St. Petersburg University
Graduate School of Management
Master in Management Program
Artificial Intelligence readiness in Russian and
Swiss-based mechanical and industrial engineering
companies
Master’s Thesis by the 2nd year student
Daniil A. Fetisov
Research advisor:
Dr. Tatjana A. Samsonowa
Associate Professor
St. Petersburg
2017
ЗАЯВЛЕНИЕ О САМОСТОЯТЕЛЬНОМ ХАРАКТЕРЕ ВЫПОЛНЕНИЯ
ВЫПУСКНОЙ КВАЛИФИКАЦИОННОЙ РАБОТЫ
Я, Фетисов Даниил Алексеевич, студент второго курса магистратуры направления
«Менеджмент», заявляю, что в моей ВКР на тему «Готовность внедрения искусственного
интеллекта в российских и швейцарских промышленно-производственных компаниях»,
представленной в службу обеспечения программ магистратуры для последующей передачи в
государственную аттестационную комиссию для публичной защиты, не содержится элементов
плагиата.
Все прямые заимствования из печатных и электронных источников, а также из
защищенных ранее выпускных квалификационных работ, кандидатских и докторских
диссертаций имеют соответствующие ссылки.
Мне известно содержание п. 9.7.1 Правил обучения по основным образовательным
программам высшего и среднего профессионального образования в СПбГУ о том, что «ВКР
выполняется индивидуально каждым студентом под руководством назначенного ему научного
руководителя», и п. 51 Устава федерального государственного бюджетного образовательного
учреждения высшего профессионального образования «Санкт-Петербургский государственный
университет» о том, что «студент подлежит отчислению из Санкт-Петербургского
университета за представление курсовой или выпускной квалификационной работы,
выполненной другим лицом (лицами)».
Y
28.09.2017
STATEMENT ABOUT THE INDEPENDENT CHARACTER
OF THE MASTER THESIS
I, Fetisov Daniil Alekseevich, second year master student, program “Management”, state that
my master thesis on the topic “Artificial Intelligence Readiness in Russian and Swiss Based
Mechanical and Industrial Engineering Companies”, which is presented to the Master Office to be
submitted to the Official Defense Committee for the public defense, does not contain any elements of
plagiarism.
All direct borrowings from printed and electronic sources, as well as from master theses, PhD
and doctorate theses which were defended earlier, have appropriate references.
I am aware that according to paragraph 9.7.1. of Guidelines for instruction in major
curriculum programs of higher and secondary professional education at St.Petersburg University “А
master thesis must be completed by each of the degree candidates individually under the supervision
of his or her advisor”, and according to paragraph 51 of Charter of the Federal State Institution of
Higher Professional Education Saint-Petersburg State University “a student can be expelled from St.
Petersburg University for submitting of the course or graduation qualification work developed by
other person (persons)”.
Y
!2
АННОТАЦИЯ
Автор
Фетисов Даниил Алексеевич
Название ВКР
Готовность внедрения искусственного интеллекта в
российских и швейцарских промышленно-производственных
компаниях
Направление подготовки
38.04.02 «Менеджмент»
Год
2017
Научный руководитель
Самсонова Татьяна Александровна
Описание цели, задач и Целью данного исследования является изучение факторов,
основных результатов
влияющих на готовность к имплементации Искусственного
Интеллекта в промышленно-производственных компаниях в
регионе, лидирующем в инновациях (Швейцария), и в
отстающем в инновациях регионе (Россия). Для того чтобы
достигнуть этой цели, была проведена консультация с
экспертом промышленно-производственной отрасли, была
применена Модель Принятия Технологий, дополненная
внешними переменными, был составлен и проведен опрос
(102 респондента), а также был осуществлен статистический
ана лиз. Ре зульт аты исследования показывают, что
осуществимость имплементации Искусственного Интеллекта
намного важнее для компаний, чем потенциальная выгода.
Следственно есть перспектива разъяснить компаниям
потенциальную выгоду от имплементации Искусственного
Интеллекта, таким образом, способствуя его внедрению во
всей отрасли.
Ключевые слова
Искусственный Интеллект, принятие технологий, Модель
Принятия Технологий, Швейцария, Россия
ABSTRACT
!4
Master Student's Name
Daniil Fetisov
Master Thesis Title
Artificial Intelligence Readiness in Russian and Swiss Based
Mechanical and Industrial Engineering Companies
Main field of study
38.04.02 «Management»
Year
2017
Academic Advisor's Name
Tatjana A. Samsonowa
Description of the goal, tasks The purpose of this study is to analyze the factors influencing
and main results
readiness levels towards Artificial Intelligence solutions
implementation by mechanical and industrial engineering
companies in a region leading in innovation (Switzerland) and in a
region lagging behind (Russia). In order to achieve this goal, the
research consults an industry expert, uses Technology Acceptance
Model enriched with external variables, designs and conducts a
related survey (102 valid responses) and carries out statistical
analysis. The results of the study show that feasibility is much
more important for companies than potential benefits of
implementation. Therefore, there is an opportunity to educate
companies about the benefits of Artificial Intelligence, thus driving
its implementation in the industry.
Keywords
Artificial Intelligence, technology adoption, Technology
Acceptance Model, Switzerland, Russia
Table of contents
Generating Table of Contents for Word Import ...
!5
INTRODUCTION
Artificial Intelligence (also known as Machine Intelligence; later referred to as AI) is
the science and engineering of making machines do tasks which require intelligence when
done by human beings (McCarthy, 2007). Thus the concept of AI opposes NI (Natural
Intelligence). A particular use (and the most common one) of AI is through intelligent
computer programs.
The increasingly rapid growth of available data incentivized most companies across
various industries to use more structured approach in collecting, processing and storing it.
The companies which intend to be the leaders in the market and reap the benefits first, have
to use a wide range of data analytics tools powered by elements of AI - this is the minimum
requirement for them to stay competitive in the VUCA1 world (Davenport, 2013). Leading
mechanical and industrial engineering companies are no exception: they are now on the
threshold of massive integration of AI solutions (Faggella, 2017), potentially propelling the
overall development of AI and encouraging companies from other industries to follow their
lead.
In order to understand the current situation of AI implementation in industrial
engineering companies better, this research compares state of affairs for two countries Switzerland (one of the leaders of AI use in corporate sector) and Russia (the country lagging
behind, especially for industrial engineering companies). This allows getting a more
comprehensive picture for the analysis.
1 VUCA - VUCA world
(volatile, uncertain, complex, and ambiguous – concept which was created by U.S.
Army War College referring to the new reality after the Cold War)
!6
The research goal of this study is to analyze the adoption of AI solutions by
mechanical and industrial engineering companies in a region leading in innovation
(Switzerland) and a region lagging behind (Russia).
For the research goal to be accomplished, it is necessary to fulfill the following subgoals:
1. Understand the current developments in AI;
2. Design a survey determining status quo of AI solutions in mechanical and industrial
engineering companies;
3. Conduct the survey among middle-level and senior-level employees of mechanical
and industrial engineering companies;
4. Test AI solutions adoption using Technology Acceptance Model (TAM);
5. Enrich traditional TAM with external variables based on theoretical review and sensecheck with an industry expert (in this context - Organizational Resistance to Change,
Perceived Risks and Supplier Support);
6. Compare results of AI acceptance in Russia and Switzerland;
7. Give recommendations and make theoretical and practical contributions
Thus it is possible to formulate the research questions as follows:
1. How do the external variables (Organizational Resistance to Change, Perceived Risks
and Supplier Support) influence the adoption of AI solutions by mechanical and
industrial engineering companies in TAM framework?
2. What are the main differences in AI adoption between the leaders and the laggards –
that is to say Swiss and Russian companies?
3. What are the potential drivers and barriers in AI adoption by mechanical and
industrial engineering companies?
The research explores the following systematic processes for gathering better
understanding of the topic:
•
Literature review – analysis of existing researches on the topic;
•
Theoretical modeling – interview with an industry expert, development of
extended Technology Acceptance Model;
!7
•
Survey design – determination of the variables, formulation of questions;
•
Statistical analysis – data collection, data analysis and graphic representation.
This research is structured in the following manner: introduction, three chapters,
conclusion, list of references and five appendices.
Introduction points out the relevance of this study; research goal, sub-goals, research
questions as well as overview of the structure are presented.
The first chapter studies previous researches on the topic of Artificial Intelligence and
its applications in business, gives an overview of mechanical and industrial engineering
industry, compares AI acceptance in Russian and Swiss companies and examines Technology
Acceptance Model (enriched with external variables). Hypotheses for this research are also
developed in this chapter.
In the second chapter research design and research model are developed, related
survey is created and empirical research is conducted.
In the third chapter the research findings, theoretical and managerial implications,
potential drivers and barriers as well as limitations are outlined. Also the comparison between
Russian and Swiss companies is drawn.
Conclusion summarizes the results, recommendations and potential for future
research.
!8
CHAPTER 1. THEORETICAL BACKGROUND
1.1 Artificial Intelligence (AI): brief overview
First of all let us take a look at the definition of AI given by Oxford English
Dictionary: “It is the theory and development of computer systems able to perform tasks
normally requiring human intelligence”; thus AI is based on the principles of human
cognition. The researchers mostly distinguish the following five elements of human
intelligence used in AI building principles: learning, reasoning, problem-solving, perception,
and language-understanding (Copeland, 2012).
There are many types of learning, but the most common ones used in intelligent
programs are rote 2 learning and generalization. The former is basically a simple
memorization of individual things – e.g. mate-in-x moves in chess or Sudoku engines: they
simply try out all the possible moves until the successful outcome. The latter is based on the
principle of learning the situations so that machine performs better in similar situations they
have not previously come across – e.g. if a machine encounters a word with suffix ‘-ment’
2
Rote learning - learning by memorization without proper understanding or reflection; mechanical learning
(Oxford English Dictionary)
!9
and is told that it is a noun once, it can predict that words with similar suffixes are nouns as
well.
Reasoning means drawing appropriate conclusions based on presented data. There are
two types of conclusions: inductive and deductive. In deductive conclusions, if premises are
true, then the conclusion is true (e.g. ‘Company X can either take a loan or offer its equity;
company X didn’t take a loan, thus it offered its equity’). In inductive conclusions premises
support the conclusion, but do not necessarily guarantee it is true (‘All clients of bank X
receive 5% cash back; Ivan received 5% cash back, thus he is a client of bank X’ – not
necessarily true). There has been a significant breakthrough in teaching machines to draw
inferences; however reasoning includes drawing conclusions which are relevant to the task,
and data scientists are now struggling to make AI differentiate relevant conclusions from
irrelevant (the so-called noise).
In terms of problem-solving methods it is possible to outline two types: specialpurpose and general-purpose. Similar to learning principles, special-purpose method is
designed to work out a specific problem, whereas general-purpose method deals with a wide
range of various problems. An example of a general-purpose method in AI is means-end
analysis – a program selects from the possible means (actions), executes them, and repeats if
necessary until the current state is transformed into a pre-defined goal state (e.g. a robot is
programmed to pick up boxes until there is nothing left).
In terms of perception, the environment is examined by various sensors, then
information is processed and analyzed and appropriate response is made. Currently artificial
perception is well-developed – cleaning robots are roaming offices, collaborative robots
allow employees to work together at factories and autonomous cars can drive at moderate
speed almost without any accidents. This element is predicted to grow fastest in the near
future – at a CAGR3 of 7.67% during 2017-2021 (Zervos, Ghaffarzadeh and Harrop, 2017).
It is rather easy to formulate certain phrases/sentences using a language (including
artificial languages) and its syntax. However it is much harder to understand them. Modern
AI is still not completely capable of creating comprehensive system of language
3
CAGR - is a business and investing specific term for the geometric progression ratio that provides a constant
rate of return over the time period
!10
understanding. Some scientists believe that this step might be the most important one in
creating next generation AI (Copeland, 2012).
1.2 AI development and its applications in business
Some researchers believe that the AI expansion started out with the invention of
computer in the early 1940s (Panczyk and Rudzinski, 2002). The majority however think that
actual development of AI began a dozen years later at the AI conference in Dartmouth
College in 1956 – the so-called Dartmouth summer research project on AI. During this
workshop approximately 20 scientists and mathematicians brainstormed and argued about the
possibilities of machines “behaving intelligently” (Veale, 2001).
1.2.1 Expert systems
The first commercial application of AI was made in the late 1970s with the
introduction of expert systems – computer software that attempts to mimic the reasoning of a
human specialist (Jackson, 1998); expert systems became one of the first (if not the first)
genuinely successful applications of AI (Russell and Norvig, 2010). Expert systems were
introduced to solve complex problems based on drawing meaningful inferences (rule-based
system). One of the biggest advantages of expert systems is that they demonstrate the logic
behind every inference – why a particular decision was made, why certain options were
eliminated etc.
There were several successful expert systems at the early stage – one of the first was
XCON (or Expert CONfigurer, later on called R1). It was developed to validate technical
correctness of customers’ orders and guide the assembly of such orders for Digital Equipment
Corporation (DEC). The program was a definite success: XCON achieved from 95 to 98%
accuracy while validating and sorting orders and drastically increased the speed of assembly.
The overall net return for DEC thanks to XCON implementation was estimated to be more
than $40 million per year (Blecker and Friedrich, 2005). Another example of a successful
expert system was Mycin – a program which identified diseases based on patients’ symptoms
and other factors. The expert system also recommended treatment and dosage of medicine –
according to the data of patients: weight, allergies etc. A special commission at Stanford
Medical School concluded that Mycin suggested appropriate treatment in 65% of the cases –
!11
a better result than that of human experts (average score – 52.5%), who made decisions based
on the same factors as Mycin (Yu, 1979).
In the 1980s expert systems were spreading even more rapidly. Among the leaders at
the high-end expert systems market were such companies as Xerox and Texas Instruments.
However after such hype of the 80s, expert systems ceased to be a separate AI concept in the
1990s. Instead, such systems were integrated with other solutions (such as PC) in accordance
to the businesses’ needs and the new VUCA world (volatile, uncertain, complex, and
ambiguous – concept which was created by U.S. Army War College referring to the new
reality after the Cold War).
The further progress of AI continued in late 1990s – this was primarily due to the
surge in computing power and meticulous work of computer engineers (Mead and Kurtzweil,
2006). Such events as Deep Blue beating chess champion Garry Kasparov in 1997, Stanford
robot autonomously driving in regular traffic for 131 miles in 2005 and IBM Watson winning
in “Jeopardy!” game in 2011, show the skyrocketing potential of AI capabilities.
1.2.2 Artificial Neural Networks
With the development of AI another concept started to spread and applied across
various industries – Artificial Neural Networks (ANN). ANN is a computing system which
consists of layers of hidden nodes and layers of output nodes, which react to the external
inputs (Wang, 2003). Nodes are similar to neurons in brains – they are interconnected with
each other. Please refer to exhibit 1 for visual representation of an ANN.
!12
Exhibit 1. ANN example
The applications of ANN are enormous and cover most of the AI use cases – from
predictive analytics and decision-making tools to pattern recognition and data mining.
ANNs are not programmed to execute the same operations; instead they learn to
recognize specific patterns. Researchers point out 3 main types of such learning: supervised,
unsupervised and reinforcement learning (Suzuki, 2011).
The key distinctive feature of supervised learning is that it has pre-planned target
output. ANN learns by setting values of its parameters for any valid input values after having
seen output values; the training data is made up of pairs of input and desired output values
(Suzuki, 2011). The process of supervised learning usually includes several stages. First of all
it is necessary to recognize specific type of training data. Then one has to gather training data
which meets the criteria for solving a certain problem. After that it is necessary to translate
training data into an appropriate code which is comprehensible for ANN. Then ANN
conducts the training by itself. Lastly, the performance of ANN is assessed after the learning
with the test data set – it has not been given to ANN for learning but has similar structure as
!13
the learned data. One of the most common uses of supervised learning of ANN is pattern
recognition.
Unlike supervised learning, unsupervised learning does not have any target outputs.
The specific feature of this type of learning is that ANN receives only examples which are
not tagged. ANN determines the data structure and looks for certain patterns by itself.
Unsupervised learning is mostly used for tackling various estimation problems such as
clustering, filtering, forecasting and estimation of statistical distributions; the model which
resorts to unsupervised learning the most is self-organizing map (Kohonen, 1989).
Lastly, in case of reinforcement learning, examples are normally not given to ANN;
rather they are created via ANN’s interactions with the environment. ANN collaborates with
the environment so that it finds optimal actions to receive long-term reward (the ability to
learn from their environment is generally considered one of the greatest benefits of ANNs).
Reinforcement learning is often integrated into ANN’s general learning algorithm. It is most
frequently applied to deal with sequential decision-making problems, such as game engines
(checkers, chess, go), telecom and others.
Thus ANN has many benefits that make it indispensable in terms of commercial use –
it is very good at problem-solving (e.g. pattern recognition), decision-making as well as
forecasting (these areas are especially important for the industry studied in this paper –
mechanical and industrial engineering companies).
1.2.3 Deep learning
Deep learning is rather similar to ANN in terms of pattern recognition techniques. It is
based on feature learning, which allows a system “to automatically discover the
representations needed for feature detection or classification from raw data” (Bengio,
Courville and Vincent, 2013).
Deep learning applies backpropagation method in order to establish structure in
datasets. Deep learning techniques largely contributed to the significant development in such
areas as visual recognition, speech recognition, video processing, object detection and many
others (LeCun, Bengio and Hinton, 2015).
The main difference between Artificial Neural Networks and Deep Learning (even
though these models are much interconnected most of the time) is the amount and structure of
!14
data it is applied to. Deep Learning platforms have more hidden nodes (layers) than ANNs
and normally deal with significantly bigger datasets.
1.2.4 Robotic Process Automation
Broadly speaking Robotic Process Automation (RPA) is a type of software which
allows imitating human behavior while executing tasks within a certain process. It can carry
out repetitive tasks much faster and with higher precision than human employees, moreover
RPA does not get tired of executing similar tasks over and over again. RPA systems benefit
both executives and employees: routine tasks can be done faster and more accurately, while
employees can do other tasks more focused on creative side, emotional intelligence and
customer interaction (Willcocks, 2016).
RPA, unlike most of other AI solutions, is designed to carry out simple tasks, e.g.
entering purchase invoices in a company’s Enterprise Resource Planning system. Most of the
time RPA systems have to be provided with specific instructions; they rarely allow any
variability in decision-making process.
1.2.5 Virtual Agents
Virtual Agents (sometimes referred to as Intelligent Agents or Autonomous Intelligent
Agents) receive information about the environment using sensors (or similar tools) and then
execute the responsive action in order to achieve a goal. Virtual Agents vary in types;
according to Russell and Norvig (Russell and Norvig, 2010), there are 5 main types of Virtual
Agents:
1. Simple reflex agents;
2. Model-based reflex agents;
3. Goal-based agents;
4. Utility-based agents
5. Learning agents
!15
The first group initiates a response based solely on the current state of environment
completely ignoring historical data. Its decision-making follows the so-called conditionaction rule (if there is certain condition then agent executes a reciprocal action).
Model-based reflex agents can operate within partially observable environments. By
collecting data from the environment, the model gets the general understanding of how the
environment works.
Goal-based agents are more developed than their model-based counterpart in the way
that they additionally use the information about specific goals of the model. Thus goal-based
agents may choose a path among multiple options so that the chosen path achieves the final
goal. Goal-based agents are flexible since it is possible to modify the knowledge base which
supports the model’s decision-making.
If goal-based agents operate within a binary framework (goal is achieved or not
achieved), utility-based agents select an action which maximizes the desired utility (utilitybased agents choose the path which satisfies the goal to the biggest extent to put it simple –
i.e. what is the best outcome among all the probabilities).
Learning agents, unlike the former 4 types, initially operate in an unknown
environment thus becoming more competent. Learning agents have 2 core elements – the
learning element and the performance element. The former is responsible for making
improvements in the model and the latter deals with choosing appropriate responsive actions
with the environment. Given the complexity of this type of agent, it is by far the most
sophisticated.
The most common use case for virtual agents is automated online assistant (chatbot)
and similar customer service and marketing tools.
1.2.6 Natural Language Processing
Natural Language Processing (NLP) has relatively narrower focus than other AI
solutions – it deals with creating software capable of processing a natural language. Even
though the focus is narrow, the task itself is one of the most complicated and complex ones.
Some relatively successful NLP products already exist in the market as of 2017 (e.g. IBM’s
Watson, Amazon’s Alexa, Apple’s Siri, Google Translate and others), however the full
!16
integration of natural language and machine perception has not yet been achieved, and there
is still huge commercial potential behind this AI solution.
1.2.7 Hybrid systems
The last AI method analyzed in this research is hybrid systems. Those are “systems
that use more than one problem-solving technique to solve a problem” (Gray and Kilgour,
1997). This paper analyzes 2 types of hybrid systems since they have the widest range of use
in business – namely those methods are Fuzzy Expert Systems and Data Mining.
First of all here is a brief explanation what fuzzy logic is. Unlike classical Boolean
logic with only 2 possible values (True or False), Fuzzy logic operates with a range of values.
Each of these values reflects various degrees of truth on a scale between completely false and
completely true (Meana et al., 2016). This approach creates resemblance with human
reasoning where there are few absolute values and many grey areas.
Fuzzy expert systems are expert systems which operate based on fuzzy logic
principles. Fuzzy logic finds the most common use in such systems (Kantrowitz et al., 2001).
These systems are quite efficient in business since they allow more precision in decisionmaking process. The main use cases of such systems in business are in planning, designing
and as a decision support tool.
Data mining (also known as Knowledge discovery databases or Information
discovery) uses AI to find useful insights in huge amounts of data. The software allows
discovering relationships and associations which are not so easy to find by a human being,
the main constraint being the amount of time necessary to complete the task (Port, 2001).
Normally the Data mining process works like this: first data is loaded into Data
mining database, then Data mining techniques are applied, after that the software finds
correlations, trends and unusual patterns, and lastly the software interprets the results
(normally via visualization tools such as Wave, Tableau and others) (Brown, 2012).
Data mining process uses a number of different techniques for data analysis in order
to gather useful information. Here are the main methods for data mining (Gheorghe and
Petre, 2014):
● Clustering – the tool discovers a finite number of categories to describe the data;
!17
● Classification – data items are divided into one of several predefined categories;
● Regression – a function mapping data items to a real-valued prediction variable is
created;
● Association rule learning – a model describing significant dependencies between
variables is created;
● Deviation detection – a model discovering the most significant changes compared
to previously measured data or benchmark is created.
Both Fuzzy expert systems and Data mining tools can be very effective in commercial
use since they allow to make more precise decisions based on human-like logic, and deal
with huge amounts of information, analyze it very quickly and give meaningful insights.
1.2.8 Conclusion
With the exponential growth of data, both internal and external, the companies are
currently facing big challenges with data analysis (which information is actually relevant?)
and decision making (what should I do based on the information provided?). The AI solutions
mentioned above are designed to tackle this problem, most of the times even more effectively
than human employees can.
While most companies today are interested in implementing the AI opportunities, they
only see such solutions as supportive tools for their management. Some researchers however
have a different standpoint here. For example, Marketing Director of Yandex Andrei Sebrant
in his recent interview to Malina.am shared the following view: humans are objectively
worse at analyzing information, detecting patterns, making predictions and recommendations,
people should consider allowing machines not only giving insights to people, but also making
decisions themselves (in his example Sebrant mentioned ANNs, but meant machine
intelligence in general).
1.3 Mechanical and industrial engineering companies’ overview
!18
As it follows from the name, mechanical and industrial engineering is made up of 2
elements: mechanical engineering and industrial engineering. The former profession consists
of specialists who work on design, development and production of various mechanical
systems; this field is rather broad and most of the times it also includes industrial engineering
according to the researchers (Katz and Talmi, 2017). However industrial engineering has its
own differences. It stands in between engineering and business (much closer to engineering
though). Industrial engineers have to take daily operational business decisions regarding
many aspects: quality control on site, ensuring maximum efficiency of manufacturing
processes, optimization of productivity of workers and even performing cost analyses. The
combination of these 2 professions led to creation of mechanical and industrial engineering
companies.
A good example of such foundation is ABB – a Swiss-Swedish multinational
corporation mainly operating in power, industrial automation and robotics. This conglomerate
was created in 1988 after the merger of 2 companies: Swedish ASEA and Swiss Brown,
Boveri & Cie (BBC). The history of these 2 companies goes back to the end of the 19th
century – they were both founded by electrical engineers. ASEA started manufacturing and
selling light bulbs and generators, while BBC produced motors, steam turbines and
transformers. This collaboration of engineers turned out to be quite fruitful – ABB today is
one of the biggest players in the industry with operations all over the world, in approximately
100 countries (ABB official website, 2017); it currently holds 314th position of 500 biggest
companies worldwide by revenue (Fortune 500, 2017).
Since the 1970s the industry has become a leader in development and application of
high technology, integrating the first AI solutions among other things. In spite of the fact that
mechanical and industrial engineering industry is traditionally considered as the one
producing machinery and hardware, it has moved significantly towards the service industry –
the companies install equipment, train personnel, conduct maintenance and repair works.
Such services have 2 main benefits: they significantly increase revenues and also reduce
exposure to low-cost competition (Vieweg, 2012).
According to some researchers there are 4 major factors influencing industries to
adopt AI solutions: substantial budget, large amount of organized data and the ability to
acquire AI experts, data scientists and additional talent (Faggella, 2017). The first factor
!19
describes a company’s ability to invest in a technology which may not necessarily lead to fast
ROI. In terms of big amount of structured data, companies which possess such information
can get much greater value out of it – more insights, more patterns, better possible forecast
for the future. Lastly, in order to implement a new AI tool it is necessary to attract experts in
this field. The main attraction here is generally considered to be money, but it also includes
brand appeal of the company.
Keeping in mind those 3 factors, it is possible to assume that mechanical and
industrial engineering companies are on the verge of mass implementation of AI solutions.
Such companies have sufficient budgets to invest in such solutions, at least the leaders (total
revenue of the companies in this industry in Fortune 500 list (15 overall) was approximately
$650 billion in 2016). Such companies also have huge amounts of data; however data
management processes in such companies are often in poor state, especially if they do not
have standardized processes across the markets (interview with ABB Group Vice-President of
Sales & Marketing, 2017). As mentioned before, mechanical and industrial engineering
companies have significant budgets, so it is not a problem for them to attract AI experts and
other talent with high salaries; also the brand image of most of such companies is positive
(e.g. renewable energy companies such as GE and Siemens), which also plays a big role in
talent acquisition. Overall, theoretically such companies should be about to implement more
and more AI solutions in the near future.
Currently the level of adoption of AI in the industry varies significantly. Some
industry leaders have been trying to integrate AI solutions for decades – e.g. Siemens have
been conducting in-depth research in this area for more than 30 years and have implemented
several of them – e.g. artificial neural networks in steel mills (Siemens official website,
2017), while others are still looking up to the industry leaders and assessing AI potential. This
research intends to understand the overall readiness of mechanical and industrial engineering
companies to implement AI solutions and their possible implications from middle
management and senior management perspective.
1.4 Russian and Swiss companies’ comparison
This research intends not only to look at readiness for AI adoption in mechanical and
industrial engineering companies, but also to further narrow down the scope and compare
!20
such companies in 2 countries - Russia and Switzerland. The reason for this exact comparison
lies in the level of development in AI. Russian companies are lagging behind their
counterparts, and Swiss companies are considered the leading force in innovation and as a
consequence AI implementation. Therefore in order to stay competitive and lead the change,
Russian companies have to constantly monitor and follow some of the best practices of the
leaders. Thus this research will analyze and compare Russian and Swiss mechanical and
industrial engineering companies, outline the most important findings and suggest
recommendations.
So why exactly does this research draw comparisons with Swiss companies? Why not
other countries leading in AI implementation and the level of innovation in general? The
traditional metric for a country’s innovativeness level is the number of research papers
published in this country (Cheng and Krumwiede, 2017). However, the sheer number of
papers does not necessarily give the full picture for understanding the situation - a huge
proportion of papers may not have sufficient citations, thus questioning the quality of such
papers.
There is another approach when comparing research papers - the so-called FieldWeighted Citation Impact (FWCI). This metric shows the relation between the number of
citations of researchers’ publications and the average number of citations received by all
other similar publications (Aldieri, Kostemir and Vinci, 2017). This metric is important to
show the real value of research papers. For example, China was the leading country in the
number of AI publications from 2011 to 2015, having published approximately 40% more
papers than the second country in this list - the USA. However, in terms of FWCI China was
only 34th, significantly lagging behind the leaders.
FWCI looks up similar publications in Scopus database; the publications are
determined based on 3 characteristics: the same year of publication, type of research and
studied discipline. For example, FWCI of 1.00 means that a certain publication has been cited
exactly the same number of times as an average number of similar publications in the world.
Thus after comparing countries’ research papers on AI using FWCI metric, we discover that
Switzerland holds the first position with FWCI of 2.71 (Source: Elsevier/Scopus database).
Another factor making Switzerland the most suitable country for comparison in AI
readiness levels with Russia is the level of innovation of the countries. In order to determine
!21
this parameter, the best approach is to use Global Innovation Index (GII), which was
developed by Professor Soumitra Dutta of INSEAD in 2007. Under GII the term ‘innovation’
is considered as “the implementation of new or significantly improved products (goods or
services), a new process, a new marketing method, or a new organizational method in
business practices, workplace organization, or external relations”, borrowing the definition
from Oslo Manual developed by OECD (Organization for Economic Co-operation and
Development) (OECD and Eurostat, 2005). In terms of the framework itself, GII is based on
the innovation efficiency ratio, which in turn is calculated with innovation input and
innovation output indices (the former relies on such parameters as institutions, human capital
and research, infrastructure, market and business sophistication, while the latter is based on
knowledge and technology outputs and creative outputs; each of these parameters also has
several sub-parameters) (Global Innovation Index report, 2017). All things considered,
according to GII report in 2017, Switzerland is in the first place (it actually holds its first
position for the 7th consecutive year), while Russia is only 45th.
Thus in order to have a comprehensive understanding of AI readiness of companies
and look at the best practices of leaders and pain-points of laggards, the best option in the
framework of this research is to compare Russian companies with the Swiss ones.
1.5 Adoption of AI by companies: Technology Acceptance Model
After the development of AI and other adjacent technological advancements, many
researchers started to investigate practical use and adoption of AI by companies in particular.
Since users’ adoption is crucial for emerging technologies, technology acceptance became
one of the most important fields for researchers. They started looking for the factors which
would influence the adoption of such technologies and eventually came up with several
models which satisfied their criteria. One of the very first models was Theory of Reasoned
Action (TRA) developed by Icek Ajzen and Martin Fishbein in 1980. This theory tries to
predict people’s actions based on 2 factors – pre-existing attitudes and behavioral intentions.
Another model which aims to analyze factors influencing intentions and behavior of people is
Theory of Planned Behavior (TPB) also developed by Icek Ajzen. TPB is widely recognized
in social psychology as very efficient; it tries to explain consumer behavior in different
situations, conditions and domains (Klöckner and Verplanken, 2012). The main idea of this
!22
theory is that intentions of consumers are built on three blocks – their attitude, subjective
norm (perceived social pressure to engage or not to engage in a behavior) and perceived
behavioral control (people's perceptions of their ability to perform a given behavior) (Ajzen,
1985).
TPB was later transformed and became the underlying foundation of another model
more applicable to business – Technology Acceptance Model (TAM) developed by Davis in
1986. This model (along with several modifications – e.g. extended TAM, also known as
TAM2) is still widely used for understanding how individuals along with organizations might
adopt a new technology and which factors influence their decision (Lin, Shih and Sher,
2007). Its main purpose is to foresee the factors which motivate users to accept, use and stay
loyal to a certain technology, such as various AI tools for companies in this case (Chiou and
Shen, 2012).
According to TAM (see Exhibit 2), the actual use of a certain technology is directly
influenced by its users’ motivation (Behavioral Intention to Use - BI). While the previous
statement might seem rather obvious, others are not necessarily so straightforward. BI
depends on perceived Attitude Towards Using (AT), which in turn is affected by 2 factors:
Perceived Usefulness of technology (PU) and its Perceived Ease of Use (PEOU). The former
factor (PU) is described by Davis as “the degree to which a person believes that using a
particular system innovativeness which makes the technology better than its predecessor in
the minds of users”. Also PU has a direct impact on BI. The latter factor (PEOU) reflects the
level of complexity, which is evaluated based on how difficult a new technology is for
understanding for its users. Finally TAM includes External Variables, which affect both PU
and PEOU. They heavily depend on the technology which is being evaluated as well as other
practicalities, such as industry, users themselves etc.
!23
(
Exhibit 2. Technology Acceptance Model (source – Davis, 1989)
According to the updated version of TAM, the researchers (Davis and Venkatesh)
became skeptical about the importance of Attitude Towards Using in influencing Behavioral
Intention to Use, so the former parameter was removed from extended TAM. They thought
that AT did not completely mediate relationship between PU and PEOU with BI (Venkatesh
and Davis, 2000). Therefore, PU and PEOU become two of the most important factors
influencing the intention to use a technology, AI solutions in case of this research. If we drill
down even further, some researchers give a definite statement that PU plays a major role in
recurrent use of a technology, whereas PEOU may not have substantial and long-lasting
direct effect (Premkumar and Bhattacherjee, 2008); thus it is possible to conjecture that PU
will also be the most important parameter in this research.
In order to measure PU and PEOU, the researchers (Ajzen, Fishbein and Davis)
applied the following method - there were 5 bipolar adjectives with a seven-point scale (from
the highest to the lowest degree). According to a number of researchers, this method proved
to be reliable, provided high-quality results which were easy to measure, was easy to conduct
and did not take long time to carry out (Zaichkowsky, 1985). That is why this model was
selected among other technology adoption models. Given the advantages of TAM, a similar
approach will be conducted in the empirical part of this research.
!24
1.6 Adoption of AI by companies: external variables in TAM
After careful analysis of existing publications on AI adoption in Scopus, EBSCO and
Google Scholar databases (the following keywords were used, either in combination with one
another or alone: “Artificial Intelligence”, “AI”, “Expert System”, “Decision Support
System”, “Decision Making System”, “Artificial Neural Network”, ”ANN“, “Deep
Learning“, “Data Mining”, “Process Automation“, “Virtual Agent”, “Natural Language
Processing”, “Technology Adoption”, “Technology Acceptance Model”, “TAM”), it turned
out that most researchers focused on applications of AI solutions; publications on their
adoption or integration were scarce. Most papers analyzing AI focused on Artificial Neural
Networks applications, usually focused on a specific industry - most notably physics and
engineering, e.g. Paguio and Dadios (2012) or Sumathi and Bansilal (2016). Overall business
context was rather underrepresented - most publications on AI solutions were concerned with
specific applications of AI in such industries as medicine, chemistry and engineering.
Therefore the topic of adoption of AI solutions by companies was not covered sufficiently.
One particular publication was related to the adoption of an AI solution in corporate
sector and was elaborated quite well - the research paper by G.Rigopoulos, J.Psarras and
D.Askounis (2008). The work was exploring users’ (responsible employees of a bank)
attitude towards adoption of Decision Support Systems (DSS) in their daily work. The
research used a revised Technology Acceptance Model for measuring adoption attitudes,
focusing on PU and PEOU. However this research method showed rather limited results and
did not fully explore the underlying factors influencing DSS adoption due to the lack of
additional external variables as proposed by Davis (1989). Despite this limitation, the
research provides a solid foundation for our research. TAM proved to be an efficient model in
the context of this publication; all 6 of the initial hypotheses were supported. This research
also intends to use TAM and complement it with additional external variables for better
understanding of AI adoption readiness.
In terms of selection of exact external variables for TAM, researchers agree that
there is no universal rule of thumb when making this decision (Legris, Ingham and Collerette,
2003). In order to determine the external variables, a number of technology adoption and user
!25
acceptance publications (mostly evaluating acceptance of an IS technology) using TAM or
similar models were analyzed. The researchers considered these variables to indirectly
influence the final technology acceptance decision. After the analysis a list of variables
potentially suitable in context of this research was made (see Table 1).
Table 1. Research of publications on external variables used in TAMs and similar models
Authors
Technology
analyzed
Model
External variables
Data collection
method
Brock, Khan
(2017)
Big Data
TAM
Organizational
learning capabilities
Survey with 359
respondents
Cho, Sagynov
(2015)
E-shopping
Attitude model
(7-point Likert
scale)
User acceptance, risk
perception, trust
Survey with 216
respondents
Chong, Ooi,
Lin, Tan
(2010)
Online
banking
TAM
Government support,
trust
Survey with 103
respondents
Chyou, Kang,
Cheng
(2012)
QR code
TAM
Social influence,
awareness
knowledge,
facilitating conditions
Survey with 287
respondents
Juan, Lai, Shih Building
(2016)
Information
Modeling
Customized
model (partly
includes TAM)
Organizational
resistance to change
Survey with 300
respondents
Lin, Persada,
Nadlifatin
(2014)
E-Learning
System
TAM
Interactivity
perception
Survey with 302
respondents
Lurudusamy,
Thurasamy
(2016)
Broadband
UTAUT
(Unified Theory
of Technology
Acceptance and
Use of
Technology)
Risk perception,
Survey with 450
perceived
respondents
innovativeness, social
influence,
performance
expectancy, effort
expectancy
Internet
!26
Mingxing,
Jing, Yafang
(2014)
Mobile
Payment
Systems
TAM
Trust perception, risk
perception
Ortega Egea,
González
(2010)
Electronic
Records
System
TAM
Institutional trust, risk Survey with 254
perception
respondents
Pai, Huang
(2010)
Business
Information
Systems
TAM
Information quality
Survey with 294
respondents
Pantano, Rese, Augmented
Baier
Reality
(2017)
TAM
Quality of
information
Survey with 318
respondents
Robinson,
Marshall,
Stamps
(2005)
Technology
TAM
Support services and
trainings
Survey with 218
respondents
Shih, Chiu,
Chang, Yen
(2008)
RFID
TP/NP
technology
adoption model
Organizational
resistance to change,
operation efficiency
Survey with 134
respondents
Wu, Wang
(2004)
Mobile
TAM
Cost compatibility,
risk perception
Survey with 310
respondents
for
Survey with 196
respondents
salespeople
Commerce
Table 1 (cont.). Research of publications on external variables used in TAMs and similar models
After making a list of existing external variables in TAM framework which are
potentially applicable to this research as well, it was necessary to determine which variables
are of interest in context of the research - i.e. for determining AI solutions acceptance factors
in mechanical and industrial engineering companies.
In order to do so the researcher contacted Group Vice-President of Sales of Marketing
& Sales in ABB and conducted an interview regarding this matter (August, 2017). Even
before Mr. Vice-President was presented with the list of potential external variables, he
pointed out that the support of AI solution supplier impacts PEOU and PU; he gave an
example of recent AI data analytics tool in ABB - extensive online and later on-site trainings
!27
were necessary for the employees to feel confident at using the tool and getting the most out
of it. Also Mr. Vice-President recommended choosing risk perception and organizational
resistance to change as external variables since from his perspective they had the biggest
potential to influence the acceptance of AI solutions.
Thus taking into consideration previous research studies reinforced with the interview
with ABB Vice-President, also keeping in mind the industry and geographical context of
research, it is possible to formulate the following research questions:
1. How do the external variables (Perceived Risks, Organizational Resistance to Change,
Supplier Support) influence the adoption of AI solutions by mechanical and industrial
engineering companies in TAM framework?
2. What are the main differences in AI adoption between the leaders and the laggards –
that is to say Swiss and Russian companies?
3. What are the potential drivers and barriers in AI adoption by mechanical and
industrial engineering companies?
1.7 Hypotheses development
As it has been mentioned before, in the framework of TAM Perceived Usefulness
(PU) and Perceived Ease of Use (PEOU) play the most important role in influencing Attitude
Towards Using a certain technology (Davis, 1989). Moreover, PEOU was meant to affect PU
of a technology. Also Davis conjectured that Attitude Towards Using (AT) directly influenced
Behavioral Intention to Use (BI). These relations of parameters have already been supported
and proved significant by a number of previous publications (e.g. King and He, 2006).
Keeping in mind theoretical background of TAM mentioned above, this research
conjectures that original TAM complemented with additional external variables suited to this
specific case is capable of predicting employees’ attitudes towards using AI solutions by the
company. Considering these variables used in the original model by Davis (1989), and
additional external variables, it is possible to make the following hypotheses based on the
model:
Hypothesis 1: Perceived Usefulness (PU) has direct positive influence on Attitude
Towards Using (AT);
!28
Hypothesis 2: Perceived Ease of Use (PEOU) has direct positive influence on Attitude
Towards Using;
Hypothesis 3: Perceived Ease of Use has direct positive influence on Perceived
Usefulness;
Hypothesis 4: Perceived Usefulness has direct positive influence on Behavioral
Intention to Use (BI)
Hypothesis 5: Attitude Towards Using has direct positive influence on Behavioral
Intention to Use
Even though the hypotheses of the classical TAM were successfully tested on a
number of technologies (e.g. Brock and Khan, 2017), these technologies were mostly IS
technologies and not AI. Although some works covered an AI solution acceptance in
corporate sector (e.g. Rigopoulos, Psarras and Askounis, 2008), such publications were
limited to a single solution only (Rigopoulos focused the research on AI powered Decision
Support System only) and did not explore the whole spectrum of AI. Therefore, this research
will bring additional value to this field of study.
In terms of external variables for TAM in the framework of this research, 3 parameters
were selected as mentioned before: Perceived Risks, Organizational Resistance to Change
and Supplier Support. In order to better incorporate these variables in our model, a number of
publications using these parameters were analyzed.
The existing researches using Perceived Risks as an external variable appeared in
publications in 2 variations: either as a parameter indirectly influencing BI (Cho, 2004)
through PU, or as having direct influence (Mingxing, Jing and Yafang, 2014). However most
researchers tend to use the former approach, so we will follow it as well.
Hypothesis 6: Perceived Risks have direct negative influence on Perceived Usefulness
No matter how beneficial a new technology may be for a company, it will inevitably
face a certain degree of resistance (Lippert and Davis, 2006). There is clear logic behind this
idea - new technology will require the change of set methods in a company, thus making
employees dive into different environment, or at least slightly increase their level of stress.
!29
That is why Organizational Resistance to Change may play a significant role in a technology
adoption and appeared in several publications (e.g. Carr et al., 2010).
Hypothesis 7: Organizational Resistance to Change has direct negative influence on
Perceived Usefulness
Lastly, Supplier Support was used in several researches regarding technology
acceptance as an external variable in TAM (Robinson, Marshall and Stamps, 2005); also this
variable was proposed to be included in the model of this research by ABB Vice-President as
mentioned above. A number of studies showed that Supplier Support can have a significant
impact on mitigating employees’ resistance to a new technology as well as increasing the
utilization of such technology (Parthasarathy and Hampton, 1993). Thus, we expect this
variable to affect both PU and PEOU.
Hypothesis 8.1: Supplier Support has direct positive influence on Perceived
Usefulness
Hypothesis 8.2: Supplier Support has direct positive influence on Perceived Ease of
Use
1.8 Research model: extended TAM
Upon reviewing theoretical background to the research and consulting an industry
expert (ABB Group Vice-President), 8 hypotheses were put together. Thus an extended TAM
was developed (see Exhibit 3), which expects to measure the parameters influencing
Behavioral Intention to Use AI solutions by middle and senior management employees in
mechanical and industrial engineering companies.
!30
Y
Exhibit 3. Extended Technology Acceptance Model with hypotheses
This model contains 4 basic variables (PU, PEOU, AT and BI) as well as 3 additional
ones; it is meant to measure the presented variables’ influence on AI solutions acceptance in
mechanical and industrial engineering companies.
CHAPTER 2. RESEARCH METHOD
2.1 Research design
This research encompasses several methods of gathering data. First, an extensive
literature analysis was carried out. Then after getting initial understanding of the researched
topic, an in-depth interview with Group Vice-President of a major mechanical and industrial
engineering company (ABB) was conducted; its primary focus was on aligning the method of
gathering data and establishing the most suitable external variables, as well as discussing the
general readiness levels and use cases of AI implementation in the industry. Lastly, based on
the review of similar researches and the conducted interview, TAM was taken as the main
research tool for this study. It consists of 7 variables; each of these includes 2 or 3 statements.
The statements were compiled based on the relevant publications and expert interview and
adopted to the context of this research. Each statement is based on a seven-point Likert scale
!31
(from 1 (strongly disagree) to 7 (completely agree)). Overall, there are 19 statements related
to variables of TAM (part 3 of the survey) and 7 questions aimed at better understanding the
respondents themselves - name of the company, country of the company’s operations for a
respondent, position and department the respondent works in, questions measuring current/
potential AI usage in the company and hierarchical levels of AI usage.
It is worth mentioning that it is the whole company that adopts AI solutions, not the
individual employees. But the only way of measuring the acceptance for the company is
through its employees. Due to this fact, the respondents of the survey were asked to use a
mechanical and industrial engineering company as a frame of reference while taking the
survey - as a result we received individual answers taken through the prism of the whole
company.
The survey starts with several introductory questions, after that a brief description of
AI is given in order to refresh respondents’ memory of the concept or to educate them. Lastly,
19 TAM statements go after this information block.
The survey is offered to the respondents either in English or in Russian based on the
respondent’s country of work (during the testing phases of the survey it was discovered that
employees from Russian-based companies had significant difficulties with understanding of
the questions of the survey, thus it was translated into Russian).
2.2 Sample description
The survey was presented to two categories of respondents - employees from
mechanical and industrial engineering companies (middle or senior management levels) from
either Swiss-based companies or Russian-based ones. This means that companies do not
necessarily have headquarters in Switzerland or Russia, but have offices and ongoing
operations in these countries.
The survey was sent directly to senior and middle-level managers; those employees
were also asked to share the results with their colleagues. Because of the personal influence
of certain senior-level executives, the survey was shared with a large number of employees. It
is rather difficult to estimate the exact number of employees this survey was offered to, but it
was approximately 160 people. The total number of respondents is 102 employees (54 from
!32
Switzerland and 48 from Russia; please refer to Appendices 1 and 4 in order to examine
survey sample as well as interesting insights from the responses). Therefore, the rate of
response to the survey is 62,5%.
2.3 Statistical analysis
In order to perform the analysis, 2 types of statistical software tools were used - IBM
SPSS Statistics and its added-on module AMOS 24.0.
First of all, in order to test the fit of the model, this research applied Confirmatory
Factor Analysis. Based on it, several indices were calculated, such as ratio of Chi-Square to
Degrees of Freedom (DF), Incremental Fit Index (IFI), Tucker-Lewis Index (TLI) and
Comparative Fit Index (CFI). After that Convergent Validity, Composite Reliability and
Discriminant Validity were calculated.
Lastly in order to test the hypotheses, this research resorts to structural equation
modeling method (i.e. path analysis using latent variables – questionnaire items).
2.4 Measurement model & structural model
The statistical analysis was conducted using item-total correlation technique. This
method allows determining the degree of correlation between statements used in the survey;
it also allows to relate the statements with corresponding variables.
Next, Confirmatory Factor Analysis was carried out in order to test the fit
measurement of the model.
The model proved to be an adequate fit. The relation of Chi-Square to Degrees of
Freedom equals 2.386, the result which does not exceed the threshold accepted by most
researchers (e.g. Hu and Bentler, 1999) (please refer to Table 2 in order to compare the fit
indices with the respective thresholds). The levels of IFI, TLI and CFI do not exceed the
maximum amount either, having 0.96, 0.938 and 0.957 respectively, showing a very good fit
indeed.
In order to test Convergent Validity, AVE (Average Variance Extracted) values of
latent variables was analyzed. Each of them is more than 0.5, which is the threshold, thus
proving their validity. Moreover, AVE indices are all higher than squared correlations
!33
between variables. Composite Reliability is measured by more than 0.7, also exceeding the
threshold. All these indices confirm the fit of the model.
Table 2. Fit indices compared to thresholds (Hu, Bentler, 1999)
Index
Recommended value
Measurement model value
Chi-Square over DF
<3
2.386
IFI
> 0.9
0.96
TLI
> 0.9
0.938
CFI
> 0.9
0.957
RMSEA
< 0.1
0.055
CHAPTER 3. RESULTS AND DISCUSSION
3.1 Structural equation modeling results
After testing the fit, the research applied structural equation modeling for testing the
hypotheses. This model allows establishing the existence of statistical significance of the
influence of external variables on Perceived Usefulness and Perceived Ease of Use.
!34
Moreover, the relationships between variables of the basic model were tested in order to
validate it.
After the analyses, it was established that all 9 hypotheses are supported by the data
(please refer to Table 3).
Table 3. Results of hypotheses
#
Hypothesis
Estimate
Result
1
Perceived Usefulness -> Attitude Towards Using
0.078
Supported
2
Perceived Ease of Use -> Attitude Towards Using
0.62
Supported
3
Perceived Ease of Use -> Perceived Usefulness
0.88
Supported
4
Perceived Usefulness -> Behavioral Intention to Use
0.442
Supported
5
Attitude Towards Using -> Behavioral Intention to Use
0.134
Supported
6
Perceived Risks -> Perceived Usefulness
0.377
Supported
7
Organizational Resistance to Change -> Perceived
Usefulness
0.224
Supported
8.1
Supplier Support -> Perceived Usefulness
0.092
Supported
8.2
Supplier Support -> Perceived Ease of Use
0.11
Supported
The first 5 hypotheses were focused on the basic model, thus confirming its validity.
The latter 4 dealt with the impact of external variables on either PU (e.g. PR, ORC and SS) or
PEOU (SS).
Let us examine the former group first. Here we can see that both PU and PEOU have
statistically significant impact on AT with Estimates of 0.078 and 0.62 respectively (p <
0.05). It is worth mentioning that PEOU has bigger impact on AT than its counterpart PU.
Thus we can induce that for an average mid-level or senior-level employee of a mechanical
and industrial engineering company the key factor for forming an attitude towards an AI
solution is how easy it is to learn it and what efforts are necessary for its use, not the benefits
it can potentially bring. Therefore, hypotheses 1 and 2 are supported by the data.
!35
Moreover, the ease of use of a technology has a very strong effect on its perceived
usefulness, which follows from SEM results (Estimate equals 0.88, p < 0.05). Hypothesis 3 is
confirmed.
Hypotheses 4 and 5 are concerned with the impact on the intention to use (BI) AI
solutions in the companies by 2 factors - PU and AT. Both of these variables proved to
influence BI having Estimates of 0.442 and 0.134, thus supporting the hypotheses.
Since all of the hypotheses related to classical TAM are supported, it is possible to
conclude that in the context of this research TAM is an acceptable instrument for
investigating the adoption factors of AI solutions by mechanical and industrial engineering
companies.
Now let us consider the second group of variables - the ones that measure the impact
of external variables on PU and PEOU. First we will focus on variables which are
conjectured to influence PU. Perceived Risks, Organizational Resistance to Change and
Supplier Support all proved to be statistically significant antecedents of PU. Also since PR
and ORC were hypothesized to have negative influence on PU, these variables were
considered as reversed-scored in SPSS, meaning that their Likert scale scores were reversed
for the correct analysis.
It is worth mentioning that ORC was proposed to be included in the model by ABB
Group Vice-President during our interview, and the modeling proved his professional
conjecture right.
Supplier Support also proved to impact PEOU. However, SS has much smaller
influence on PU than other external variables, having an Estimate of 0.092 (also relatively
small impact on PEOU - 0.11). In spite of this fact, hypotheses 8.1 and 8.2 proved to be
statistically significant.
Taking everything into account, all of the hypotheses mentioned above (1 through 8.2)
are supported by statistical analyses.
Due to the fact that all the hypotheses were supported by SEM, it is possible to answer
Research Question 1 - the external variables ORC and PR have direct negative influence of
PU, while SS have direct positive impact both on PU and PEOU.
3.2 Comparison of Russian and Swiss companies’ survey responses
!36
Initially this research was considering statistical analysis based on TAM for
respondents of Russian and Swiss-based companies separately. However, the preliminary
analyses showed that the responses for two samples are very similar, thus such comparison
appeared to be pointless. Therefore, two samples were combined and analyzed together using
TAM. However, the responses to the multiple choice questions in Part 1 turned out to be
rather insightful (please refer to Appendix 4).
The split between the countries is rather similar - 54 respondents work in Switzerland
while 48 are employed in Russia. Most of the companies are large multinational corporations
headquartered outside of Russia (84%). The company which had the most respondents from
was ABB (46%).
In terms of current status of AI solutions in enterprises, more than a half of Swiss
respondents stated that AI solutions either have been implemented already or are being
implemented at the moment. Less than 6% of the respondents stated that no AI solution
currently exists in their company, while the plurality of respondents (37%) said that the idea
of AI solutions integration has been proposed and now it is being evaluated. In Russia there is
a very different situation: less than 8% of the companies are currently using AI solutions.
Moreover, looking closely at individual responses it appears that those companies using AI
are all foreign companies operating in Russia (e.g. Schneider Electric, Siemens and ABB).
37.5% of the companies are not even planning to integrate AI solutions in the near future, and
the biggest group (45.8%) is now evaluating this possibility.
The most common AI solutions in Swiss companies appeared to be Data Mining Tools
and Expert Systems. In Russian companies the majority of respondents stated that there were
no AI solutions in their company. Not surprisingly the next most popular responses were
similarly Data Mining Tools and Expert Systems (again these were foreign companies).
The most popular departments using AI solutions were Sales, Operations and Finance
in Switzerland and Sales, Operations and Marketing in Russia. This may explain the reason
for Virtual Agents being the third most popular AI solution in Russia - Virtual Agents are very
commonly used in Marketing (Forbes, 2017).
The hierarchical use of AI solutions is rather similar across countries - senior
management represents the biggest part, twice as much as middle management.
!37
Having analyzed the data from 2 countries, now it is possible to answer Research
Question 2. It is important to emphasize that even though there is a significant difference in
the use of AI solutions and its potential implementation, Russian companies are rather open
to AI solutions acceptance according to TAM; no inherent resistance to adoption of AI was
found during data analysis. Therefore, we can conclude that Russian companies are just as
ready to implement AI solutions and are motivated by similar factors as their Swiss
colleagues. There may be other factors influencing AI implementation in Russian companies,
thus there is an opportunity for future research.
3.3 Potential drivers and barriers
After careful analysis of the results of statistical modeling, it is possible to outline
several drivers and barriers towards adoption of AI solutions by industrial engineering
companies.
As it has been mentioned before, Perceived Ease of Use has much greater impact on
Attitude Towards Using than Perceived Usefulness. This may be explained by the risk-averse
behavior of big organizations - usually they do not want to change a standardized process
unless this change does not incur significant problems with high level of certainty, is easy to
implement and there is a clear benefit of the implementation; also employees, especially the
middle level, may not necessarily understand clear benefits of AI solutions implementation,
which can be a barrier for acceptance of AI.
As it is shown in Table 3, there is a very strong influence on PU by PEOU. This may
mean that the easier the use of an AI solution is for an employee, the more benefits they find
in such a solution. Therefore, it is reasonable to assume that a potential driver for AI
implementation in companies can be the education of employees with regards to AI benefits.
Also the support of AI solutions’ suppliers to the employees using the AI has the potential to
facilitate the integration and thus acceptance; therefore, it is another driver for AI
implementation.
3.4 Theoretical and Managerial contribution
This research is focused on understanding the underlying factors influencing adoption
of AI solutions by mechanical and industrial engineering companies. The research applies
!38
Technology Acceptance Model in order to assess the adoption factors of AI. TAM has not
been used often for AI adoption research; only a handful of related studies exist in EBSCO,
SCOPUS and Google Scholar databases. Neither has TAM been widely used for measuring
the acceptance of a technology by an organization and not individual consumers. The
classical model is also refined with additional external variables (Perceived Risks,
Organizational Resistance to Change and Supplier Support), which have been determined
after analysis of related researches and the interview with an industry expert; the role of these
factors gives a new perspective on the research of AI acceptance.
Moreover, few researches focused on the assessment of a technology implementation
(especially powered by AI) in mechanical and industrial engineering companies, making this
study useful for the companies in this industry.
Furthermore, the comparison between Russian and Swiss AI status quo (or
comparison between leading and underperforming regions in terms of AI) has been
investigated insufficiently.
The model proves that Organizational Resistance to Change and Perceived Risks have
significant direct negative correlation with Perceived Usefulness. Thus companies may
disregard substantial rewards for AI implementation, prioritizing risks over benefits. Future
researches on this topic may investigate the ways of changing this point of view so that it is
easier to integrate AI solutions across an organization and get the most out of this promising
technology. For example, a significant step forward in AI adoption in a company could be
education of employees about AI solutions’ benefits and the ways of overcoming risks and
barriers towards implementation.
This research is especially useful for corporate stakeholders, i.e. mechanical and
industrial engineering companies. It compares Russian companies to the leaders – the Swiss
ones, so that the former can analyze the data, realize that the acceptance towards AI solutions
implementation is similar to their Swiss colleagues and possibly look for opportunities for its
implementation.
Since there is a fast rise of AI development now, the AI solutions are going to be
integrated in more and more companies, therefore organizations embracing AI in the near
future will most likely benefit from the technology becoming early adopters. The use of AI
!39
solutions is not limited to specific departments or functional areas: its potential is significant
and it is likely to continue to grow exponentially in the near future (Purdy, Daugherty, 2016).
3.5 Limitations
The research also has a number of limitations which may be improved in the future
researches.
First of all, the survey sample is relatively limited (102 respondents), even though
sufficient for this Master Thesis. In order to have more precise results for analyses, this
sample should be increased by 100% (approximately 100 responses for each country).
The next limitation is the external variables. This research took 3 variables after
analyzing related publications and checked them with an industry expert. However, there may
be other suitable external variables explaining significant part of variability in the model.
This is an opportunity for future research.
Also the research equalized the notion of employees’ responses with company’s
standpoint. It should be said that employees’ responses may still be subjective (they may not
know the precise state of affairs in their company) and not correspond to the company’s
vision. However, this research considers that in order to gain a company’s perspective, the
best way is to get each employee’s perspective.
Lastly, only 2 countries are analyzed, which may limit the scalability of this research.
A more extensive research may be conducted, combining several countries from each group
of AI solutions implementation - leaders, average and laggards.
!40
CONCLUSION
This research is aimed at investigating the adoption factors of AI solutions by
mechanical and industrial engineering companies in an innovation leading region and a
region with AI implementation level falling behind the leaders. Being able to adapt fastdeveloping AI solutions into a company’s workflow can be the distinctive feature bringing
the company to a leading position due to increased productivity, cost reduction and other
factors. Even though the concept of AI is not new, only few companies truly understand and
reap the benefits of this technological development, thus there is space for companies to use
AI and drastically improve their competitive position.
In order to understand the factors influencing AI adoption by companies, it is
necessary to analyze the factors impacting acceptance of individual employees of these
companies. After investigating existing methods of technology adoption in theoretical and
empirical researches Technology Acceptance Model was chosen for this research (it has been
widely used for Information Technologies adoption studies, but has not been very common
for investigating AI adoption). This research uses classical TAM (Davis, 1989). In addition to
classical TAM, 3 additional external variables were added to it; they were selected after
thorough analysis of existing literature and an interview with the industry expert (Group
Vice-President of ABB).
Keeping in mind research objective and research questions, a total of 9 hypotheses are
developed, where the first 5 are a part of classical TAM and the latter 4 are concerned with
external variables. The method for this research is quantitative with some elements of the
qualitative (i.e. interview with industry expert).
!41
The data was gathered via surveys sent directly to the employees of mechanical and
industrial engineering companies in Switzerland and Russia. The surveys consisted of 2 main
parts - questions targeted at understanding status quo of AI solutions in companies and
questions related to the acceptance of AI solutions by employees. Questions in the second
part were developed based on previous researches using TAM and adapted to the context of
this research. Each tested variable included at least 2 questions (since during the statistical
analysis some questions could prove to be exceedingly correlated and thus had to be excluded
- this indeed happened in this research as well). The questions from the second part were
measured using a seven-point Likert scale.
The data from the surveys was analyzed in IBM SPSS Statistics first and then in
AMOS 24.0. After the cleansing of data, exploratory factor analysis was conducted (EFA),
after that confirmatory factor analysis (CFA) and lastly the structural equation modeling
method. These types of analyses are commonly used when applying TAM, especially the
former two (EFA and CFA).
The results of statistical analyses reflected that the research model met the fit
requirements (reliability, convergent validity as well as discriminant validity) and proved to
be relevant in the framework of this research.
Moreover, 9 out of 9 hypotheses developed for the research proved to be supported.
The former 5 hypotheses of the classical TAM were supported, thus proving the model to be
appropriate. The latter 4 hypotheses were supported as well, meaning that the chosen external
variables indeed influence perceived usefulness or perceived ease of use.
After the analyses mentioned above, the survey responses of employees from
Switzerland and Russia were compared. Since the results of TAM questions were very
similar, the research focused on the questions from the first part of the survey - i.e. status quo
of AI solutions. Here big discrepancies were detected. Much more AI solutions were
implemented in Switzerland than in Russia. However, the levels of acceptance towards AI did
not vary significantly across the countries, thus it potentially can mean that there are
favorable conditions for AI implementation in Russian companies; though thorough analyses
need to be conducted in order to confirm this hypothesis, thus there is an opportunity for
future research.
!42
Bigger impact on AI acceptance by Perceived Ease of Use rather than Perceived
Usefulness may be an evidence of risk-averse behavior of industrial engineering companies,
which are hesitant to change a standardized process unless this change is easy to implement
and it has a straightforward benefits. This poses a barrier towards AI implementation, but
there is a solution to that – educating employees of all levels about the benefits of AI. Also
the suppliers of AI solution should help the employees in using such solutions, thus driving
the acceptance of AI even faster.
LIST OF REFERENCES
1. Aldieri, Luigi, Maxim Kotsemir, and Concetto Paolo Vinci. 2017. "The Impact Of
Research Collaboration On Academic Performance: An Empirical Analysis For Some
European Countries". Socio-Economic Planning Sciences. doi:10.1016/j.seps.
2017.05.003.
2. Angelov, Plamen. 2013. "Autonomous Learning Systems: From Data Streams To
Knowledge In Real-Time". Choice Reviews Online 51 (02): 51-0926-51-0926. doi:
10.5860/choice.51-0926.
3. Anson, Mark, Frank Fabozzi, and Frank Jones. 2011. "Investing In Common
Stock". The Handbook Of Traditional And Alternative Investment Vehicles, 17-39.
doi:10.1002/9781118258248.ch2.
4. Bengio, Y. (2009). Learning Deep Architectures for AI. Foundations and Trends® in
Machine Learning, 2(1), pp.1-127.
5. Bengio, Y., A. Courville, and P. Vincent. 2013. "Representation Learning: A Review
And New Perspectives". IEEE Transactions On Pattern Analysis And Machine
Intelligence 35 (8): 1798-1828. doi:10.1109/tpami.2013.50.
6. Blecker, Thorsten, and Gerhard Friedrich. 2005. Mass Customization Information
Systems In Business. Norwood Mass.: Books24x7.com.
!43
7. Brock, Vitor, and Habib Ullah Khan. 2017. "Big Data Analytics: Does Organizational
Factor Matters Impact Technology Acceptance?". Journal Of Big Data 4 (1). doi:
10.1186/s40537-017-0081-8.
8. Brown, Martin. 2012. "Data Mining Techniques". IBM Official Website.
9. Buchanan, Bruce. 2005. "A (Very) Brief History Of Artificial Intelligence". AI
Magazine 26 (4).
10. Cheng, Colin C.J., and Dennis Krumwiede. 2017. "What Makes A Manufacturing
Firm Effective For Service Innovation? The Role Of Intangible Capital Under
Strategic And Environmental Conditions". International Journal Of Production
Economics 193: 113-122. doi:10.1016/j.ijpe.2017.07.007.
11. Chiou, Jyh-Shen, and Chung-Chi Shen. 2012. "The Antecedents Of Online Financial
Service Adoption: The Impact Of Physical Banking Services On Internet Banking
Acceptance". Behaviour & Information Technology 31 (9): 859-871. doi:
10.1080/0144929x.2010.549509.
12. Cho, Yoon C., and Esen Sagynov. 2015. "Exploring Factors That Affect Usefulness,
Ease Of Use, Trust, And Purchase Intention In The Online
Environment". International Journal Of Management & Information Systems
(IJMIS) 19 (1): 21. doi:10.19030/ijmis.v19i1.9086.
13. Copeland, B. Jack. 2012. Turing: Pioneer Of The Information Age. Oxford
University Press.
14. Davenport, Thomas, and Jeanne Harris. 2007. "Competing On Analytics: The New
Science Of Winning". Choice Reviews Online 44 (11): 44-6322-44-6322. doi:
10.5860/choice.44-6322.
15. Davenport, Thomas. 2013. "The Analytics Advantage". Deloitte Analytics Report.
16. Davis, Fred D. 1989. "Perceived Usefulness, Perceived Ease Of Use, And User
Acceptance Of Information Technology". MIS Quarterly 13 (3): 319. doi:
10.2307/249008.
17. Dessimoz, J., Koehler, J. and Stadelmann, T. (2015). AI in Switzerland. AI Magazine,
36(2), p.102.
18. Estrella-Ramon, Antonia, Manuel Sánchez-Pérez, and Gilbert Swinnen. 2016. "How
Customers’ Offline Experience Affects The Adoption Of Online Banking". Internet
Research 26 (5): 1072-1092. doi:10.1108/intr-03-2015-0092.
19. Faggella, Daniel. 2017. "Artificial Intelligence Adoption - What It Takes For
Industries To Change". Huffington Post.
20. Garijo, D. (2017). AI buzzwords explained. AI Matters, 3(1), pp.4-8.
!44
21. Gheorghe, Mihaela, and Ruxandra Petre. 2014. "Integrating Data Mining Techniques
Into Telemedicine Systems". Informatica Economica 18 (1/2014): 120-130. doi:
10.12948/issn14531305/18.1.2014.11.
22. Hintz, Erin. 2017. "Thinking About Chatbots For Marketing? Consider This". Forbes.
23. Jackson, Peter. 1998. Introduction To Expert Systems. 3rd ed. Harlow [u.a.]: AddisonWesley.
24. Juan, Yi-Kai, Wan-Ying Lai, and Shen-Guan Shih. 2016. "Building Information
Modeling Acceptance And Readiness Assessment In Taiwanese Architectural
Firms". Journal Of Civil Engineering And Management 23 (3): 356-367. doi:
10.3846/13923730.2015.1128480.
25. Katz, R. and Talmi, I. (2017). Teaching Industrial Design in Mechanical Engineering:
A Laboratory for Design and Engineering. The International Journal of Engineering
Education, 33(2(A), pp.672–679.
26. King, W. and He, J. (2006). A meta-analysis of the technology acceptance
model. Information & Management, 43(6), pp.740-755.
27. Klöckner, C. and Verplanken, B. (2012). Yesterday’s habits preventing change for
tomorrow? The influence of automaticity on environmental behaviour. Environmental
psychology: An introduction, pp.197-209.
28. Kohonen, T. (1989). Self-organization and associative memory. 3rd ed. Berlin:
Springer.
29. LeCun, Y., Bengio, Y. and Hinton, G. (2015). Deep learning. Nature, 521(7553), pp.
436-444.
30. Legris, P., Ingham, J. and Collerette, P. (2003). Why do people use information
technology? A critical review of the technology acceptance model. Information &
Management, 40(3), pp.191-204.
31. Lippert, S. and Davis, M. (2006). A conceptual model integrating trust into planned
change activities to enhance technology adoption behavior. Journal of Information
Science, 32(5), pp.434-448.
32. McCarthy, J. (2007). What is Artificial Intelligence?. Stanford University.
33. McKinsey & Company. (2017). Leslie Willcocks - The next acronym you need to
know about: RPA (robotic process automation). [online] Available at: https://
www.mckinsey.com/business-functions/digital-mckinsey/our-insights/the-nextacronym-you-need-to-know-about-rpa [Accessed 28 Sep. 2017].
34. Mead, W. and Kurzweil, R. (2006). The Singularity Is near: When Humans Transcend
Biology. Foreign Affairs, 85(3), p.160.
!45
35. Meana-Llorián, D., González García, C., Pelayo G-Bustelo, C., García, V. and
Lovelle, J. (2016). IntelliSenses: Sensing the Internet of Things. Information Systems
and Technologies (CISTI).
36. Mendoza, R. (2010). Business Intelligence 2.0. International Journal of Business
Intelligence Research, 1(4), pp.63-76.
37. Parthasarathy, M. and Hampton, R. (1993). The role of piracy in the diffusion of a
software product: a propositional framework. AMA Winter Educators' Conference
Proceedings.
38. Premkumar, G. and Bhattacherjee, A. (2008). Explaining information technology
usage: A test of competing models. Omega, 36(1), pp.64-75.
39. Purdy, M. and Daugherty, P. (2016). Why Artificial Intelligence is the future of
growth. Accenture.
40. Rigopoulos, G., Psarras, J. and Th. Askounis, D. (2008). Web Support
System for Group Collaborative Decisions. Journal of Applied Sciences,
8(3), pp.407-419.
41. Robinson, L., Marshall, G. and Stamps, M. (2005). Sales force use of technology:
antecedents to technology acceptance. Journal of Business Research, 58(12), pp.
1623-1631.
42. Rudzinski, W. and Panczyk, T. (2002). The Langmuirian Adsorption Kinetics
Revised: A Farewell to the XXth Century Theories?. Springer, 8(1), pp.23-34.
43. Russell, S. and Norvig, P. (2010). Artificial Intelligence: A Modern
Approach. Pearson.
44. Sebrant, A. (2017). How not to lose the future. [online] Malina.am. Available at:
http://malina.am/video/2017/8/17/andrey_sebrant [Accessed 28 Sep. 2017].
45. Snoj, B., Pisnik Korda, A. and Mumel, D. (2004). The relationships among perceived
quality, perceived risk and perceived product value. Journal of Product & Brand
Management, 13(3), pp.156-167.
46. Sumathi, S. and Bansilal (2016). Artificial neural network application for voltage
control and power flow control in power systems with UPFC. Emerging Research in
Electronics, Computer Science and Technology (ICERECT).
47. Suzuki, K. (2011). Artificial Neural Networks - Industrial and Control Engineering
Applications. InTech.
48. Veale, T. (2001). Key Ideas in Artificial Intelligence. Dublin City University.
49. Venkatesh, V. and Bala, H. (2008). Technology Acceptance Model 3 and a Research
Agenda on Interventions. Decision Sciences, 39(2), pp.273-315.
!46
50. Vieweg, H. (2012). An introduction to Mechanical Engineering: Study on the
Competitiveness of the EU Mechanical Engineering Industry. ECORYS.
51. Wang, S. (2003). Interdisciplinary Computing in Java Programming. The Springer
International Series in Engineering and Computer Science.
52. Yu, V. (1979). Antimicrobial Selection by a Computer. JAMA, 242(12), p.1279.
53. Zervos, H., Ghaffarzadeh, K. and Harrop, J. (2017). Sensors for Robotics:
Technologies, Markets and Forecasts 2017-2027. IDTechEX.
APPENDICES
Appendix 1
Survey sample. Part 1
Question
Response options
Basic information
[Includes country of work of the respondent,
company, job title, department]
How would you evaluate your company’s
readiness for Artificial Intelligence (AI)
use?
·
·
·
·
·
!47
No AI solutions exist or are under
consideration
An AI solution has been proposed and
it is being evaluated
Based on the evaluation, an AI solution
has been accepted and is being
implemented
An AI solution exists and is being used
AI failure (an AI solution has gone into
decline and has been phased out)
Provided there is one, what type(s) of AI
solution(s) is there in your company?
· Expert System/Fuzzy Expert System
· Decision Making tool
· Artificial Neural Network
· Deep Learning Platform
· Data Mining tool
· Robotic Process Automation
· Virtual Agents (e.g. chatbots)
· Natural Language Processing tool
· Other Machine Learning Platform
(please specify)
· Other (please specify)
In which functional areas are the AI
solutions used in your company?
·
·
·
·
·
·
·
·
Finance
Planning
Marketing
Sales
HR
Operations
Entire company
Other (please specify)
At which hierarchical employee levels are
the AI solutions used in your company?
·
·
·
·
·
Managing Director
Senior Management
Middle Management
Line Management
Other (please specify)
Survey sample. Part 2
Tested variable
Perceived
Usefulness
Questions
1) Using AI solutions can improve performance of the workflow in
my company
2) Using AI solutions can increase productivity of the workflow in
my company
3) AI solutions can help to accomplish tasks faster
!48
Perceived Ease
of Use
4) I expect that setting up AI solutions will not cause major
problems
5) I expect that learning to use AI solutions will not be difficult
6) I expect that AI solutions will be easy to use
Attitudes
Towards Using
7) The use of AI solutions will benefit my company
8) Using AI solutions is a good idea
9) My company is constantly tracking available AI solutions
Intention to Use
10) I am willing to test AI solutions’ capabilities on my projects
11) I would recommend other companies to start using AI solutions
Perceived Risks
12) There is a high probability of losses for our company if an AI
solution is implemented
13) There is a high chance of potential failure to using AI solutions
Supplier Support
14) It would be important for our AI solution supplier to provide
extensive on-site training for its users
15) It would be important for our AI solution supplier to provide
online training for its users
16) It would be important for our AI solution supplier to provide
training manuals and reference materials for users
Organizational
Resistance to
Change
17) My Company would be among the last to try a new technology
even if it appeared promising
18) My Company is reluctant to adopt a new technology
19) My Company finds reasons not to implement a new technology
Appendix 2
Descriptive statistics
!49
Y
Appendix 3
Statistical model
Y
Model specifications
!50
Y
Model result
Y
Regression weights
Y
!51
Standardized regression weights
Y
Residual covariances matrix
Y
!52
Standardized residual covariances matrix
Y
Variances
Y
!53
Y
Appendix 4
Useful country-specific insights from the survey
Y
Switzerland
!56
Y
Russia
[How would you evaluate your company’s readiness for Artificial Intelligence (AI) use?]
Y
!58
[Provided there is one, what type(s) of AI solutions is there in your company?]
Y
[In which functional areas are the AI solutions used in your company?]
Y
[At which hierarchical employee levels are the AI solutions used in your company?]
!59
Y
Appendix 5
List of survey participants
Company
Country
Position/department
ABB
Switzerlan
d
Business analyst
Abb
Switzerlan
d
Lean Management Expert - Financial Transformation
Eaton
Switzerlan
d
Key Account Manager
ABB
Switzerlan
d
Senior Data Analyst
Siemens AG
Switzerlan
d
Regional Coordinator, Communciations
GE
Switzerlan
d
Business analyst
ABB
Switzerlan
d
Division Controller, low voltage products
Eaton
Switzerlan
d
Project leader, operations
GE
Switzerlan
d
Operations & Direct Sourcing Leader
ABB
Switzerlan
d
Global SCM Training Excellence Manager
!60
Abb
Switzerlan
d
Human resources manager
ABB
Switzerlan
d
Supply Planner
Siemens
Switzerlan
d
Communications Manager
ABB
Switzerlan
d
IS Enterprise Architect
Abb
Switzerlan
d
Senior Business Analyst
ABB
Switzerlan
d
Business analyst
GE
Switzerlan
d
Project Manager
Siemens
Switzerlan
d
Supply Chain Manager
ABB
Switzerlan
d
Business Specialist for Supply Chain
Abb
Switzerlan
d
Customer Manager
ABB
Switzerlan
d
Group Head Quality & Supply Chain at ABB
Siemens
Switzerlan
d
Business Development Manager at Siemens Switzerland
ABB
Switzerlan
d
Global Business Development Manager
ABB
Switzerlan
d
Global Supply Chain Manager
Abb
Switzerlan
d
Head of Sales, Power Grids division
GE
Switzerlan
d
Product Manager
Abb
Switzerlan
d
Data Scientist
ABB
Switzerlan
d
Head of Indirect Materials and Services
!61
ABB
Switzerlan
d
Strategy Consultant
Abb
Switzerlan
d
Product Manager
Siemens
Switzerlan
d
Finance Transformation Manager
Abb
Switzerlan
d
Division controller
ABB
Switzerlan
d
Product manager
Siemens
Switzerlan
d
Project manager
ABB
Switzerlan
d
Business controlling
GE
Switzerlan
d
Lean Finance Management
Siemens
Switzerlan
d
Talent acquisition manager
GE
Switzerlan
d
HR manager
Abb
Switzerlan
d
Project leader
ABB
Switzerlan
d
Account manager
ABB
Switzerlan
d
HR consultant
GE
Switzerlan
d
Account manager
ABB
Switzerlan
d
IS architect
Siemens
Switzerlan
d
Project manager
Abb
Switzerlan
d
Senior data analysts
GE
Switzerlan
d
Account manager
!62
ABB
Switzerlan
d
Business analyst
Siemens
Switzerlan
d
Sales manager
Abb
Switzerlan
d
Product Manager
Eaton
Switzerlan
d
Account manager
ABB
Switzerlan
d
Head of government relations
ABB
Switzerlan
d
Product Manager
General Electric
Switzerlan
d
Sales manager
ABB
Switzerlan
d
Business Developer
Шнейдер
Электрик
Russia
Менеджер проектов. Департамент по решениям и
проектам.
SIEMENS AG
Russia
Customer Service, Automotive Service Manager
Силовые
Машины
Russia
Руководитель продаж ключевым клиентам
Russia
Старший менеджер по тепломеханическому
оборудованию
Силовые
Машины
Russia
Начальник службы технологических систем
управления
АББ
Russia
менеджер по внедрению проектов
силовые
машины
Russia
проектный менеджер
Шнейдер
Russia
Менеджер по развитию бизнеса
Силовые
Машины
Глобал Электро Russia
Директор
Сименс
Russia
Менеджер отдела закупок
Турбоатом
Russia
Старший менеджер, отдел по работе с персоналом
Шнейдер
Электрик
Russia
Руководитель проектов
!63
Шнейдер
Электрик
Russia
Руководитель проектов
Шнейдер
Электрик
Russia
проектный менеджер
Шнейдер
Электрик
Russia
Руководитель проектов
АББ
Russia
Менеджер отдела продаж
Силовые
Машины
Russia
Начальник отдела автоматизации
Шнейдер
Электрик
Russia
Начальник отдела управления персоналом
АББ
Russia
Старший менеджер по продажам
Шнейдер
Электрик
Russia
Менеджер проектов
Шнейдер
Russia
Менеджер проекта
Шнейдер
Электрик
Russia
Менеджер по развитию бизнеса
Шнейдер
Russia
Специалист отдела продаж
АББ
Russia
Руководитель проектов
АББ
Russia
Финансовый аналитик
АББ
Russia
Финансовый аналитик
Силовые
Машины
Russia
Руководитель проектов
Силовые
Машины
Russia
Финансовый аналитик
АББ
Russia
Проектный менеджер
Силовые
Машины
Russia
Специалист по продажам
АББ
Russia
Финансовый аналитик
АББ
Russia
Менеджер по работе с персоналом
Шнейдер
Электрик
Russia
Специалист по продажам
ABB
Russia
Key account manager
Шнейдер
Russia
Специалист транспортного отдела
!64
Шнейдер
Электрик
Russia
Менеджер отдела продаж
АББ
Russia
Специалист по продажам
Абб
Russia
Старший специалист по продажам
Силовые
Машины
Russia
Директор по работе с корпоративными клиентами
АББ
Russia
Специалист по продажам
АББ
Russia
Руководитель проектов
Силовые
Машины
Russia
Специалист по продажам
Силовые
Машины
Russia
менеджер, отдел продаж
Зименс
Russia
Специалист отдела продаж
Силовые
Машины
Russia
Менеджер по продажам
Шнейдер
Электрик
Russia
Специалист по закупкам
Силовые
Машины
Russia
Начальник производства
Силовые
Машины
Russia
Специалист по продажам
!65
Отзывы:
Авторизуйтесь, чтобы оставить отзыв