (Amin, 2020, p. 1479-1480;
Phillips et al., 2014, p. 8).

Fairness
Among all the authenticity criteria, fairness holds the most significance. Fairness pertains to
whether the research has accounted for the diverse experiential realities of the participants
when relaying the responses obtained from the conducted interviews (Amin et al., 2020, p.
1479; Phillips et al., 2014, p. 10). It means that the researcher should avoid any form of
favouritism or discrimination towards any individual, group, or community involved in the
research. By interviewing individuals with a range of roles within each company, we were
able to obtain a more comprehensive understanding of the perspectives and experiences of the
58
participants, thus avoiding any potential bias or discrimination towards any particular group
or individual. For example, some of the respondents were logistic managers and others were
supply chain planners or digitalization managers. This gave us the opportunity to approach the
process from several lived realities of our respondents.

59
Ontological authenticity
The concept of ontological authenticity criterion is particularly important in studies that
involve people as participants. This criterion ensures that participants gain a better
understanding of their social context and receive something valuable in return for their
participation in the study (Amin et al., 2020, p. 1480; Phillips et al., 2014, p. 10). By
examining the implementation of AI in the manufacturing industry and its impact on SCM,
this study not only contributes to the knowledge of practitioners and researchers but also
meets the ontological authenticity criterion by providing participants with valuable insights
into the social situation in which they live and work. Therefore, this study offers a win-win
situation for both participants and the wider community by generating knowledge that
benefits all stakeholders involved.

Educative authenticity
The third criterion, educational authenticity, pertains to the extent to which study respondents
develop a deeper awareness and comprehension of individuals foreign of their own
stakeholder group (Amin et al., 2020, p.1480; Phillips et al., 2014, p. 10). Upon the
completion of the study, the final version was made available to the participants. By granting
access to the study's respondents, they can learn about the various strategies used by other
companies with the implementation of AI including their views on how to improve SCM.
Additionally, the study's conclusions are presented, providing participants with valuable
insights into the topic.

Catalytic authenticity
Catalytic authenticity refers to how much a research study inspires and enables action towards
clarifying the issue at hand, improving or solving the problem, or strengthening values related
to the issue. The idea behind catalytic authenticity is that simply gaining knowledge through
research is not enough to effectively address the various issues that participants may raise
during the research process. Instead, the study should encourage and facilitate action towards
addressing those issues in a meaningful way (Amin et al., 2020, p. 1480; Phillips et al., 2014,
p. 10). The aim of our study was not only for us as writers to enhance our understanding of
the subject matter, but also for study participants and readers to gain a deeper and broader
understanding and knowledge of the topic. We hope that those who take part in the study will
find useful insights that can inform the development of their implementation of AI or other
relevant areas.

Tactical authenticity
Tactical authenticity refers to the degree to which all participants in a research inquiry are
given the power and agency to take the necessary actions that are implied or proposed by the
research (Amin et al., 2020, p. 1480; Phillips et al., 2014, p. 10). To achieve tactical
authenticity, several useful procedures and techniques can be employed, including
confidentiality, negotiations regarding the types of data that will be collected and how they
will be interpreted and reported and the use of clear and detailed consent forms (Amin et al.,
2020, p. 1480). We have fulfilled this criterion by sending out clear consent forms to all our
respondents which can be seen in appendix three and have been careful with confidentiality
throughout the whole process. In addition, our contention is that this standard has been met, as
60
our model provides respondents and stakeholders with valuable insights into the process of
implementing AI in the manufacturing industry in order to improve SCM. Specifically, it
offers participants the chance to engage with the framework of our thesis when approaching
the process.

4.5 Summary over practical methodology

Table 8: Summary of practical methodology

61
Data Collection methods Primary data through interviews
Sampling technique Purposeful and snowballing
Conducting the interviews Semi-structured interviews
Through video conference calls.
Research ethics Table 7. Ethical research
principles
Analysis method Inductive
Grounded Theory
Quality criteria Trustworthiness and Authenticity

62
5.0 Empirical findings and analysis

This chapter will present the results of our data collection efforts. Firstly, we will provide a
summary of the findings. Then, we will delve into a detailed analysis of the aggregated
dimensions, in order to break down both the first-order and second-order codes that constitute
them.

5.1 Findings and data structure

The objective of our thesis project is to develop a comprehensive understanding of How

implementation of AI can improve SCM. We adopted an inductive approach to perform
thematic analysis of the collected data, specifically grounded theory, as mentioned earlier.
Furthermore, we structured our analysis into three levels, namely aggregate dimensions,
second-order codes, and first-order codes, which will be elaborated upon in the following
sections. Figure one illustrates the data structure of the analysis and shows how the three
aggregated dimensions were developed.

The first aggregate dimension, (1) Automating the production for more prosperity, is
composed of two elements, (1A) Increasing accuracy of forecasts to enable success and (1B)
Refining methods and processes in order to sustain.

Moving on, the second aggregate dimension, (2) Conserving financial funds, comprises two
elements, (2A) Enhancing the allocation of resources and (2B) Optimising the supply chain.

The third and the last aggregate dimension, (3) Integrating previous approaches to mitigate
emerging challenges, is built on two components, (3A) Reducing complicated transformation
and (3B) Working on increasing trust to the AI-system.

63
Figure 1: Overview of the data structure

64
5.2 Aggregate dimension 1: Automating the production for more prosperity

For companies to improve their SCM, our data analysis means they need to automate
production in order to create more accurate forecasts. Forecasting is being used by
employees and managers as a prediction-tool to estimate possible supply and demand on the
market. By predicting the demand, companies arrange their organisation and thereafter will be
able to know what they can supply. These forecasts proved to be the basis for crucial
decisions, decisions that determine the success or failure of the future. With AI, more accurate
forecasting becomes apparent, thus by implementing AI forecasting, prosperity is therefore
enabled (1A). Additionally, the respondents answer indicates that AI has the capacity to refine
and automate companies, which can be crucial for them to sustain within the industry (1B).

5.2.1 Second-order code 1A: Increasing accuracy of forecasts to enable success

When companies work within SCM, a large part is about accurate forecasting (1A). Accurate
refers to how precise the estimation of the future demand and supply was being made, when
comparing the approximation with the actual result. When conducting forecasting,
participants explained how they often ask themself ‘what if?’, making up scenarios and
estimating about the future, in order to increase potential achievements and increase the
accuracy (1.a.1).

Forecasting was a crucial part of the SCM, both when predicting potential demand from the
market and when estimating what supply their production could generate. In order to do
forecasting in a more profitable way, using AI, the system can create scenario analyses faster
and more specific with using less amount of time (1.a.1).

“So forecasting is the part where we work a lot with AI, to be able to calculate demand &
supply for the following 24 months” - Respondent 1

“When you are in a supply chain role, [...] you want AI to be able to generate scenarios
because those are the ones that take the most time.” - Respondent 2

"It is an incredibly volatile market [...]. To get an analysis which may have generated that
"now it looks like this", "then you can expect that the prices should start going up". I think
that would be something good, I can spontaneously feel that it would be something good for
us" - Respondent 4

Subsequently, the scenario analysis and its forecast works as the foundation for business
decisions (1.a.1). Therefore implementing AI to create scenario analyses both is beneficial but
above all, should be accurate in order for companies to rely on it and enable success through
it. Accordingly, companies who implement AI which will generate successful estimation from
their scenario analysis, generate more prosperity in the long run.

"Above all, from my side then, is that we must get better forecasting. Everything that we
produce during pre-season, we produce based on forecasting" - Respondent 1

65
As already touched upon, an essential part of forecasting to enable success is to also evaluate
the outcome, to gain experience for the future (1.a.2). Our respondents indicated that an
evaluation between the expected outcome and the actual outcome is a great feedback and
something that AI needs to to be able to produce, since it's used widely (1.a.2). By doing
evaluation and providing feedback, the AI-system also learns from it in order to gain secure
success through forecasting.

66
"There are a lot of meetings, a lot of follow-ups and then also a lot of scouting, workshops,
how do we proceed? Where are we going? What kind of projects do we have that we will run?
and how do we ensure that we solve problems in our production"
- Respondent 1

"AI will probably be a recurring thing during the summer and at weekly meetings, what
should be changed and done." - Respondent 2

To conclude, by implementing AI, companies strive to achieve accurate forecasting, they will
then be able to learn from their mistakes and plan for the future in a more profitable way. This
was proven through our data to both be of high demand within SCM and to be a profitable
option than previous methods used.

5.2.2 Second-order code 1B: Refining methods and processes in order to sustain
Refining in order to sustain (1B) refers to how the implementation can improve certain
general areas within the SCM. However, in order to truly be transparent with our data, and
give context to our findings, we have also included how our respondent has explained their
current situation (1.b.1). Firstly, a lot of administrative work can be replaced with AI and the
employee can therefore focus their time on more fulfilling assignments (1.b.2). Additionally
with AI and in combination with SCM, companies are able to gain competitiveness (1.b.3),
which is of importance since they need to stay relevant in the manufacturing industry. Lastly,
our respondents expressed the importance of their production, which the implementation of
AI can aid with predicting disruption or interference more in detail (1.b.4).

Our respondents provided diverse perspectives on their current situations (1.b.1) regarding AI
implementation, which can be displayed in figure two. Some participants were in the process
of implementing AI and had recently witnessed the positive outcomes of their implementation
efforts. Others had made more significant progress along the AI spectrum. By considering
these varied viewpoints, we have gained valuable insights into the specific areas within SCM
where AI has demonstrated improvements.

The accumulation of knowledge gathered from these different perspectives spans across the
industry. Consequently, we have been able to identify the specific domains within SCM that
have benefited from AI implementation. It is worth noting that those who have fully
established AI within their operations recognize its significant advantages and believe that the
benefits outweigh any potential drawbacks. Conversely, participants who have not made as
much progress in AI development often perceive more obstacles than opportunities, which is
demonstrated in figure two.

Overall, our research has allowed us to comprehend the contrasting experiences and
perspectives related to AI implementation in SCM. This comprehensive understanding
enhances our ability to assess the benefits and challenges associated with AI adoption and
provides valuable insights for organisations considering or currently undertaking AI
initiatives in their supply chain management processes.

67
Figure 2: Illustrative image of the current situation

According to our participants in the supply chain management field, AI has played a
significant role in completing or has the potential to complete a substantial amount of their
administrative work (1.b.2). This use of AI technology serves as a valuable support for
employees, allowing them to save time on tedious assignments and redirect their focus
towards more rewarding and strategic work.

By automating administrative tasks through AI, SCM professionals can streamline their
workflow and increase overall productivity. Time-consuming activities such as data entry,
documentation, and repetitive data analysis can be efficiently handled by AI systems. This
enables employees to dedicate their time and expertise to more complex decision-making,
problem-solving, and relationship-building aspects of their roles.

The ability to offload administrative work to AI not only enhances efficiency but also
contributes to employee satisfaction and engagement. By relieving employees from mundane
and repetitive tasks, they can engage in more intellectually stimulating and fulfilling work,
leading to increased job satisfaction and higher morale within the SCM team. Furthermore,
leveraging AI for administrative tasks minimises the potential for human error and improves
accuracy and consistency in data processing and analysis. This, in turn, enhances the overall
quality of SCM operations and decision-making.

Overall, the integration of AI technology in SCM has the potential to significantly alleviate
the burden of administrative work, allowing employees to focus on higher-value activities.
This not only improves operational efficiency but also enhances employee satisfaction and the
overall effectiveness of SCM processes

"AI makes the efficiency greater, because I can polish my data and wait for the result, I don't
68
have to sit and work with my excel sheets. [...] Sitting and doing laboratory work and picking

69
takes a lot of time, and then you get less time to analyse what it was you got out and work on
those deviations." - Respondent 1

"From a Supply Chain perspective, I think AI is good, when you have more than one site and
you have several assumptions you have to make, then AI can help a lot because it's something
that previous tools can't help with at all. Without AI, you have to do manual calculations.” -
Respondent 2

"Mainly when it comes to efficiency, it's about lost working hours due to inefficient working
methods that we have today without the AI tool" - Respondent 2

In addition to making employees more time-efficient, AI implementation within supply chain

management (SCM) has the potential to enhance working methods and increase
competitiveness (1.b.3). By leveraging AI technologies, SCM professionals can optimise
processes, improve decision-making, and gain a competitive edge in several ways.

AI can analyse vast amounts of data and provide valuable insights and recommendations that
can inform strategic decision-making within SCM. It can identify patterns, trends, and
anomalies in data that may not be readily apparent to human analysts. These insights can
enable more informed and data-driven decision-making, leading to improved operational
efficiency, cost reduction, and better resource allocation. As previously mentioned, companies
can improve their forecasts by using AI and therefore provide more success for the company.
Further, they will also be able to use AI to improve their competitiveness, targeting markets
that they previously have not intended to, polishing current methods to become more efficient
and improving their margins to be able to provide greater customer service in terms of price
and time-frame.

“Since we work with such incredibly large sums and incredibly large volumes, AI is almost
something that's necessary, to continue to stay relevant in the world market.”
- Respondent 2

"If we don't have an idea on how AI can help more in a business and don't do AI, then we
won't survive against competitors. It's not just fun and games for the geeks, but it's really
business critical that we implement it" - Respondent 6

"Therefore, I mean that AI is competitive because… You want to be faster, better quality and
better trust, have a higher trust" - Respondent 5

By embracing AI and leveraging its capabilities within SCM, companies can transform their
working methods, optimise operations, and enhance their competitiveness in the marketplace.
The ability to make data-driven decisions, streamline processes, and leverage predictive
insights positions organisations for success in a rapidly evolving business landscape.

The final aspect highlighted by our respondents (1.b.4) is how AI aids in preventing issues
within the manufacturing industry. Given the industry's heavy reliance on production, AI
70
plays a crucial role in enabling SCM to operate proactively. By utilising AI technologies,
companies can predict potential machine shutdowns, identify when machines require
maintenance or repairs, and plan ahead accordingly. This proactive approach leads to
improved efficiency and mitigates production disruptions.

71
Through AI-driven predictive maintenance, SCM professionals can leverage historical and
real- time data from sensors and equipment to anticipate potential failures or breakdowns. AI
algorithms can analyse patterns and indicators of machine performance, providing early
warnings and insights into when maintenance or repair activities should be scheduled. This
allows companies to prevent unexpected production downtime and optimise the utilisation of
their manufacturing resources.

By effectively managing machine maintenance and repairs, companies can reduce unplanned
disruptions, optimise production schedules, and minimise the associated costs of downtime.
AI's predictive capabilities enable better resource allocation, ensuring that maintenance
activities are performed at the most opportune times, avoiding unnecessary interruptions to
production.

"During production, where you can also use AI in many ways, where you can see how
machines work, when they will break down and when I need to do maintenance. Cause no
factory feels good on stand by" - Respondent 1

"What we can't do is that we can't cause production in an industry like this to stop, because
then it won't be fun." - Respondent 3

“Actually implementing AI is because you want to maintain control over your production
[...]” - Respondent 2

Therefore, by leveraging AI to predict and prevent potential issues, companies can optimise
their SCM processes, reduce downtime, enhance production efficiency, and ultimately
improve their competitiveness within the manufacturing industry. The proactive nature of AI-
enabled SCM allows companies to stay ahead of potential disruptions and operate more
efficiently, ultimately benefiting both the company and its customers.

5.3 Aggregate dimension 2: Conserving financial funds

In order for manufacturing companies to have the opportunity to improve the SCM through
the implementation of AI, our data analysis shows that it is important to strive to save money
by effectively distributing the resources you have available. Instead of wasting or overusing
resources, it can be crucial to try to use them in a strategic and efficient way to reduce costs
and maximise value. What has been seen is that by implementing AI, (2A) resources can be
allocated in a better way so that one can improve their contingency management and to (2B)
preserve financial resources by optimisation in the SC.

5.3.1 Second-order code 2A: Enhancing the allocation of resources

When dealing with the SC, allocate the resources property can pose challenges and one
approach that has been shown from our data collection is that by applying AI (2.a.1), it can
decrease contingencies to improve data quality, and (2.a.2) a more consistent and even flow
can be fostered. In conclusion, companies therefore try to solve these challenges and allocate
resources better by implementing AI.

72
What our data analysis showed us was that within the SC, the occurrence of unexpected
events is normal. Unplanned incidents appear that the employees or manager needs to handle.
Often, they manage to do so, but the “abnormal event” makes their data quality poor. What
AI then

73
has been able to aid, is to prevent unpredictability by reducing contingencies. Therefore by
implementing AI, the SCM will be able to improve their data quality since they have less
contingencies (2.a.1) and provide better control and planning over it. With improved data
quality, they can allocate their resources more appropriately to the operational organisation.
By allocating their resources correctly, the financial resources can be saved.

“Throughout 2021 and large parts of 2022 there were problems with deliveries in, mainly due
to Asia and China closing and opening, closing and opening. So no goods came out of there.
It makes the history look very strange, if you compare it to a normal season. [...] that's our
biggest challenge now, to ‘clean’ our data and check 'What was the abnormality?'” -
Respondent 1

“If you throw in bad data, then you get bad data out, it's that simple.” - Respondent 4

“With data quality then, to do it well you have to have control over your data and you have to
have high quality. What is meant by high quality then? Well, you have to know what it
means, where it comes from and what it is supposed to fulfil." - Respondent 6

"We make a plan and then something happens a week in the future and then we have to throw
the whole plan away, the whole plan becomes invalid and we start over from scratch. There I
think that AI would have a good role [...]" - Respondent 2

Our respondents has emphasised the importance of establishing an even flow within the SC to
enhance efficiency (2.a.2). Our data analysis findings strongly support the notion that
companies with a consistent and smooth flow of operations can achieve higher levels of
efficiency. This efficiency, in turn, translates into several benefits, including the ability to
allocate more time and resources to other critical activities, ultimately resulting in cost
savings for the company.

“We win when we can automate the simple flow”- Respondent 3

“ and that's probably the first efficiency I miss - that I make time for a formal, manual thing,
instead of evaluating and optimising my team's goods flow as well as possible” - Respondent
2

“Instead of having an operator there who works 6 shifts and there is a new guy or girl every
shift who thinks differently from the other. AI becomes a simple flow that is fairly
standardised, and in this way it becomes more efficient and saves money.” - Respondent 3

The goal to create an even flow can be effectively achieved through the utilisation of AI. The
‘even flow’ that our respondents refer to are the processes and changes from when goods
enter production, until they are shipped, where they want to reduce any large variations and
deviations. This will create an even flow, making it more predictable and therefore
manageable. By leveraging AI technologies and algorithms, companies can optimise the
allocation of resources, minimise disruptions, and ensure a consistent and smooth flow of
74
materials, products, and information throughout the SC and also to save more money.
5.3.2 Second-order code 2B: Optimising the supply chain
The respondents emphasise the importance of preserving financial resources within the SC.
They suggest that one effective approach to achieve this is by maintaining inventory at an

75
appropriate level (2.b.1), which can greatly enhance inventory management and streamline
processes. This strategy offers several benefits, including cost reduction and improved
operational efficiency. Moreover, saving the most out of resources due to financial
considerations (2.b.2) becomes crucial as it enables substantial cost savings.

According to our respondents, the management of inventory at an appropriate level (2.b.1) is a

critical aspect of SCM. This is particularly significant in manufacturing industries, where
inventory levels tend to represent a substantial portion of the balance sheet. SCM endeavours
to optimise this aspect by ensuring that the required inventory is available while avoiding the
accumulation of unnecessary stocks, which can have a detrimental impact on financial
resources. By carefully controlling inventory levels, SCM aims to strike a balance between
meeting demand and avoiding excessive storage costs. It recognizes the importance of having
the right amount of inventory on hand to fulfil customer orders promptly and efficiently.
However, according to our respondents, AI holds the potential to revolutionise inventory
management within SCM.

"It costs a lot of money to sit on too much inventory, so there is a need for AI, to get a better
forecast model, so that we see that we are at reasonable inventory levels, and above all when
inventory optimization, that we have proper safety stock.”
- Respondent 1

"it always has to do with this balance between inflow from a process that in theory should run
24/7 and then balance the flow out with cost parameters [...] and then we theoretically
optimise inventory after that" - Respondent 3

The advent of AI presents remarkable opportunities to enhance inventory management. AI's

predictive capabilities, coupled with its ability to assist with demand planning and supply
planning, enable organisations to regulate and maintain optimal inventory levels more
efficiently. Additionally, AI can automate numerous processes, resulting in reduced time and
effort expended by employees.

Furthermore, our respondent also indicated that in order to preserve financial resources it is
crucial to optimise the general SC (2.b.2). The SCM always aims to save as much as possible,
to enhance their financial resources. Efficiently optimising the general supply chain is a
critical strategy for organisations to maintain a competitive edge and financial stability. By
streamlining processes, reducing waste, and eliminating inefficiencies, SCM aims to achieve
cost savings and resource optimization. The respondents mean that this can be accomplished
through AI.

"Our industry is quite old school, we don't have fixed market prices, so of course I want to
optimise as much as possible because otherwise the margin will be far too low and we won't
make it around, [...]. There you want the AI to think for itself in a suggestion perspective, you
want to be able to receive suggestions" - Respondent 2

"But it has to do with all that, save, become efficient and if you save money, maybe resources
76
have been saved, that you don't need to have had an extra man, an extra operator, or an extra
truck". - Respondent 3

77
To conclude, by leveraging AI algorithms and advanced analytics, the companies can gain
valuable insights and make data-driven decisions to enhance overall SC performance. They
will have the opportunity to save the most of the resources and conserve financial funds.

5.4 Aggregate dimension 3: Integrating previous approaches to mitigate emerging challenges

The last theme that has been developed from our data collection is previous approaches
challenging the acceptance of AI. This theme has been developed due to the fact that a lot of
the respondents have experienced challenges with both the implementation and the acceptance
of AI. The theme has therefore been developed into two parts, which is that different
approaches can make the transformation difficult (3A) and that it can be difficult to trust the
AI-system (3B).

5.4.1 Second-order code 3A: Reducing complicated transformation

Our data analysis has revealed that transitioning and switching systems to AI can be a
challenging endeavour, as indicated by our respondents. The process of adopting AI is
considered a complex internal process, presenting significant obstacles for companies aiming
to embrace AI successfully. Integrating AI into existing systems and workflows requires
careful planning, coordination, and technical expertise (3.a.1), which can make the process
daunting. Moreover, the nature of work within the respondent's domain adds another layer of
difficulty (3.a.2). The continuous variation in their work poses unique challenges when it
comes to transforming their operations with AI. The dynamic nature of their tasks and
responsibilities may necessitate adaptability and flexibility in integrating AI solutions. The
need to accommodate diverse workflows and ever-changing demands further complicates the
transformation process. Therefore AI

Implementing AI involves several key challenges, including selecting appropriate AI

technologies, integrating them with existing systems, ensuring data compatibility and quality,
and addressing organisational and cultural barriers. These complexities can result in
difficulties in achieving a smooth and successful integration of AI within the company's
operations. Overcoming these challenges requires careful planning, dedicated resources, and a
comprehensive understanding of the organisation's unique requirements and goals (3.a.1).
Companies must invest in the necessary expertise, training, and change management
processes to navigate the complexities of AI implementation and ensure a successful
transformation.

The results indicate that the existing internal logistics or organisational structure can act as a
barrier to the successful implementation of AI. This suggests that simply implementing AI
into an unchanged structure may not yield the desired improvements in SCM. Instead, a new
structure and planning procedure may be necessary to fully leverage the benefits of AI in
SCM.

By introducing AI, companies have an opportunity to reevaluate and redesign their internal
logistics and processes. AI can provide insights, recommendations, and automation
capabilities that enable more efficient and effective planning, coordination, and execution of

78
SCM activities. This may involve redefining roles and responsibilities, optimising workflows,
and aligning the organisational structure to better support AI-enabled SCM practices.

Implementing AI-driven solutions often requires a holistic approach that encompasses not
only technological aspects but also organisational and cultural changes. This includes
fostering a data-driven culture, promoting collaboration across departments, and providing
training and

79
support for employees to adapt to the new ways of working. By reimagining the structure and
planning procedures within the company, organisations can harness the full potential of AI in
improving SCM. The integration of AI as an organiser can help untangle complicated internal
processes and pave the way for enhanced operational efficiency, better decision-making, and
improved competitiveness.

"It is our internal logistics that makes it complicated." - Respondent 3

"If we throw in a bad sales plan, it doesn't matter how much we look at all the generated
parameters, we have to have a good plan in as well and a well-crafted plan.”
- Respondent 1

“When it gets too complicated, usually, then the automation doesn't help much. It will not be
as easy to achieve something good, if it happens all the time, or for complicated parameters.”
- Respondent 2

Based on our analysis, we have identified a common challenge faced by many manufacturing
companies operating in SCM when it comes to integrating AI. This challenge arises from the
substantial day-to-day variations and deviations encountered in their operations (3.a.2).

The dynamic nature of SCM operations introduces complexity and unpredictability, making it
more challenging to seamlessly integrate AI solutions. These variations can stem from factors
such as fluctuating customer demands, changing market conditions, supplier disruptions, and
evolving production requirements. These fluctuations pose significant obstacles when
attempting to implement AI systems that rely on consistent and stable data inputs.

To effectively integrate AI into SCM operations, companies must account for these variations
and deviations. They need to develop AI models and algorithms that can adapt and respond to
real-time changes, making use of available data to generate accurate and relevant insights.
This may involve incorporating machine learning and advanced analytics techniques that can
dynamically adjust to the evolving conditions of the manufacturing environment.

Furthermore, companies must ensure that their data collection processes capture the necessary
information to account for variations and deviations. Robust data management practices,
including data cleansing, normalisation, and validation, become essential to handle the
diversity and variability present in SCM operations. Addressing these challenges requires a
comprehensive understanding of the specific variations and deviations encountered in SCM
processes. By incorporating AI technologies that can effectively handle these dynamics,
companies can harness the full potential of AI to optimise operations, improve decision-
making, and enhance overall supply chain performance.

"There are no normal days for me, I can think I will do one thing when I come to work and
then it becomes something else entirely" - Respondent 1

"So the optimization that we do, which says that ‘this is how we should always do it’. But that
80
"always" doesn't exist. Of course, we have a plan to try to put that product on one site and the
next product somewhere else. But, two, five or eight times, we put something else there,
because we had to." - Respondent 3

"Above all, you can see that the work from day to day is changing." - Respondent 1

81
While the process of transitioning AI into SCM can be demanding, it also presents
opportunities to leverage technology. Companies that effectively address these challenges can
unlock the immense potential of AI to optimise operations, including managing and adapting
to the day- to-day variations and deviations. In addition they will be able to improve decision-
making, and gain a competitive advantage in their industry.

5.4.2 Second-order code 3B: Working on increasing trust to the AI-system

Another challenge of accepting AI is the lack of trust in the system (3B). It has emerged that
when implementing AI, it does not always live up to the expectations as thought (3.b.1). Also,
if there is no transparency people are starting to be suspicious and uncertain (3.b.2), why is
this good for us? This lack of openness results in employees and managers still using old
systems and methods. Additionally, certain respondents have reported encountering
challenges during the implementation of AI, which include factors like maintenance demands,
time commitments, and heightened complexity. The objective is to streamline the roadmap to
ensure a clear vision of the end goal (3.b.3).

Implementing AI can bring about high expectations for organisations seeking to leverage its
potential benefits. However, some respondents have found that the reality of AI
implementation does not always align with their initial expectations (3.b.1). Challenges have
arisen when the companies have encountered situations where AI does not perform as they
anticipated which has led to that they feel some annoyance towards AI.

"[...] it happens quite often that something starts spinning, then we don't get all the data in on
time and you experience a frustration when you come to work and these runs are not finished
and then you get a little discouraged." - Respondent 1

“I take out a sort of laser-read volume and still fact-check it, because AI has been wrong so
many times.“ - Respondent 2

So even though AI has its upside, frustration and negative experience has emerged when
using the system. Therefore in order for the implementation to be beneficial for SCM, more
contributions to developing the system's accuracy are acquired, so expectations from
employees and users are encountered.

Moreover, the aspect of transparency has emerged as a significant concern among some of the
respondents and they therefore are suspicious towards the AI-system (3.b.2), as its absence
creates an environment of uncertainty within companies. This suspicion makes employees
hold onto old methods and techniques, which work against the integration of AI. Due to a lack
of basic understanding or familiarity with the new AI system, employees are facing
challenges in trusting the system. As a result, they find themselves working twice as hard to
correct and verify the data generated by the system, aiming to ensure its accuracy.

“People don't trust data, especially not data they don't understand how it came about. You can
implement very advanced models, which spit out a result, but if you don't understand where

82
the result comes from, then you have uncertainty." - Respondent 1

“Do I just want to go in and trust those numbers without knowing how it's happened behind it?
[...] I want to know the background. [...] if I have a forecast this week that changes to next

83
week, I have to be able to understand, explain and defend the differences between them, to the
other departments.” - Respondent 2

What our respondents have expressed is their need for transparency and more communication
regarding what AI produces, in order for them to understand and grasp the decision-process
the AI-system has explored.

It's very much a communication issue as well, not just how you should know techniques or
something but a lot of communication. And trust each other and the technology. - Respondent
5

For the companies and employees to place trust in an AI system, it becomes essential that they
have a clear understanding of its underlying principles. Without comprehending the system's
background, establishing trust becomes a challenge. Working on increasing trust becomes
therefore crucial when integrating previous approaches with AI.

The last building-block for 3B revolves around the respondents' shared experience regarding
the complexity involved in introducing AI, as it demands considerable time and maintenance
commitments for companies. People tend to see the barriers of implementing AI due to the
complexity, instead of what AI can contribute in the long run. (3.b.3). So there is a challenge
of acceptance in AI for the respondents.

The final component of 3B revolves around the collective experience of respondents

concerning the complexity associated with the adoption of AI. Implementing AI requires
significant time and ongoing maintenance commitments, which companies often perceive as
barriers rather than recognizing the long-term benefits AI can provide (3.b.3). Consequently,
there is a challenge of acceptance among respondents regarding AI. To address this challenge,
a clearer roadmap is necessary to guide companies in integrating AI into their systems. Our
earlier analysis of aggregated dimensions reveals that AI can bring substantial advantages to
companies, such as time savings, reduced complexity, and cost savings. However, despite
these benefits, respondents currently hesitate to implement AI due to concerns about
complexity, time investment, and financial costs which presents a problem that needs to be
addressed.

“In this particular department, it's too early now, in 5 years - absolutely, but now we let
everyone else test AI first. We can see with the kind of manufacturing that we have, we're not
going to benefit because it's too complex.” - Respondent 6

"perhaps that AI is not as easy to implement in this industry” - Respondent 3

"This is too complex for us now to invest in because it doesn't have the short-term impact
because our business is not that mass-produced." - Respondent 5

"I think that if we look at continuous production, it would take time before setting up such a
system. You would have to feed it with basic data and only know which data is relevant to the
84
system and to obtain in order to then be able to process it. I think that would be a pretty
complicated task.” - Respondent 4

"I can imagine that AI can be quite a lot of work to get to the basics, so to speak." -
Respondent 2

85
"People have understood what AI is now and you have to invest time in it, in order for it to
work.” - Respondent 5

In conclusion, the adoption of an AI system presents a significant barrier for companies

aiming to transform their operations. The uncertainty and the complex investment that this
requires makes it not worth it because they don't know if it will be good or if you can trust
that it will work. To overcome this obstacle and facilitate a smoother transition, it is crucial to
establish a streamlined roadmap that provides a clear vision and goals. Such a roadmap
ensures that the company can place trust in the transformation process and navigate the
complexities of integrating AI effectively.

86
6.0 Discussion and Theory Elaboration

In this chapter, we have undertaken a comprehensive discussion of the previously presented

findings. The chapter is structured into individual discussions for each aggregate dimension,
which are then compared to the existing literature, revealing both similarities and differences.
Moreover, the combination of these discussions will contribute to the elaboration of a theory.
Through this theory elaboration, we will present a proposed theory that addresses the research
question, how the implementation of AI can improve the SCM.

6.1 Automating production for more prosperity

Consistent with prior research, our findings reinforce the notion that AI holds promise in
enhancing the SCM (Sithole et al., 2016, p. 18). Both existing literature and our own study
underscore the advantages of AI in optimising production through the analysis of production
data and offering recommendations for process enhancements (Min, 2010, pp. 16-17).
Additionally, AI aids in predictive maintenance by analysing sensor data to anticipate
maintenance requirements (Min, 2010, pp. 16-17). Being able to forecast potential
circumstances has shown to be of great advantage from both our results and literature (Bughin
et al., 2017, p. 22; Li et al., 2020; Piplani et al., 2021). It becomes reasonable why the
manufacturing industry strives to automate and implement AI. By implementing automation
within SCM, manufacturing companies gain the ability to swiftly adapt to evolving market
demands and customer preferences, they can create both potential demands and supply from
the market. Through accurate scenario analyses and forecasting, automation empowers
companies to make informed decisions and respond effectively to changes. Integrating
automation in production and SCM, coupled with receiving feedback and evaluations, has
enhanced companies’ adaptability. This increased adaptability has the potential to drive
prosperity by enabling companies to proactively meet customer needs, optimise operations,
and capitalise on emerging opportunities. Hence, the driving forces from manufacturing
companies to reduce costs, accurate forecasting and fainting competitive edge within the SC
can all be pursued by implementing AI (Sithole, 2016, p. 18; Thun et al., 2021, p. 727).

What previous studies have failed to highlight is the importance of feedback and evaluation
even after the implementation of AI. Reim et al., (2020, p. 188) emphasise how feedback and
evaluation during the process of implementation is extremely important. What our study has
shown is that feedback and evaluation is something that is required in SCM, not as a tool to
increase the AI-system, but as an instrument within AI to improve SCM. AI should be able to
generate feedback and evaluation, not only receive it, in order for SCM to be improved.

In addition, there appears to be a potential disparity between the existing literature and our
own results. Prior studies (Bughin et al., 2017, p. 22) assert that AI-based approaches can lead
to quantifiable reductions in errors and cost savings. However, our findings do not provide
specific numerical evidence to support such claims. While our study acknowledges the
advantages of AI, it remains uncertain to what extent it has improved production, reduced
errors, increased savings, and saved time. Consequently, the return on investment (ROI) for
AI implementation remains indefinite. Given the substantial investments being made into AI
(Schmidt et al., 2021,
87
p. 170; Strohm et al., 2020, p. 5530; Schaeffer et al., 2018, p. 4-5), the ROI becomes
increasingly crucial, especially considering the significant implementation costs associated
with AI (Schmidt et al., 2021, p. 170). This contradiction between the literature and the
practical implementation of AI arises from the fact that while AI can refine and automate
processes to enhance a company's competitiveness (Thun et al., 2021, p. 727), the true
impact behind

88
increased competitiveness and prosperity remains unclear, even though everything points
towards AI.

6.2 Conserving financial funds

The implementation of AI systems offers manufacturing companies the opportunity to

streamline their SC operations, but effective resource management and cost-saving measures
are crucial for success. Our data analysis emphasises the significance of this aspect. Existing
literature supports the idea that AI development and implementation in the industry can lead
to increased productivity, cost reduction, and improved decision-making processes (Pannu,
2015,
p. 79). However, the literature also acknowledges the high costs and time-consuming nature
of AI implementation. It is recognized as a significant investment and a potential barrier for
companies (Huang et al., 2018). Additionally, within SCM, the technical complexity further
adds to the implementation costs (Kohli et al., 2021b). Our respondents and the literature
agree that implementing AI has the potential to yield long-term time and cost savings.
However, previous research also highlights the substantial financial and time requirements
involved in AI implementation. This creates a decision-making barrier, as companies must
carefully evaluate whether the expected benefits of AI in terms of future time and cost savings
outweigh the initial investment and time commitments.

Moreover, our findings reveal that it is crucial to allocate the resources to manage
contingencies and by incorporating AI, the SCM can enhance the quality of their data as it
reduces the number of contingencies they encounter. This aligns with existing literature that
argues that predictive maintenance is an advantage of implementing AI within the
manufacturing industry (Kumar et al., 2019b). Through the utilisation of ML algorithms to
analyse sensor data from various devices, AI has the capability to anticipate potential machine
failures or other unforeseeable events. However when it comes to data quality, the literature
and our results contradict each other. According to Kumar et al. (2019b), literature indicates
that AI can assist in identifying abnormalities. However, our findings reveal that
manufacturing companies encounter challenges in SCM that are related to this issue. Instead
of relying on AI alone, companies in the manufacturing industry need to provide input to the
system, specifying faults or abnormalities. Unfortunately, this process degrades the data and
renders it ineffective.

Our findings also indicate that the respondents in our study recognize the significance of
having an even flow within the SC to ensure efficiency and the utilisation of AI enables the
attainment of this objective as a viable goal. This finding is consistent with existing literature,
which suggests that the adoption of AI-based systems can play a crucial role in achieving such
a smooth flow (Gao et al., 2020). By leveraging AI technology, companies can optimise
various aspects of their supply chain operations and overcome potential bottlenecks or
disruptions. Additionally, as Wang et al. (2019) discovered, implementing an AI-based
inspection system can reduce the inspection time by 60% and improve defect detection
accuracy by 20%, ultimately improving the overall quality of products and the flow in the
system. Overall, the convergence of our study's findings with the existing literature

89
underscores the importance of an even flow in the supply chain and highlights the potential
role of AI-based systems in achieving this objective. By streamlining processes, reducing
inspection time, and improving defect detection accuracy, AI technologies have the capacity
to enhance operations and facilitate a smooth flow throughout the supply chain. However,
previous evidence from our result demonstrates that when the initial data quality is
inadequate, predictive maintenance may provide erroneous indications, rendering AI unable
to generate or establish a consistent flow.

90
Furthermore, the literature highlights the significance of utilising ML algorithms for
optimising inventory management as a means to conserve financial resources. By leveraging
these algorithms, companies can analyse inventory levels and accurately forecast demand,
allowing them to ensure the availability of the right materials at the right time. This
optimization leads to cost reduction and minimization of waste, aligning with the goals of
improving the SCM (Papageorgiou, 2009, p. 1931-1932; Pournader et al., 2021, p. 7). The
findings from our respondents further reinforce the importance of effective inventory
management within SCM, particularly in the context of manufacturing industries. Maintaining
the appropriate inventory levels becomes crucial for balancing supply and demand, as
excessive stocks tie up financial resources and can lead to unnecessary costs. Conversely,
inadequate inventory levels can result in stockouts and customer dissatisfaction. Therefore,
SCM strives to achieve the optimal balance, ensuring the availability of required inventory
while minimising the storage of surplus stocks. In this regard, AI emerges as a valuable tool
to help attain the goals of effective inventory management. In addition, Pournader et al.
(2021, p. 7) is also saying that ML algorithms can analyse the availability of raw materials
needed for production and predicts that there may be a shortage in the coming weeks which
can solve the problem with both mange unforeseeable events, creates an even flow and
optimise inventory.

Lastly, according to our respondents, optimising to save money is also essential for the
companies. They discuss the importance of allocating resources and optimise the SC with AI
in order to enhance their financial resources. Existing literature is highlighting that when you
streamline the SC with AI and the production quality, reduce defect rates and increase
production output the manufacturing companies can achieve cost savings and improve
customer satisfaction simultaneously (Wang et al., 2019, p. 1475). This streamlining of
operations leads to cost reduction and improved utilisation of resources. Overall, optimising
the SC with the aid of AI can have significant financial benefits for manufacturers according
to both existing literature and our findings. Efficient allocation of resources and strategic
adoption of AI can lead to cost savings and enhanced financial resources for companies.
However, challenges related to data quality and the need for manual input in AI systems
should be addressed to ensure their effectiveness.

6.3 Integrating previous approaches to mitigate emerging challenges

Numerous studies have emphasised the complexity associated with applying AI in SCM
(Dash et al., 2019, p. 44; Van Hoek, 1998). However, it is important to note that the definition
of complexity found in the existing literature may not align with the definition derived from
our own research findings. While prior studies have likely explored various dimensions of
complexity in AI-based SCM, the specific context and scope of our research may have
uncovered unique aspects of complexity. The complexity expressed by our respondents
primarily revolves around the variation and dynamic nature of their work, where no two days
are alike. In contrast, the existing literature portrays SCM as a cross-functional department, to
highlight its inherent complexity. This difference in perspectives can be attributed to the
specific context of our research and the experiences shared by the respondents. The day-to-
day operations within SCM can involve managing multiple stakeholders, coordinating various

91
activities, and navigating uncertainties, which contribute to the perceived complexity
described by our participants. While the literature may emphasise the cross-functional nature
of SCM (Dash et al., 2019, p. 44), it is essential to recognize and incorporate the real-world
experiences and perspectives of practitioners. By acknowledging the unique complexities
encountered by individuals within SCM, we gain a more comprehensive understanding of the
challenges and intricacies involved in effectively managing supply chains. This divergence
could be attributed

92
to differences in research methodologies, data sources, or the specific industry or supply chain
under investigation. Therefore, it is crucial to consider and acknowledge the potential
variance in defining complexity in AI-SCM. By recognizing this distinction, we contribute to
the broader understanding of the multifaceted nature of complexity in SCM and highlight the
need for further investigation and refinement of the concept in future studies.

In addition, implementing AI systems has shown to meet a certain resistance from the
industry, within SCM, underlined by both our results and by literature (Huang et al., 2018;
Sharma et al, 2021, p. 3). According to literature the resistance is mainly from the employees
due to the significant changes to existing processes and systems that AI brings with it.
However, our results did not indicate that the employees had resistance towards the changes,
more that it is the previous processes and systems that are hard to transform into AI, while the
employees outlook was positive towards change and streamlining with AI. Our results
explained how previous procedures and internal logistics makes the adaptation or the
integration more difficult.

Our findings align with the viewpoints expressed by Sharma et al. (2021, p. 3) and Kumar et
al. (2019a), who argue that the implementation of AI can be challenging, particularly due to
issues related to data quality and its impact on outcomes. These studies suggest that AI may
not always perform as intended and that errors in AI-driven processes can have serious
consequences, such as disruptions in the SC or incorrect inventory management. The
disappointment expressed by respondents in our study regarding AI's failure to meet initial
expectations is consistent with the concerns raised by Sharma et al. (2021) and Kumar et al.
(2019a). One key factor highlighted by these studies is the critical role of data quality in AI's
performance. If the data used to train AI models is incomplete, biased, or of poor quality, it
can significantly impact the accuracy and reliability of the outcomes. The disappointment
expressed by respondents when AI fails to meet initial expectations highlights the need for
careful consideration of data quality, realistic expectations, and ongoing monitoring and
improvement of AI systems.

Furthermore, what our study showed, similar to literature, was that there is a suspicion
towards AI and when its been implemented into a company (Wang et al., 2021; Nagurney et
al., 2021,
p. 400; Reim et al, 2020, p. 182; Thun et al., 2021, p. 736). Our results indicated that the
reason why employees are suspicious is because of the lack of transparency that AI holds.
Literature often explains that transparency is one of the challenges when it comes to AI (Reim
et al., 2020,
p. 182; Hasija & Esper, 2022, p. 389). Often when a process has been executed by AI, the
user of the system gets no insight to why and/or how AI has generated the outcome. What
literature then argues for is that transparency and explainability is a combination of being able
to interpret the knowledge as well as trustworthiness of the systems (Reim et al., 2020, p.
182). What our result demonstrated was that the lack of transparency and explainability
results in employees returning to old methods and processes, to assure AI has considered
everything and computed correctly. Due to the frequent involvement of suppliers and
customers in supply chain decisions, the external visibility and associated risks of AI outputs
93
are generally higher (Hasija & Esper, 2022, p. 390). Which we argue is why there is a higher
level of suspicion surrounding AI in supply chain management compared to other areas in
business.

As explained previously by Thun et al. (2021, p. 736), addressing the challenge of

establishing trust in AI systems is a complex task that necessitates several considerations.
Companies aiming to instil trust in AI systems must focus on ensuring compatibility with
existing systems, resolving network speed and stability issues, and addressing concerns
related to data security. People will not be able to trust the system, if they can not understand
the background (Hasija

94
& Esper, 2022, p. 389), which our result also indicated. This result shows how literature's
estimation of overreliance on AI might be mistaken. Kohli et al. (2021b) explains that a risk
with implementing AI, is that individuals might over-rely on AI, resulting in complacency
and a lack of human oversight, which is the complete opposite to our result.

Continuously, the lack of popular acceptance of certain AI applications has sparked a debate
regarding the factors contributing to their failed implementation. Some argue that poor
technical performance is the primary cause, while others point to organisational and/or social
characteristics (Strohm et al., 2020, p. 5526; Frank et al., 2019, p. 80). Trust is influenced not
only by the technological aspects but also by the company responsible for its implementation
and its effectiveness in communicating relevant information (Reim et al., 2020, p. 182). Our
results align with this observation. Respondents often expressed two key aspects: (1) working
in parallel with the system or (2) controlling the outcome of AI. This is primarily due to their
lack of basic understanding or knowledge of the new AI system in comparison to the familiar
old system. Additionally, respondents emphasised the importance of collecting all relevant
data required for the AI system to function effectively. That is why we argue that companies
implementing AI lack the action of actually integrating AI. Our findings suggested, as well as
literature, that there might be both a social and organisational aspect to the implementation of
AI (Strohm et al., 2020, p. 5526; Frank et al., 2019, p. 80). To manage that, we argue that
investments need to be done towards integrating AI, not only implementing it.

Lastly, our findings indicate that companies within the manufacturing industry aspire to adopt
AI technology in order to streamline their everyday operations and simplify their systems,
resulting in cost reduction and time savings. However, as mentioned before, the
implementation of AI presents challenges in terms of complexity, time commitment, and
increased costs (Huang et al., 2018). These factors create a sense of indecisiveness among
companies regarding the suitability of AI for their specific needs. This uncertainty stems from
the need for careful consideration of factors like the company's resources, technical
capabilities, and readiness for change. Furthermore, our result shows that companies in the
manufacturing industry believe that SCM in this industry is far too complex to do anything
about it now. You can see that the respondents think it's something good and will probably
come in the future, but right now it doesn't work because it's because it's a far too complex
industry. This is contrary to previous literature that indicates the urgency of implementing AI
in order to maintain market presence and competitiveness (Sharma et al., 2021, p. 7-9;
Pournader et al., 2021, p. 7). Pournader et al. (2021, p. 7) also indicates that ML in
manufacturing is expected to grow a lot in the coming years. Interestingly, despite this
recognition, there seems to be a prevalent wait-and-see mentality among companies. Many
are hoping that someone else will pioneer the solution, making it easier for others to follow
suit. Consequently, those with sufficient capital are poised to seize the opportunity early,
securing market share and gaining a competitive edge.

6.4 Automating for more prosperity and conserving financial funds, comes at the cost of
integrating previous approaches

Our discussion in combination with our empirical findings have facilitated the elaboration of
a theory. The inductive approach has faced criticism for its perceived lack of starting close
95
enough to the phenomenon being studied, as pointed out by Shepherd and Sutcliffe (2011, p.
363). However, in the context of our research on how the implementation of AI can improve
SMC, we have employed a thoughtful data analysis using a grounded theory approach. This
approach has allowed us to develop a theory that is closely aligned with the data collected.
Throughout our findings we have distinguished two aggregated dimensions which point

96
towards how the implementation of AI improves SCM, whereas the third aggregated
dimension portrays the gloomy side of AI. The pursuit of automating processes in order to
achieve prosperity and conserve financial resources through the implementation of AI does
come with the challenge of integrating previous approaches. While AI offers significant
potential for improvement, it often requires companies to navigate the complexities of
incorporating this new technology into their existing systems, culture and workflows.

Integrating AI involves considering the compatibility and cohesiveness of the new AI systems
with the pre-existing processes and technologies. This may require modifications or
adaptations to ensure seamless integration and minimise disruptions. Companies must assess
the impact of AI implementation on their current operations, data infrastructure, and
workforce to identify potential gaps or conflicts that need to be addressed. The integration
process can involve challenges such as data compatibility, system interoperability, and the
need for retraining or upskilling employees. Companies may need to invest in new
infrastructure, update their data management practices, or establish communication protocols
between AI systems and existing tools. This integration process requires careful planning,
effective change management, and collaboration across departments.

Further, organisations must consider the cultural, social and organisational aspects of
integrating AI. Resistance to change and the need for cultural adaptation can pose hurdles to
successful integration. Efforts should be made to involve employees in the process,
communicate the benefits of AI integration, and provide training and support to ensure a
smooth transition and increase trust towards the AI-system. It is evident that employees
recognize the long-term improvements that can be achieved through AI implementation. They
understand the potential benefits and advantages that AI can bring to their work
processes and overall organisational performance. However, our research also indicates that
there are significant challenges that hinder the successful implementation of AI. One of the
key challenges identified is the volume and complexity of the tasks involved in implementing
AI. The process of integrating AI into existing systems and workflows can be complex and
time-consuming. It requires careful planning, coordination, and technical expertise to ensure a
smooth and effective implementation. This challenge can pose a barrier to organisations
seeking to adopt AI, as the resources and efforts required may be substantial.

Additionally, the successful implementation of AI often entails significant investments.

Companies need to allocate financial resources for acquiring AI technologies, developing
customised solutions, and providing training and support for employees. The costs associated
with implementing AI can be a deterrent for organisations, especially for those with limited
budgets or competing investment priorities. While integrating AI may require adjustments and
investments, the potential benefits it offers, such as increased efficiency, improved decision-
making, and cost savings, can outweigh the challenges. Organisations that approach
integration with a strategic mindset and a focus on aligning AI with their existing approaches
can pave the way for a successful and prosperous transformation.

Addressing these challenges may involve strategic planning, setting realistic expectations, and
developing a phased approach to AI implementation. It is important to prioritise key areas
97
where AI can have the most significant impact and to allocate resources accordingly.
Additionally, organisations may explore partnerships or collaborations with AI solution
providers or leverage external expertise to mitigate the challenges and accelerate the
implementation process. By acknowledging the challenges and taking proactive measures to
address them, organisations can overcome the barriers to successful AI implementation and

98
reap the long-term benefits and improvements that AI offers. This may involve building a
supportive organisational culture, providing training and upskilling opportunities for
employees, and establishing clear communication channels to manage expectations and
facilitate the adoption of AI technologies.

In summary, while there are challenges associated with integrating previous approaches when
implementing AI, a thoughtful and comprehensive approach to integration can lead to long-
term prosperity by harnessing the full potential of AI technology while leveraging existing
strengths and resources.

99
 RESEARCH METHODOLOGY

Research
Research comprises “creative work undertaken on a systematic basis in order to increase the
stock of knowledge, including knowledge of man culture and society, and the use of this stock
of knowledge to device new applications”. It is used to establish or confirm facts, reaffirm the
results of previous work, solve new or existing problems, and develop new theories.

Research Methodology
Research methodology means providing an accurate, detailed account of the methods and
procedures that are used in the study or the experiment. Research methodology deals with the
research design used and methods used to present the study.

1.6.1 RESEARCH DESIGN

According to F.N. Kerlinger research design is the plan, structure, and strategy of investigation
conceived so as to obtain answers to research questions and to control “variance”
A research design is a systematic plan to study a scientific problem. The design of the
study defines the study type (descriptive, correlational, semi-experimental, experimental,
review, meta-analytic) and sub-type (e.g.: descriptive, longitudinal case study), research
question, hypothesis, independent and dependent variables, experimental design and if
applicable, data collection method and statistical analysis plan. Research design is the
framework that has been created to seek answers to research.
A research design is an arrangement of conditions for collection and analysis of data in a manner
that aims to combine relevance to the research to the economy in procedure. The research design
is the conceptual structure within which research is conducted; it constitutes the blueprint for the
collection, measurement and analysis of data.
Simply It is a basic plan, that guides the data collection and analysis phases of the project. It
provides a frame work, which specifies the type of information to be collected, the resources of
data collection procedure define research design

. Descriptive Research:
In this research it deals with descriptive research type. It includes survey and fact finding
enquiries of different kinds. The major purpose of descriptive research is description of the state
of affairs as it exists at present. It is the study designed to depict the participants in an accurate
way.it is the most widely-used research design as indicated by the theses, dissertations and
10
0
research report of institutions its common means of obtaining information include the use of the
questionnaire, personal interviews with the aid of study guide or interview schedule, and
observation either participatory or not.
In social and business research, we quite often use the term ex post facto research for
descriptive studies. The main characteristics of this method is that the researcher has no control
over variable, he can only report what is happening of what has happened. Most ex- post facto
research projects are used for descriptive studies in which the researcher seeks to measure such
items for example, frequency of shopping, and preference of the people over similar data.

Research design adopted for this study is descriptive research. Descriptive method was adopted
because it deals with description of the state of affairs as exists at present.

STATEMENT OF THE PROBLEM

“A study on pros and cons of AI on the performance of supply chain management
practices of service sector with special reference to Coimbatore city

PERIOD OF STUDY
The period of the study is for 3 months.
SAMPLING DESIGN
Sampling method refers to the rules and procedures by which some elements of the population
are included in the sample. Some common sampling methods are simple, stratified, and cluster
sampling.
POPULATION
Population taken for this study was mainly concentration on employees of service sector in
Coimbatore city. Population of the organization was 200 employees.
SAMPLE SIZE
Sampling is a process of selecting a subset of randomized number of members. The sample size
of my study was 100 employees. This phase consists of the data analysis of the data collected
based on the satisfied random probabilistic to sample technique. The data collected were
analyzed using the following methods.
SAMPLING TECHNIQUES
The collecting data is the sampling process. When a decision is the made to use the sample a
number of factors must be taken to consideration
In this study I’m going to use simple random sampling method.
Simple random sampling:
A simple random sample is a subset of individuals chosen from a larger set. Each individual is
10
1
chosen randomly and entirely by chance, such that each individual has the same probability of
being chosen at any stage during the sampling process, and each subset of individuals has the
same probability of being chosen for the sample as any other subset of individuals. This process
and technique is known as simple random sampling.
The various steps involved in the sampling process are,

▪ Identifying target population

▪ Determining sampling frame

▪ Selecting sampling procedure

▪ Determining sampling size

Obtaining information from respondent

SOURCES OF DATA
There are several ways of collecting primary and secondary data. Questionnaire method is used
here for collecting information from the employees.

⮚ Primary data

⮚ Secondary data

Basic search for data: -

PRIMARY DATA: Information collected from employees through Questionnaire with
employees and discussion held with HR manager.
SECONDARY DATA: Website, Books and Company records, employee records
PRIMARY DATA
Primary data were collected by the researcher from their original sources. Primary data are first-
hand information collected through various methods such as observation, interviewing,
questionnaires” and so on. The primary data are those which are collected as a fresh and for the
first time and which consist of original in character. The data are collected by listing to the
general talks with department heads and also concentrated on direct observation. Primary data is
a data which is collected first time from the field. It is collected for specific purpose of solving a
particular research problem.
Primary data is information collected by the researcher directly through instruments such
as interview, observation, questionnaire and schedule.
Primary data collected for this study is through the following methods

10
2
▪ Questionnaire method

▪ Interviewing method

Questionnaire Method:
It is a research instrument consisting of series of questions for the purpose of gathering
information from the employees. In this method the questionnaire has been prepared and
distributed among the employees of service sectors

▪ Interviewing Method:

Under the interviewing method, the information is collected by way of direct asking
questions to the supervisor. It is the verbal conversation with the supervisor, with the
objective of collecting relevant information for the purpose of research
SECONDARY DATA
Secondary data means data that are already available that is they refer to the data which have
already been collected and analyzed by someone else: when research utilizes secondary data,
then he has to look in to various sources from where he can obtain them. In this case, they
certainly not confronted with the problems that are usually associated with the collection of
primary data.
STATISTICAL TOOLS

⮚ Simple percentage analysis

⮚ Chi-square test

SIMPLE PERCENTAGE METHOD

Simple percentage is one of the tools used for this study. This method is nothing but the average
of collected frequency. It is one of the simple and prominent methods used for analysis. The
value is calculated from the following formula
Equation
Number of respondents
Simple Percentage = * 100
Total number of respondents

10
3
CHI-SQUARED TEST
A chi-squared test, also referred to as test (or chi-square test), is any statistical hypothesis test in
which the sampling distribution of the test statistic is a chi-square distribution when the null
hypothesis is true. Chi- square tests are often constructed from a sum of squared errors, or
through the sample variance. Test statistics that follow a chi-squared distribution arise from an
assumption of independent normally distributed data, which is valid in many cases due to the
central limit theorem. A chi-squared test can be used to attempt rejection of the null hypothesis
that the data are independent.
Chi-square test is one of the simple and most commonly used non-parametric tests in statistical
work. Also considered a chi- square test is a test in which this is asymptotically true, meaning
that the sampling distribution (if the null hypothesis is true) can be made to approximate a chi-
square distribution as closely as desired by making the sample size large enough.
The chi-squared test is used to determine whether there is a significant difference between the
expected frequencies and the observed frequencies in one or more categories.

Equation
Formula for Chi-square test:

∑ (O-E)2
Chi-square test =X2= E
E
Where,
O = Observed frequency
E = Expected frequency

Expected frequency = (Row total * column total)/ grand total.

Degree of freedom = (c-1) (r-1)
Where,
C = No of columns
r = N o of rows

The level of significance is 5%.

10
4
10
5
 Gender
Valid Cumulative
Frequency Percent Percent Percent
Valid Male 79 79.0 79.0 79.0
Female 21 21.0 21.0 100.0
Total 100 100.0 100.0

INTERPRETATION
From the above table it is revealed that 79% of the respondents are male and 21% are female.

10
6
 Age
Valid Cumulative
Frequency Percent Percent Percent
Valid Below 30 59 59.0 59.0 59.0
30-40 32 32.0 32.0 91.0
41-50 5 5.0 5.0 96.0
51-60 4 4.0 4.0 100.0
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 59% of the respondents falls under the age group of 30.
32% of respondents falls under the category of 30-40. Remaining 5% and 4% falls under the
category of 41-50 and 51-60 respectively

10
7
 Educational Qualification
Valid Cumulative
Frequency Percent Percent Percent
Valid Plus two 2 2.0 2.0 2.0
Degree 59 59.0 59.0 61.0
PG 39 39.0 39.0 100.0
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 59% of the respondents falls under the category of
degree. 39% of respondents belongs to PG. Remaining 2% belongs to plus two.

10
8
 Experience
Valid Cumulative
Frequency Percent Percent Percent
Valid Less than 10 27 27.0 27.0 27.0
year
11year-20year 43 43.0 43.0 70.0
21year-30year 25 25.0 25.0 95.0
Above 30 year 5 5.0 5.0 100.0
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 43% of the respondents have an experience 11-20 year.
10
9
27% of the respondents have an experience of less than 10 years. Remaining 25% and 5%
belongs to 21-30 years and above 30 years respectively.

I believe that AI in our firm provide enhanced accuracy and precision in

decision making
Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 35 35.0 35.0 35.0
Agree
Agree 53 53.0 53.0 88.0
Neutral 8 8.0 8.0 96.0
Disagree 4 4.0 4.0 100.0
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 88% of the respondents are agreed their firm provide
enhanced accuracy and precision in decision making. 8% of the respondents are neutral in their
11
0
opinion and 4% are disagreed.

Our company having cost efficiency performance while using AI prediction

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 36 36.0 36.0 36.0
Agree 50 50.0 50.0 86.0
Neutral 9 9.0 9.0 95.0
Disagree 3 3.0 3.0 98.0
Strongly 2 2.0 2.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

11
1
From the above table it is revealed that 86% of the respondents are agreed their company having
cost efficiency performance while using AI prediction .9% of the respondents are neutral in their
opinion and 5% are disagreed.

I am well aware about the demand forecasting using AI is accurate and

very useful for the firm
Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 27 27.0 27.0 27.0
Agree
Agree 56 56.0 56.0 83.0
Neutral 13 13.0 13.0 96.0
Disagree 4 4.0 4.0 100.0
Total 100 100.0 100.0

11
2
INTERPRETATION

From the above table it is revealed that 83% of the respondents are agreed that they are aware
about the demand forecasting using AI is accurate and very useful for the firm.13%of the
respondents are neutral in their opinion and 4% are disagreed.

I believe that AI enables faster and more efficient decision making

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 21 21.0 21.0 21.0
Agree
Agree 57 57.0 57.0 78.0
Neutral 21 21.0 21.0 99.0
Disagree 1 1.0 1.0 100.0
Total 100 100.0 100.0

11
3
INTERPRETATION

From the above table it is revealed that 78% of the respondents are agreed they believe that AI
enables faster and more efficient decision making.21%of the respondents are neutral in their
opinion and 1% are disagreed.

I believe that AI reduce human error

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 45 45.0 45.0 45.0
Agree 41 41.0 41.0 86.0
Neutral 6 6.0 6.0 92.0
Disagree 6 6.0 6.0 98.0
Strongly 2 2.0 2.0 100.0
Disagree
Total 100 100.0 100.0

11
4
INTERPRETATION

From the above table it is revealed that 86% of the respondents are agreed they believe that AI
reduce human error. 6%of the respondents are neutral in their opinion and 8% are disagreed.

I am well aware about use of AI enhance productivity

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 37 37.0 37.0 37.0
Agree 50 50.0 50.0 87.0
Neutral 8 8.0 8.0 95.0
Disagree 3 3.0 3.0 98.0
11
5
 Strongly 2 2.0 2.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 87% of the respondents are agreed they are aware about
use of AI enhance productivity. 8%of the respondents are neutral in their opinion and 5% are
disagreed.

I believe that AI based predictions can develop more agile planning

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 29 29.0 29.0 29.0
Agree 47 47.0 47.0 76.0
Neutral 18 18.0 18.0 94.0
Disagree 5 5.0 5.0 99.0

11
6
 Strongly 1 1.0 1.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 76% of the respondents are agreed they believe that AI
based predictions can develop more agile planning.18%of the respondents are neutral in their
opinion and 6% are disagreed.

I am aware about AI prediction useful to adapt with market changes and other
economic disruptions
Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 28 28.0 28.0 28.0
Agree 56 56.0 56.0 84.0

11
7
 Neutral 12 12.0 12.0 96.0
Disagree 3 3.0 3.0 99.0
Strongly 1 1.0 1.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 84% of the respondents they are aware about AI
prediction useful to adapt with market changes and other economic disruptions agreed 12%of
the respondents are neutral in their opinion and 4% are disagreed.

Companies' efforts to reduce environmental and water pollution, greenhouse

gases, and threats posed by waste generated in the industrial environment
throughout the supply chain
Valid Cumulative
Frequency Percent Percent Percent

11
8
Valid Strongly Agree 37 37.0 37.0 37.0
Agree 45 45.0 45.0 82.0
Neutral 13 13.0 13.0 95.0
Disagree 4 4.0 4.0 99.0
Strongly 1 1.0 1.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 82% of the respondents that their efforts to reduce
environmental and water pollution, greenhouse gases, and threats posed by waste generated in
the industrial environment throughout the supply chain 13%of the respondents are neutral in
their opinion and 5% are disagreed.

I believe that use of AI predictions reduce human expertise and intuition

Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 35 35.0 35.0 35.0
Agree

11
9
 Agree 53 53.0 53.0 88.0
Neutral 8 8.0 8.0 96.0
Disagree 4 4.0 4.0 100.0
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 88% of the respondents believe that use of AI
predictions reduce human expertise and intuition 8%of the respondents are neutral in their
opinion and 4% are disagreed.

I’m aware that improper data lead to have wrong predictions using AI

12
0
 Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly Agree 36 36.0 36.0 36.0
Agree 50 50.0 50.0 86.0
Neutral 9 9.0 9.0 95.0
Disagree 3 3.0 3.0 98.0
Strongly 2 2.0 2.0 100.0
Disagree
Total 100 100.0 100.0

INTERPRETATION

From the above table it is revealed that 86% of respondents are aware that improper data lead
to have wrong predictions using AI.9% neutral in their opinion and 5% are disagreed.

12
1
 I am aware that the implementation cost of AI prediction is high
Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 27 27.0 27.0 27.0
Agree
Agree 56 56.0 56.0 83.0
Neutral 13 13.0 13.0 96.0
Disagree 4 4.0 4.0 100.0
Total 100 100.0 100.0

56
60

50

40
27 Series1
30
Series2
20 13

10 4

0
Strongly Agree Agree Neutral Disagree

INTERPRETATION

From the above table it is revealed that 83% of respondents are aware that implementation cost
of AI is high.13% neutral in their opinion and 4% are disagreed

12
2
 I believe that AI prediction is a barrier for smaller organisation
Valid Cumulative
Frequency Percent Percent Percent
Valid Strongly 21 21.0 21.0 21.0
Agree
Agree 57 57.0 57.0 78.0
Neutral 21 21.0 21.0 99.0
Disagree 1 1.0 1.0 100.0
Total 100 100.0 100.0

57
60

50

40
Series1
30 21 21
Series2
20

10 1

0
Strongly Agree Agree Neutral Disagree

INTERPRETATION

From the above table it is revealed that 78% of respondents are agreed that AI is a barrier for
smaller organisations.21% neutral in their opinion and 1% are disagreed

12
3
 Conclusion

Our purpose for the thesis was to contribute with guidance regarding the future of AI. We
aspired to provide guidance and insights for the future of AI by examining its implementation
and development within the complex environment of SCM. AI is expected to offer solutions
to significant challenges and drive innovation, but its implementation presents its own set of
challenges for organisations. Furthermore, since accurate forecasting regarding AI can be
highly beneficial in navigating this transformative technology landscape, we have intended to
provide beneficial insights to establish accurate forecasting. Through our grounded theory
analysis conducted on the data collected from qualitative interviews, we have detected
findings that serves as the basis for answering the research question of this thesis, being:

- How can implementation of Artificial Intelligence improve Supply Chain

Management?

Our study and the existing literature highlight the potential improvements of implementing AI
into supply chain management (SCM) within the manufacturing industry. The study and
existing literature support the idea that implementing AI can lead to significant improvements
in various aspects of SCM. One key advantage of AI in SCM is its ability to automate
production processes. By leveraging AI technologies, manufacturers can enhance the
accuracy of their forecasts, enabling them to better anticipate customer demands and adjust
their production accordingly. This helps in avoiding overproduction or underproduction,
leading to optimized inventory levels and reduced costs.

Moreover, AI can refine and optimize existing processes within the supply chain. By
analysing vast amounts of data, AI algorithms can identify inefficiencies and areas for
improvement. This allows companies to streamline their operations, minimize waste, and
enhance overall productivity. Financial savings are another notable benefit of implementing
AI in SCM. With AI's resource allocation and optimization capabilities, companies can
effectively manage their resources, such as raw materials, machinery, and labour. By
optimizing these resources, manufacturers can reduce costs and avoid unnecessary
expenditures. Furthermore, AI empowers decision-making within the supply chain. AI can
provide valuable insights and recommendations for strategic decision-making. This enables
companies to make informed choices regarding production planning, inventory management,
transportation logistics, and more. With improved decision-making, manufacturers can drive
prosperity and achieve significant cost savings. In summary, the integration of AI in SCM
within the manufacturing industry can automate production processes, increase forecast
accuracy, refine existing processes, conserve financial resources, and enhance decision-
making capabilities. These advantages contribute to increased efficiency, reduced costs, and
overall competitiveness in the market.

However, what our study further explains is the several challenges and considerations that
need to be addressed for successful implementation. Firstly, feedback and evaluation are
crucial

12
4
within AI systems to improve SCM. While previous studies have overlooked this aspect, our
study conclude the need for AI to generate feedback and evaluation, rather than just receiving
it. This feedback loop ensures continuous improvement and enhances SCM effectiveness. In
addition, effective resource management and cost-saving measures are crucial for successful
AI implementation. While AI can optimise production, streamline SC operations, and
improve decision-making, the high costs, complexity, and time requirements associated with
implementation pose challenges. The complexity associated with applying AI in SCM can
vary depending on the context and perspectives of practitioners. The unique complexities
encountered by individuals within SCM, such as managing multiple stakeholders and
navigating uncertainties, need to be considered to gain a comprehensive understanding of the
challenges involved. In addition, companies need to carefully evaluate the expected benefits
of AI and consider whether they outweigh the initial investment and time commitments. Even
though effective inventory management through AI can lead to cost reduction and minimise
waste. AI algorithms can analyse inventory levels and accurately forecast demand, helping to
achieve the optimal balance between supply and demand. However, challenges related to data
quality should be addressed to ensure the effectiveness of AI in inventory management.

AI can contribute to achieving an even flow in the supply chain by streamlining processes,
reducing inspection time, and improving defect detection accuracy. However, initial data
quality and the integration of AI with existing processes and systems can pose difficulties.
Data quality is a significant factor in AI performance. Poor data quality can lead to errors and
disruptions in SCM. While AI has the potential to analyse sensor data and anticipate
maintenance requirements, challenges related to data quality and the need for manual input
should be addressed for effective implementation.

Moreover, resistance to AI implementation exists, but it may not necessarily stem from
employees' resistance to change. Instead, the resistance lies in the difficulty of transforming
previous processes and systems into AI-compatible ones. Transparency and explainability of
AI systems are crucial in building trust and addressing suspicions. Trust in AI systems
requires compatibility with existing systems, addressing concerns related to data security, and
effective communication of relevant information. Over-reliance on AI and complacency
should be avoided, and human oversight should be maintained.

In summary, while AI holds promise in improving SCM within the manufacturing industry,
careful consideration of feedback and evaluation, cost-benefit analysis, data quality,
complexity, resistance, and trust is necessary for successful implementation and reaping the
full benefits of AI.

12
5
FINDINGS
1. 79% of the respondents are male and 21% are female
2. 59% of the respondents falls under the age group of 30.
3. 59% of the respondents falls under the category of degree.
4. 43% of the respondents have an experience 11-20 year.
5. 88% of the respondents are agreed their firm provide enhanced accuracy and precision in
decision making.
6. 86% of the respondents are agreed their company having cost efficiency performance
while using AI prediction.
7. 83% of the respondents are agreed that they are aware about the demand forecasting
using AI is accurate and very useful for the firm.
8. 86% of the respondents are agreed they believe that AI reduce human error.
9. 87% of the respondents are agreed they are aware about use of AI enhance productivity.
10. 76% of the respondents are agreed they believe that AI based predictions can develop
more agile planning.
11. 84% of the respondents they are aware about AI prediction useful to adapt with market
changes and other economic disruptions agreed.
12. 82% of the respondents that their efforts to reduce environmental and water pollution,
greenhouse gases, and threats posed by waste generated in the industrial environment
throughout the supply chain.
13. 88% of the respondents believe that use of AI predictions reduce human expertise and
intuition.

12
6
12
7
 Practical recommendations

1. Based on our empirical findings, we present practical recommendations derived from this
study to provide valuable insights for organizations currently involved or anticipating
involvement in the implementation process of AI in the context of SCM. These
recommendations may also be relevant for other industries than just the manufacturing
industry.

2. Our first recommendation is to embrace change management and to make a cultural

adoption. In the context of implementing AI in SCM, embracing change management and
fostering cultural adoption are crucial steps for success. Change management is essential
to ensure a smooth transition and acceptance of new processes. Some practical steps to
embrace change can be to educate stakeholders about the benefits of AI in SCM,
highlighting the potential improvements in efficiency, cost reduction, and decision-
making capabilities. Communicate the need for change and the rationale behind it.
Another step can be to create a comprehensive plan outlining the steps, timelines, and
resources required for AI implementation. Clearly define roles and responsibilities, and
establish communication channels to keep everyone informed.

3. throughout the process. Cultural adoption can be achieved through Fostering a

collaborative environment where teams from different departments work together to
integrate AI into SCM processes. Encourage knowledge sharing, open communication,
and cross-functional collaboration to maximise the potential of AI.

4. Our second recommendation is to overcome complexity through strategic planning. To

navigate the complexities of AI integration in SCM, organisations should invest in
strategic planning to ensure a smooth implementation process. By dedicating time and
financial resources to develop well-defined strategies, organisations can effectively
address the unique challenges associated with AI adoption in SCM. Organisations should
formulate comprehensive strategies that encompass all aspects of AI implementation in
SCM. These strategies should outline specific objectives, timelines, and resource
allocation, taking into account the complexities of SCM operations. The strategic planning
should also involve a thorough analysis of potential risks and challenges associated with
AI integration in SCM. Organisations should proactively identify and address these risks
through contingency plans and mitigation strategies. This may involve considering data
security, privacy concerns, system integration complexities, or regulatory compliance.

5. We strongly advise prioritising investments in data management as our third practical

recommendation. The significance of high-quality data cannot be overstated when it
comes to effectively implementing AI in SCM. To ensure the accuracy, completeness, and
consistency of data, organisations should establish robust data governance practices. This
entails implementing processes for data cleansing, employing data validation techniques,
and formulating data integration strategies. Furthermore, allocating resources towards
advanced technologies and infrastructure that facilitate streamlined data collection,

12
8
 storage, and retrieval is crucial.

6. Our final recommendation is to establish clear performance metrics in order to evaluate

the impact of AI implementation in SCM, organisations should define clear performance
metrics aligned with their strategic objectives. These metrics may include cost savings,
productivity improvements, customer satisfaction, and supply chain responsiveness. By
regularly monitoring and analysing these metrics, organisations can assess the
effectiveness of AI integration and identify areas for further improvement. Furthermore, it
is crucial for organisations to establish a feedback loop that facilitates continuous
improvement. By regularly monitoring and analysing the defined performance metrics,
organisations can gather valuable insights into the effectiveness of AI integration in SCM.
This ongoing evaluation allows them to identify strengths, weaknesses, and areas for
further improvement. It also enables them to adapt their strategies and optimise AI
systems to maximise benefits and align with their evolving strategic objectives. Moreover,
organisations should foster a culture of data-driven decision-making, where insights
derived from performance metrics are used to inform future actions and drive continuous
advancements in their AI-driven supply chain processes.

7. In conclusion, this study provides practical recommendations for the successful

implementation of AI in SCM. These recommendations are derived from empirical
findings and aim to offer valuable insights to individuals involved or anticipating
involvement in the AI implementation process in SCM, regardless of the industry.

Limitations and future research

Despite making a valuable contribution to the field of the implementation of AI in order to

improve SCM, it is important to acknowledge the limitations of this thesis. First of all, a
significant limitation arises from the time constraints imposed on the study. With only a brief
period of approximately one and a half months for conducting the research, certain chapters
had to be prioritised over others. Consequently, some areas of the thesis could have been
explored in greater depth, and the analysis may not have been as comprehensive as desired.
Given more time, a larger volume of data could have been collected, allowing for a more
robust and extensive analysis. Additionally, the sample size of six respondents may be
considered relatively small. However, this choice was deliberate, as it allowed for a more in-
depth exploration of each participant's perspective and facilitated a comprehensive
understanding of their viewpoints. Drawing meaningful conclusions from the available data
was not compromised by the sample size.

Furthermore, it is essential to note that the thesis focused solely on the implementation phase
of the process, neglecting the maintenance and reconstruction aspects. Consequently, to gain a
holistic understanding of the respondents' perception of the AI process as a whole, it would be
valuable to investigate and incorporate insights from these other phases. By assembling
knowledge from all stages of the process, the industry can enhance their understanding and
make informed improvements in their own AI-system. A deeper understanding of these
12
9
factors can help guide organisations in making informed decisions about AI implementation
and maximise its potential benefits.

Furthermore, the discussion acknowledges the potential inconsistency between the literature
and the findings of the study. However, it is important to recognize that the realm of AI and
its integration into supply chain management is an ever-evolving field. With advancements in
AI technologies and increasing organisational expertise in implementation, it is plausible that
future studies will offer more definitive proof concerning the precise advantages and ROI
associated with AI in SCM. In essence, the present passage effectively emphasises the
prevailing disparity and raises significant points to ponder concerning the implementation and
assessment of AI in SCM.

13
0
9.0 Reference list

Aityassine, F., Aldiabat, B., Al-rjoub, S., Aldaihani, F., Al-Shorman, H. and Al-Hawary, S.
(2021). The mediating effect of just in time on the relationship between green supply chain
management practices and performance in the manufacturing companies. Uncertain Supply
Chain Management, 9(4), pp.1081–1090.

Aldiabat, K. and Le Navenec, C.-L. (2018). Data saturation: The mysterious step in grounded
theory method. The Qualitative Report, 23(1). doi:https://doi.org/10.46743/2160-
3715/2018.2994.

Alharahsheh, H.H. and Pius, A. (2020). A Review of key paradigms: positivism VS

interpretivism. Global Academic Journal of Humanities and Social Sciences, 2(3), pp.39–42.
doi:https://doi.org/10.36348/gajhss.2020.v02i03.001.

Aliyu, A.A., Singhry, I.M., Adamu, H. and Abubakar, M.A.M. (2015). Ontology,
epistemology and axiology in quantitative and qualitative research: Elucidation of the
research philophical misconception. Proceedings of the Academic Conference: Mediterranean
Publications & Research International on New Direction and Uncommon, 2(1), pp.1–26.

Amin, M.E.K., Nørgaard, L.S., Cavaco, A.M., Witry, M.J., Hillman, L., Cernasev, A. and
Desselle, S.P. (2020). Establishing trustworthiness and authenticity in qualitative pharmacy
research. Research in Social and Administrative Pharmacy, 16(10), pp.1472–1482.
doi:https://doi.org/10.1016/j.sapharm.2020.02.005.

Armelius, H. (2022). Industriproduktionens sammansättning. [online] Ekonomifakta.

Available at: https://www.ekonomifakta.se/fakta/ekonomi/produktion-och-
investeringar/industriproduktionens-sammansattning/ [Accessed 4 Apr. 2023].

13
1
Azzam, M. and Beckmann, R. (2022). How AI Helps to Increase Organizations’ Capacity to
Manage Complexity – A Research Perspective and Solution Approach Bridging Different
Disciplines. IEEE Transactions on Engineering Management, pp.1–14.
doi:https://doi.org/10.1109/tem.2022.3179107.

Bamberger, P. (2018). Clarifying what we are about and where we are going. Academy of
Management Discoveries, 4(1), pp.1–10.

Basch, J.M., Melchers, K.G., Kurz, A., Krieger, M. and Miller, L. (2020). It Takes More Than
a Good Camera: Which Factors Contribute to Differences Between Face-to-Face Interviews
and Videoconference Interviews Regarding Performance Ratings and Interviewee
Perceptions? Journal of Business and Psychology, 36(5), pp.921–940.
doi:https://doi.org/10.1007/s10869- 020-09714-3.

Bearman, M. (2019). Focus on Methodology: Eliciting Rich data: a Practical Approach to

Writing semi-structured Interview Schedules. Focus on Health Professional Education: A
Multi-Professional Journal, 20(3), pp.1–11. doi:https://doi.org/10.11157/fohpe.v20i3.387.

Bell, E. and Bryman, A. (2007). The Ethics of Management Research: An Exploratory

Content Analysis. British Journal of Management, 18(1), pp.63–77.
doi:https://doi.org/10.1111/j.1467- 8551.2006.00487.x.

Benoot, C., Hannes, K. and Bilsen, J. (2016). The use of purposeful sampling in a qualitative
evidence synthesis: A worked example on sexual adjustment to a cancer trajectory. BMC
Medical Research Methodology, 16(1), pp.1–12. doi:https://doi.org/10.1186/s12874-016-
0114- 6.

Blouin, D., Day, A.G. and Pavlov, A. (2011). Comparative Reliability of Structured Versus
Unstructured Interviews in the Admission Process of a Residency Program. Journal of
Graduate Medical Education, [online] 3(4), pp.517–523. doi:https://doi.org/10.4300/jgme-d-
10-00248.1.

Boucher, P. (2020). STUDY Panel for the Future of Science and Technology EPRS |
European Parliamentary Research Service. [online]
Available at:
https://www.europarl.europa.eu/RegData/etudes/STUD/2020/641547/EPRS_STU(2020)6415
47_EN.pdf.

Boussehaib, F.E. and Belhsen, N. (2023). Information sharing in SCM a meta analysis. Actes
du Colloque International Coopération Université-Entreprise: d’une recherche ‘sur’
l’entreprise à une recherche ‘avec’ l’entreprise, pp.549–557.

Brand, V. (2008). Empirical Business Ethics Research and Paradigm Analysis. Journal of
Business Ethics, 86(4), pp.429–449. doi:https://doi.org/10.1007/s10551-008-9856-3.

13
2
Brown, S. and Wilson, D. (2019). Industry 4.0 and the Human factor: an Overview and
Emerging Research agenda. International Journal of Management Reviews, 21(4), pp.411–
430.

Bryman, A. (1984). The Debate about Quantitative and Qualitative Research: a Question of
Method or Epistemology? The British Journal of Sociology, [online] 35(1), pp.75–92.
Available at: https://www.jstor.org/stable/pdf/590553.pdf [Accessed 2 May 2023].

13
3
Brynjolfsson, E. and McAfee, A. (2017). Machine, Platform, Crowd: Harnessing Our Digital
Future. W. W. Norton & Company.

Brynjolfsson, E., Mitchell, T. and Rock, D. (2018). What Can Machines Learn, and What
Does It Mean for Occupations and the Economy? AEA Papers and Proceedings, 108, pp.43–
47. doi:https://doi.org/10.1257/pandp.20181019.

Bughin, J., Hazan, E., Ramaswamy, S., Chui, M., Allas, T., Dahlström, P., Henke, N. and
Trench, M. (2017). Artificial Intelligence: the next Digital Frontier? McKinsey Global
Institute.

Cai, R., Zhang, M., XU, W. and Sun, Y. (2021). A predictive maintenance inventory model
considering deterioration forecasting accuracy. International Journal of Production Research,
59(4), pp.1117–1136.

Cheatham, B., Javanmardian, K. and Samandari, H. (2019). Confronting the risks of artificial
intelligence. [online] McKinsey Quarterly. Available at: https://www.cognitivescale.com/wp-
content/uploads/2019/06/Confronting_AI_risks_-_McKinsey.pdf.

Clancy, M.J. (2002). Overview of research designs. Emergency Medicine Journal, 19(6),
pp.546–549. doi:https://doi.org/10.1136/emj.19.6.546.

Clarke-Potter, K. (2019). Why Inefficient Supply Chains Are Affecting Your Customers
(And How You Can Fix It). [online] Blockhead Technologies.
Available at: https://blockheadtechnologies.com/why-inefficient-supply-chains-
are-affecting-your- customers/ [Accessed 3 Apr. 2023].

Colbert, A., Yee, N. and George, G. (2016). The Digital Workforce and the Workplace of the
Future. Academy of Management Journal, 59(3), pp.731–739.
doi:https://doi.org/10.5465/amj.2016.4003.

Collis, J. and Hussey, R. (2014). Business Research : a Practical Guide for Undergraduate and
Postgraduate Students. 4th ed. Basingstoke: Hampshire.

Cope, D.G. (2014). Methods and Meanings: Credibility and Trustworthiness of Qualitative
Research. Oncology Nursing Forum, 41(1), pp.89–91.

Coyle, J.J., Langley, C.J., Novack, R.A. and Gibson, B.J. (2017). Supply chain management a
logistics perspective. 10th ed. Boston, Mass. Cengage Learning.

Dash, R., McMurtrey, M., Rebman, C. and Kar, U.K. (2019). Application of Artificial
Intelligence in Automation of Supply Chain Management. Journal of Strategic Innovation and
Sustainability, 14(3). doi:https://doi.org/10.33423/jsis.v14i3.2105.

Delshad, S. and Safaei, A.S. (2021). A Review of Artificial Intelligence Applications in the

13
4
Manufacturing industry: Risk Analysis and recommendations. Journal of Manufacturing
Systems, 61, pp.123–134.

Dodgson, M., Gann, D., Wladawsky-Berger, I., Sultan, N. and George, G. (2015). Managing
Digital Money. Academy of Management Journal, 58(2), pp.325–333.
doi:https://doi.org/10.5465/amj.2015.4002.

13
5
Dogan, A. and Birant, D. (2021). Machine learning and data mining in manufacturing. Expert
Systems with Applications, 166, pp.1–22. doi:https://doi.org/10.1016/j.eswa.2020.114060.

Eisenhardt, K.M. (1989). Building Theories from Case Study Research. The Academy of
Management Review, 14(4), pp.532–550. doi:https://doi.org/10.2307/258557.

Elo, S. and Kyngäs, H. (2008). The qualitative content analysis process. Journal of Advanced
Nursing, [online] 62(1), pp.107–115. doi:https://doi.org/10.1111/j.1365-2648.2007.04569.x.

Elomri, A., Elhuni, A. and Abid, M. (2020). The use of artificial intelligence in supply chain
management: Risks and opportunities. Journal of Industrial Engineering and Management,
13(1), pp.12–36.

Fairris, D. and Tohyama, H. (2002). Productive efficiency and the lean production system in
Japan and the United States . Economic and Industrial Democracy, 23(4), pp.529–554.

Frank, M.R., Wang, D., Cebrian, M. and Rahwan, I. (2019). The Evolution of Citation Graphs
in Artificial Intelligence Research. Nature Macine Intelligence, 1, pp.79–85.

Gao, X., Xu, J., Wang, J., Yang, Y. and He, Y. (2020). An intelligent quality control system
for printed circuit board assembly. Journal of Manufacturing Systems, 57, pp.199–209.

Ghahremani Nahr, J., Nozari, H. and Ebrahim Sadeghi, M. (2021). Artificial intelligence and
Machine Learning for Real-world problems (A survey). International journal of innovation in
engineering, 1(3), pp.38–47.

Gharehgozli, A.H., Seyed Esfahani, S.S., Zandieh, M.A., Ramezani, A., h, j and h, j (2021).
Artificial intelligence in supply chain management: A comprehensive literature review.
International Journal of Production Research, pp.1–29.

Gill, P., Stewart, K., Treasure, E. and Chadwick, B. (2008). Methods of data collection in
qualitative research: Interviews and focus groups. British Dental Journal, [online] 204(6),
pp.291–295. doi:https://doi.org/10.1038/bdj.2008.192.

Gill, S.L. (2020). Qualitative sampling methods. Journal of Human Lactation, 36(4), pp.579–
581. doi:https://doi.org/10.1177/0890334420949218.

Glikson, E. and Woolley, A.W. (2020). Human Trust in Artificial Intelligence: Review of
Empirical Research. Academy of Management Annals, 14(2).
doi:https://doi.org/10.5465/annals.2018.0057.

Goralski, M.A. and Keong Tan, T. (2020). Artificial Intelligence and Sustainable
Development.
The International Journal of Management Education, 18(1), pp.1–9.

13
6
Grace, K., Salvatier, J., Dafoe, A., Zhang, B. and Evans, O. (2018). Viewpoint: When Will AI
Exceed Human Performance? Evidence from AI Experts. Journal of Artificial Intelligence
Research, [online] 62(62), pp.729–754. doi:https://doi.org/10.1613/jair.1.11222.

Grodal, S., Anteby, M. and Holm, A.L. (2020). Achieving Rigor in Qualitative Analysis: The
Role of Active Categorization in Theory Building. Academy of Management Review, 46(3).
doi:https://doi.org/10.5465/amr.2018.0482.

13
7
Gunasekaran, A., Papadopoulos, T., Dubey, R., Wamba, S.F., Childe, S.J., Hazen, B. and
Akter,
S. (2019). Blockchain and Big Data analytics: Opportunities and challenges. International
Journal of Production Research., 57(7), pp.1665–1679.

Guo, S., Wu, D., Zhou, C., Hu, Y. and Ding, Y. (2020). A transparency-aware framework for
AI-assisted manufacturing. IEEE Access, 8, pp.235149–235161.

Hagaman, A.K. and Wutich, A. (2016). How Many Interviews Are Enough to Identify
Metathemes in Multisited and Cross-cultural Research? Another Perspective on Guest, Bunce,
and Johnson’s (2006) Landmark Study. Field Methods, 29(1), pp.23–41.
doi:https://doi.org/10.1177/1525822x16640447.

Harari, Y.N. (2017). Homo Deus : a brief history of tomorrow. New York, NY: Harper, an
imprint of HarperCollins Publishers.

Harvey, W.S. (2011). Strategies for conducting elite interviews. Qualitative Research, 11(4),
pp.431–441.

Hasija, A. and Esper, T.L. (2022). In artificial intelligence (AI) we trust: A qualitative
investigation of AI technology acceptance. Journal of Business Logistics, 2022(43), pp.388–
412. doi:https://doi.org/10.1111/jbl.12301.

Heikkilä, J. (2002). From supply to demand chain management: efficiency and customer
satisfaction. Journal of Operations Management, [online] 20(6), pp.747–767.
doi:https://doi.org/10.1016/s0272-6963(02)00038-4.

Helm, J.M., Swiergosz, A.M., Haeberle, H.S., Karnuta, J.M., Schaffer, J.L., Krebs, V.E.,
Spitzer, A.I. and Ramkumar, P.N. (2020). Machine learning and AI: Definitions,
Applications, and Future Directions. Current reviews in musculoskeletal medicine., 13(1),
pp.69–76. doi:https://doi.org/10.1007/s12178-020-09600-8.

Hendricks, K.B. and Singhal, V.R. (2009). Demand-Supply Mismatches and Stock Market
Reaction: Evidence from Excess Inventory Announcements. Manufacturing & Service
Operations Management, 11(3), pp.509–524. doi:https://doi.org/10.1287/msom.1080.0237.

Hoon Kwak, Y. and Dixon, C.K. (2008). Risk management framework for pharmaceutical
research and development projects. International Journal of Managing Projects in Business,
1(4), pp.552–565. doi:https://doi.org/10.1108/17538370810906255.

Howard, J. (2019). Artificial intelligence: Implications for the future of work. American
Journal of Industrial Medicine, [online] 62(11), pp.917–926.
doi:https://doi.org/10.1002/ajim.23037.

Howard-Grenville, J., Nelson, A., Vough, H. and Zilber, T.B. (2021). From the Editors—
13
8
Achieving Fit and Avoiding Misfit in Qualitative Research. Academy of Management
Journal, 64(5), pp.1313–1323. doi:https://doi.org/10.5465/amj.2021.4005.

Huang, J., Cao, L. and Wan, X. (2018). Supply Chain Management in the Era of Artificial
Intelligence. IEEE Intelligent Systems, 33(5), pp.2–3.

Huang, L., Zhang, Y. and WU, Q. (2020). An artificial intelligence-based safety management
system for manufacturing plants. Journal of Manufacturing Systems, 56, pp.73–84.

13
9
Hugos, M.H. (2018). Essentials of supply chain management. 4th ed. Hoboken, New Jersey:
John Wiley & Sons, Inc.

Hult, G.T.M., Ketchen, D.J. and Arrfelt, M. (2007). Strategic supply chain management:
Improving performance through a culture of competitiveness and knowledge development.
Strategic Management Journal, 28(10), pp.1035–1052. doi:https://doi.org/10.1002/smj.627.

Jäger, N.W., Loschelder, D.D., Dettmers, L.C., Topolinski, S., Pfeiffer, T. and Botschen, G.
(2021). Rethinking the Role of Creativity in the Age of Artificial intelligence. Frontiers in
Psychology, 12(667986).

Jobin, A., Ienca, M. and Vayena, E. (2019). The global landscape of AI ethics guidelines.
Nature Machine Intelligence, [online] 1(9), pp.389–399. doi:https://doi.org/10.1038/s42256-
019-0088-2.

Jordan, M.I. and Mitchell, T.M. (2015). Machine learning: Trends, perspectives, and prospects.
Science, 349(6245), pp.255–260. doi:https://doi.org/10.1126/science.aaa8415.

Kallio, H., Pietilä, A.-M., Johnson, M. and Kangasniemi, M. (2016). Systematic

Methodological review: Developing a Framework for a Qualitative semi-structured Interview
Guide. Journal of Advanced Nursing, [online] 72(12), pp.2954–2965.
doi:https://doi.org/10.1111/jan.13031.

Kang, D., Kim, D., Kim, D. and Hong, S. (2020). Transparency and Explainability in AI-
based systems: a review. Journal of Intelligent Manufacturing, 31, pp.1147–1161.

Katoch, S., Chauhan, S.S. and Kumar, V. (2020). A review on genetic algorithm: past,
present, and future. Multimedia Tools and Applications, 80, pp.8091–8026.
doi:https://doi.org/10.1007/s11042-020-10139-6.

Kim, S.W., Kong, J.H., Lee, S.W. and Lee, S. (2021). Recent Advances of Artificial
Intelligence in Manufacturing Industrial Sectors: A Review. International Journal of Precision
Engineering and Manufacturing, 23(1), pp.111–129. doi:https://doi.org/10.1007/s12541-021-
00600-3.

Knox Lovell, C.A. (1993). Production Frontiers and Productive Efficiency. Oxford University
Press, pp.3–67.

Kohli, R., Melville, N. and Saha, B. (2021a). Augmenting Supply Chain Management with
Artificial Intelligence. Journal of Management Information Systems, 38(1), pp.1–8.

Kohli, R., Sharma, S. and Gupta, V. (2021b). Enhancing supply chain visibility and resilience
through artificial intelligence: A review and research agenda. International Journal of
Production Research, 59(6), pp.1683–1703.
doi:https://doi.org/10.1080/00207543.2020.1848596.
14
0
Korstjens, I. and Moser, A. (2018). Series: Practical Guidance to Qualitative research. Part 4:
Trustworthiness and Publishing. European Journal of General Practice, [online] 24(1),
pp.120–124. doi:https://doi.org/10.1080/13814788.2017.1375092.

Krouwel, M., Jolly, K. and Greenfield, S. (2019). Comparing Skype (video calling) and in-
person qualitative interview modes in a study of people with irritable bowel syndrome – an

14
1
exploratory comparative analysis. BMC Medical Research Methodology, 19(1).
doi:https://doi.org/10.1186/s12874-019-0867-9.

Kumar Jena, S. (2021). (PDF) Academic Journal Guide 2021 (It was earlier ABS 2021)
Ranking list. [online] ResearchGate. Available at:
https://www.researchgate.net/publication/352767191_Academic_Journal_Guide_2021_It_wa
s_earlier_ABS_2021_Ranking_list.

Kumar, S., Yadav, R. and Kumar, R. (2019a). Artificial Intelligence in Manufacturing: A

Review. Journal of Manufacturing Systems, 49, pp.1–14.

Kumar, V., Chai, K.H. and Kurnia, S. (2019b). An Empirical Examination of Transparency in
Artificial intelligence-enabled Supply Chain Decision making. Journal of Business Research,
101, pp.725–735.

Kwak, D.W., Lee, K.M., Huh, J.H. and Kim, C. (2019). Development of an Artificial
intelligence-based New Product Development System in the Electronics industry.
International Journal of Production Research, 57(3), pp.945–960.

Lambora, A., Gupta, K. and Chopra, K. (2019). Genetic Algorithm- A Literature Review.
Proceedings of the International Conference on Machine Learning, Big Data, Cloud and
Parallel Computing: Trends, Prespectives and Prospects, COMITCon 2019, pp.380–384.

Larson, E.W. and Gray, C.F. (2021). Project management: The managerial process. 8th ed.
New York, N.Y.: Mcgraw-Hill Education.

Larsson, S. and Heintz, F. (2020). Transparency in artificial intelligence. Internet Policy

Review, 9(2), pp.1–16. doi:https://doi.org/10.14763/2020.2.1469.

Laverty, S.M. (2003). Hermeneutic Phenomenology and Phenomenology: a Comparison of

Historical and Methodological Considerations. International Journal of Qualitative Methods,
2(3), pp.21–35. doi:https://doi.org/10.1177/160940690300200303.

Leo Kumar, S.P. (2018). Knowledge-based expert system in manufacturing planning: state-
of- the-art review. International Journal of Production Research, [online] 57(15-16), pp.4766–
4790. doi:https://doi.org/10.1080/00207543.2018.1424372.

Li, X., Zhang, L., Xu, X., Wu, X. and Li, Y. (2020). Artificial Intelligence and
Manufacturing: A Literature Review. Journal of Manufacturing Systems, 56, pp.258–276.

Lima, R., Da Cruz, A.M.R. and Ribeiro , J. (2020). Artificial Intelligence Applied to Software
Testing: A Literature Review. 15th Iberian Conference on Information Systems and
Technologies (CISTI) , pp.1–6. doi:https://doi.org/10.23919/CISTI49556.2020.9141124.

Lin, C.H., Wu, J.H. and Liao, Y.C. (2021). Predictive maintenance for semiconductor
14
2
manufacturing equipment using machine learning. Journal of Manufacturing Systems, 59,
pp.211–221.

Liu, Y., Tong, W., Li, Q., Yao, F., Li, Y., Li, H.X. and Huang, J. (2022). Study on
Complexity of Precast Concrete Components and Its Influence on Production Efficiency.
Advances in Civil Engineering, 2022, pp.1–18. doi:https://doi.org/10.1155/2022/9926547.

14
3
Lotfi, Z., Mukhtar, M., Sahran, S. and Zadeh, A.T. (2013). Information Sharing in Supply
Chain Management. Procedia Technology , 11(2013), pp.298–304.

Magnusson, E. and Marecek, J. (2015). Doing interview-based qualitative research: a learner’s

guide. 1:a uppl ed. Cambridge: Cambridge University Press.

Marsh, D. and Furlong, P. (2002). ”A skin, not a sweater: Ontology and epistemology in
political science”, in Marsh, D & Stoker, G. 2nd edition ed. Basingstoke: Palgrave
Macmillan, pp.7–41.

Melo, M.T., Nickel, S. and Saldanha-da-Gama, F. (2009). Facility location and supply chain
management – A review. European Journal of Operational Research, [online] 196(2), pp.401–
412. doi:https://doi.org/10.1016/j.ejor.2008.05.007.

Min, H. (2010). Artificial intelligence in supply chain management: theory and applications.
International Journal of Logistics Research and Applications, 13(1), pp.13–39.
doi:https://doi.org/10.1080/13675560902736537.

Mishra, S., Modi, S.B. and Animesh, A. (2013). The relationship between information
technology capability, inventory efficiency, and shareholder wealth: A firm-level empirical
analysis. Journal of Operations Management, 31(6), pp.298–312.
doi:https://doi.org/10.1016/j.jom.2013.07.006.

Morash, E.A. and Clinton, S.R. (1997). The Role of Transportation Capabilities in
International Supply Chain Management. Transportation Journal, 36(3), pp.5–17.

Mueller, A.E. and Segal, D.L. (2015). Structured versus Semistructured versus Unstructured
Interviews. The Encyclopedia of Clinical Psychology, pp.1–7.
doi:https://doi.org/10.1002/9781118625392.wbecp069.

Nagurney, A., Zhang, D. and Zhang, D. (2021). Artificial intelligence and supply chain
management: A comprehensive literature review and research agenda. Journal of Business
Research. Journal of Business Research, 130, pp.392–402.

Pannu, A. (2015). Artificial Intelligence and Its Application in Different Areas. International
Journal of Engineering and Innovative Technology (IJEIT), 4(10), pp.79–84.

Papageorgiou, L.G. (2009). Supply chain optimisation for the process industries: Advances
and opportunities. Computers & Chemical Engineering, 33(12), pp.1931–1938.
doi:https://doi.org/10.1016/j.compchemeng.2009.06.014.

Park, K. and Okudan Kremer, G.E. (2015). Assessment of static complexity in design and
manufacturing of a product family and its impact on manufacturing performance.
International Journal of Production Economics, 169,
pp.215–232. doi:https://doi.org/10.1016/j.ijpe.2015.07.036.

14
4
Pech, M., Vrchota, J. and Bednář, J. (2021). Predictive Maintenance and Intelligent Sensors in
Smart Factory: Review. Sensors, 21(4), p.1470. doi:https://doi.org/10.3390/s21041470.

Phillips, C.B., Dwan, K., Hepworth, J., Pearce, C. and Hall, S. (2014). Using qualitative
mixed methods to study small health care organizations while maximising trustworthiness
and

14
5
authenticity. BMC Health Services Research, 14(1), pp.1–13.
doi:https://doi.org/10.1186/s12913-014-0559-4.

Piplani, R., Gunasekaran, A. and Tiwari, M.K. (2021). Supply chain network design using a
hybrid approach of artificial bee colony and k-means clustering algorithm. International
Journal of Production Research, 59(2), pp.379–399.

Pournader, M., Ghaderi, H., Hassanzadegan, A. and Fahimnia, B. (2021). Artificial

intelligence applications in supply chain management. International Journal of Production
Economics, [online] 241(2021), p.108250. doi:https://doi.org/10.1016/j.ijpe.2021.108250.

Ratner, B. (2000). A comparison of two popular machine learning methods. DM STAT-1 –

Online Newsletter about Quantitative Methods in Direct Marketing, 4.

Reim, W., Åström, J. and Eriksson, O. (2020). Implementation of Artificial Intelligence (AI):
A Roadmap for Business Model Innovation. AI, 1(2), pp.180–191.
doi:https://doi.org/10.3390/ai1020011.

Rimol, M. (2022). Gartner Survey Reveals 80% of Executives Think Automation Can
Be Applied to Any Business Decision. [online] Gartner. Available at:
https://www.gartner.com/en/newsroom/press-releases/2022-08-22-gartner-survey-reveals-80-
percent-of-executives-think-automation-can-be-applied-to-any-business-decision.

Rothe, R. (2017). Applying deep learning to real-world problems. [online] Medium. Available
at: https://medium.com/merantix/applying-deep-learning-to-real-world-problems-
ba2d86ac5837 [Accessed 11 Apr. 2023].

Ryan, T. (2011). The Teacher/ Researcher and the Role of Pre-understanding: A Personal
Analysis. New Zealand Journal of Teachers’ Work, 8(2), pp.220–228.

Saldaña, J. (2013). The Coding Manual for Qualitative Researchers. [online] Available at:
http://emotrab.ufba.br/wp-content/uploads/2020/09/Saldana-2013-
TheCodingManualforQualitativeResearchers.pdf.

Sandelowski, M. and Barroso, J. (2003). Writing the Proposal for a Qualitative Research
Methodology Project. Qualitative Health Research, 13(6), pp.781–820.
doi:https://doi.org/10.1177/1049732303013006003.

Saunders, M., Lewis, P. and Thornhill, A. (2019). Research Methods for Business Students.
8th ed. New York: Pearson.

Schaeffer, E., Cabanes, J. and Gupta, A. (2018). Manufacturing the future: Artificial
intelligence will fuel the next wave of growth for industrial equipment companies. [online]
Accenture, pp.1–20. Available at: https://www.accenture.com/_acnmedia/pdf-74/accenture-
pov-manufacturing-digital-final.pdf [Accessed 28 Mar. 2023].
14
6
Schank, R.C. (1987). What Is AI, Anyway? AI Magazine, [online] 8(4), pp.59–65.
doi:https://doi.org/10.1609/aimag.v8i4.623.

Scherer, M.U. (2015). Regulating Artificial Intelligence Systems: Risks, Challenges,

Competencies, and Strategies. SSRN Electronic Journal, 29(2), pp.353-.
doi:https://doi.org/10.2139/ssrn.2609777.

14
7
Schmidt, T., Moeller, K., Krems, J.F. and Sichtmann, J. (2021). Cost-effectiveness of
artificial intelligence in supply chain management: Insights for small- and medium-sized
enterprises. Journal of Business Research, 132, pp.165–174.

Sharma, M., Luthra, S., Joshi, S. and Kumar, A. (2021). Implementing challenges of artificial
intelligence: Evidence from public manufacturing sector of an emerging economy.
Government Information Quarterly, p.101624. doi:https://doi.org/10.1016/j.giq.2021.101624.

Shepherd, D.A. and Sutcliffe, K.M. (2011). INDUCTIVE TOP-DOWN THEORIZING: A

SOURCE OF NEW THEORIES OF ORGANIZATION. Academy of Management Review,
36(2), pp.361–380. doi:https://doi.org/10.5465/amr.2011.59330952.

Shmueli, G. (2010). To Explain or to Predict? Statistical Science, [online] 25(3), pp.289–310.

doi:https://doi.org/10.1214/10-sts330.

Shubhendu, S. and Vijay, J. (2013). Applicability of Artificial Intelligence in Different Fields

of Life. International Journal of Scientific Engineering and Research (IJSER), 1(1), pp.28–35.

Sithole, B., Silva, S.G. and Kavelj, M. (2016). Supply Chain Optimization: Enhancing End-
to- End Visibility. Procedia Engineering, [online] 159(2016), pp.12–18.
doi:https://doi.org/10.1016/j.proeng.2016.08.058.

Solarino, A.M. and Aguinis, H. (2021). Challenges and Best ‐practice Recommendations for
Designing and Conducting Interviews with Elite Informants. Journal of Management Studies,
58(3). doi:https://doi.org/10.1111/joms.12620.

Song, L., Zhang, X., Zhou, K., Chen, S. and Wang, C. (2020). Transparent Deep Learning in
Industrial IoT for Product Quality inspection. IEEE Transactions on Industrial Informatics,
16, 3446-3455..

Squires, A. (2009). Methodological challenges in cross-language qualitative research: A

research review. International Journal of Nursing Studies, [online] 46(2), pp.277–287.
doi:https://doi.org/10.1016/j.ijnurstu.2008.08.006.

Stenbacka, C. (2001). Qualitative research requires quality concepts of its own. Management
Decision, 39(7), pp.551–556. doi:https://doi.org/10.1108/eum0000000005801.

Strohm, L., Hehakaya, C., Ranschaert, E.R., Boon, W.P.C. and Moors, E.H.M. (2020).
Implementation of artificial intelligence (AI) applications in radiology: hindering and
facilitating factors. European Radiology, [online] 30(10), pp.5525–5532.
doi:https://doi.org/10.1007/s00330-020-06946-y.

Sun, T.Q. and Medaglia, R. (2019). Mapping the challenges of Artificial Intelligence in the
public sector: Evidence from public healthcare. Government Information Quarterly, [online]
36(2), pp.368–383. doi:https://doi.org/10.1016/j.giq.2018.09.008.

14
8
Taherdoost, H. (2016). Sampling Methods in Research Methodology; How to Choose a
Sampling Technique for Research. SSRN Electronic Journal, 5(2), pp.18–27.

Thun, S., Bakås, O. and Bjørsvik Storholmen, T.C. (2021). Development and implementation
processes of digitalization in engineer-to-order manufacturing: enablers and barriers. AI &
society, 37(2), pp.725–743. doi:https://doi.org/10.1007/s00146-021-01174-4.

14
9
Timme, S.G. and Williams-Timme, C. (2003). The Real Costs of Holding Inventory. Supply
Chain Management Review, 7, pp.30–37.

Van Hoek, R.I. (1998). ‘Measuring the unmeasurable’ – measuring and improving
performance in the supply chain. Supply Chain Management; an International Journal, 3(4),
pp.187–192.

Vinuesa, R., Azizpour, H., Leite, I., Balaam, M., Dignum, V., Domisch, S., Felländer, A.,
Langhans, S.D., Tegmark, M. and Fuso Nerini, F. (2020). The role of artificial intelligence in
achieving the Sustainable Development Goals. Nature Communications, [online] 11(1), pp.1–
10. doi:https://doi.org/10.1038/s41467-019-14108-y.

Waltersmann, L., Kiemel, S., Stuhlsatz, J., Sauer, A. and Miehe, R. (2021). Artificial
Intelligence Applications for Increasing Resource Efficiency in Manufacturing Companies—
A Comprehensive Review. Sustainability, 13(6689), pp.1–26.
doi:https://doi.org/10.3390/su13126689.

Wang, H., Hu, X., Dai, H. and Li, Y. (2021). Overcoming the challenges of implementing
artificial intelligence in purchasing and supply management. Journal of Purchasing and
Supply Management, 27(1), p.100670.

Wang, L., Shen, Z. and WU, D. (2019). An artificial intelligence-based visual inspection
system for quality control of glass containers. Journal of Intelligent Manufacturing, 30(4),
pp.1475– 1484.

Weichert, D., Link, P., Stoll, A., Rüping, S., Ihlenfeldt, S. and Wrobel, S. (2019). A review of
machine learning for the optimization of production processes. The International Journal of
Advanced Manufacturing Technology, 104(5-8), pp.1889–1902.
doi:https://doi.org/10.1007/s00170-019-03988-5.

West, D.M. and Allen, J.R. (2018). How artificial intelligence is transforming the world.
[online] Brookings, Brookings Institute, pp.1–32. Available at:
https://www.brookings.edu/research/how-artificial-intelligence-is-transforming-the-world/
[Accessed 11 Apr. 2023].

Wirtz, B.W. and Müller, W.M. (2018). An integrated artificial intelligence framework for
public management. Public Management Review, 21(7), pp.1076–1100.
doi:https://doi.org/10.1080/14719037.2018.1549268.

Wu, W., Sun, Y., Cui, Y., Liu, L. and Zhang, H. (2020). Surface Defect Detection of Steel
Products Using Machine Learning Techniques. International Journal of Advanced
Manufacturing Technology,108(7-8), pp.2637–2651.

Wuest, T., Weimer, D., Irgens, C. and Thoben, K.-D. (2016). Machine learning in
manufacturing: advantages, challenges, and applications. Production & Manufacturing
15
0
Research, [online] 4(1), pp.23–45. doi:https://doi.org/10.1080/21693277.2016.1192517.

Yang, C.H., Chen, W.C., Chen, Y.Y. and Chen, S.Y. (2019). An Artificial intelligence-based
Predictive Maintenance System for Plastic Injection Molding machines. Journal of Intelligent
Manufacturing, 30(4), pp.1337–1345.

15
1
Zeba, G., Dabić, M., Čičak, M., Daim, T. and Yalcin, H. (2021). Technology mining:
Artificial intelligence in manufacturing. Technological Forecasting and Social Change, 171,
pp.1–18. doi:https://doi.org/10.1016/j.techfore.2021.120971.

15
2