Investigating AI's Role in Claims Management: Methodology Report

This report presents the research methodology employed in a qualitative study investigating the application of Artificial Intelligence (AI) in claims management within the insurance sector. The study aims to identify key factors of successful claims management, explore AI's potential, and a...Read More

Chapter 3: Methodology
3.1 Introduction
In the previous chapter, the literature regarding good claims management and the implication
of Artificial intelligence in insurance sector specifically in claims management has been
discussed. This research is an exploratory study regarding the current and potential use of
artificial intelligence in achieving the improved claims management. As also previously
discussed in introduction chapter, the objectives of this research are:
 To identify key factors of successful claims management process in insurance
 Critically explore the potential of Artificial Intelligence in insurance specifically in
claims management process.
 Barriers and challenges to implement Artificial intelligence in Claims management
process by comparing the theoretical potential to its actual potential
 To make recommendation to overcome those challenges and how it improves the
Claims Management Process.
By reviewing the literature, the theoretical aspect of first, second and third objectives have
been discussed. However, to understand the practical aspect of these objectives for example
barriers and challenges that claim management face in real world to implement AI in its
process and the actual potential of AI in claims management needs further research and data
collection from insurance professionals and by knowing from their practical experience in
claims and the use of Artificial Intelligence.
The research methodology an important part of the research process as it allows to collect the
information that is required to find out the answers of research questions. A research
methodology should be planned according to the questions that need to be answered. For the
purpose of this research Saunders, Lewis and Thornhill, (2007) research onion is adopted as
shown in the figure. Beginning with the research philosophy, this chapter will provide the
selection of research approach, research strategy, data collection method and the framework
of data analysis. For this research these components are further discussed below. This is a
qualitative research.

Figure 3.1: Research methodology model (Saunders et al., 2016)
3.2 Philosophical Paradigm
Research philosophy is essential part of any research. The quality of research will be affected
if the philosophy of research is not understood properly (Bahari, 2010). The philosophical
paradigm is the way to understand the reality of the world and studying it. It is also about the
knowledge and how to attain that knowledge. Therefore, it ultimately informs about the
research approach (Abdul and Alharthi, 2016). Ontology on the other hand is the “nature of
our beliefs about reality” (Richards, 2003). According to ontology in qualitative research
there are multiple realities on our society about one thing. It is subjected because it depends
upon the experience of about the reality.
Research philosophy in social sciences is associated with the enhancing knowledge and its
nature in the social world. Therefore, epistemological philosophical assumption is acquired
for this research. Epistemological assumption is associated with the nature of knowledge and
the method to obtain that knowledge. Since the aim of this research is to analyse the value of
Artificial intelligence (AI) in achieving improved claims management process and to address
this aim, primary data is collected by conducting interviews to gain the understanding and

knowledge of the topic area. Since epistemological assumption is about the nature of
knowledge and how to acquire that knowledge. Therefore, epistemological assumption
however, there are two types of epistemologies interpretivism and positivism.
Positivism assumes that reality is independent phenomenon and it is context free. It is
independent of humans and work on the laws of nature i-e cause-effect relationship and can
be predicted the same in future (Abdul and Alharthi, 2016). Conversely, interpretivism on the
other hand rejects this notion and instead of believing in one universal reality that cannot be
affected by humans, it assumes that there are multiples realities in our societies and is
mediated by our senses. Therefore, it is subjective. people have multiple realities depending
upon how they perceive their environment.
For the purpose of this research it is important to know about the perspectives of insurers to
understand the actual potential if AI in claims management. Therefore, this research adopts
interpretivism philosophy.
3.3 Research Approach
Saunders, Lewis and Thornhill (2007) explained two research approaches inductive approach
and deductive approach. Inductive approach is from specific observation to generalizing the
theory. It is also called bottom up approach. For qualitative research inductive approach is
acquired (Saunders, Lewis and Thornhill, 2007). Since, this dissertation is exploratory
research and qualitative in nature with the aim to analyse the value of AI in achieving
improved claims management process. Interviews are conducted to know the insights and
discovering challenges to implement AI in claims and to explore the potential benefits and
risks of AI in creating value. Therefore, inductive approach is acquired for this dissertation.
As it allows to look in the pattern and trend in an insurance industry and adoption of AI in
their systems and analysing those patterns, that can be generalised later.
In contrast deductive approach is the top down approach which starts from theory to
hypotheses to data to add to or contradict with the theory. Deductive approach is used in
quantitative study as the intent is to test the theory to search for evidences in support or
against that theory (Karen, 2010). This approach contradicts with the purpose of this
dissertation.
3.4 Research Strategy
The choice of research strategy depends upon the research aim. Since the aim of this
dissertation is to analyse the value of Artificial Intelligence (AI) in achieving improved
claims management process which is a specific area in insurance sector and needs research in

depth. For this purpose, case study strategy is acquired. Yin (2013) refers to the term case
study as an event, an entity, an individual or a unit of analysis. case study helps to understand
the problems in real world which is one of the main objectives of this research i-e comparing
a theoretical potential of Artificial Intelligence in implementing it in claims process with its
actual potential in real world. This strategy enables to understand the complex real-world
problems by collecting multiple evidences form the area of research. One of the main
advantages of this strategy is that the emergent and recent issues can be analysed in the area
of interest from the ever changing and ever evolving organisational activities (Noor, 2008).
Since the technology is always getting advanced and so the processes depending upon these
technologies .Similarly in insurance Industry it is important to understand the contemporary
issues related to the role of AI in the claims processes and how the technology is affecting
claims management in creating value and increasing the efficiency in managing claims . One
objective of this research is to explore the potential risks and benefits of AI in creating value
to the customers. The adoption of case study research strategy for this dissertation allowed
extracting the value information i-e risks and benefits of AI in claims management process
(Noor, 2008).
Another benefit of using case study strategy is that it allows generalising the findings when
multiple cases replicate the results. This dissertation has inductive research approach and
primary purpose of inductive research is to create comprehension and meaning from raw data
and its outcome is generalised theory or a model (Thomas, 2006). Hence, the case study
strategy perfectly suits the inductive approach of this research.
3.5 Research Method
Research method is based upon the finding the answers of the research questions. There are
different methods of formulating the research; however, there are two broad methods to
collect data and interpret it is qualitative and quantitative. The elementary method of research
is quantitative which includes numbers and measuring variables however, recently the
qualitative method is getting momentum among researchers (Kalra, Pathak and Jena, 2013).
In order to address objectives of this research qualitative method was used. Qualitative
research is interpretive in nature and collects non numerical data (Johnston and Vanderstoep,
2009).
Initially qualitative research was used in psychological studies where it was very difficult to
analyse the human behaviours in numeric forms. Qualitative research is used to analyse
human behaviours and generates nonnumeric data. It is used to understand people’s

behaviours, beleifs, their experiences and attitudes (Kalra, Pathak and Jena, 2013). Therefore,
it can be said that this method allows analysing the data more extensively. Therefore for the
purpose of this research is qualitative and is used to uncover the opinions, experience and is
insurers working in claims management and explore the current and potential use of AI in
claims management process which is the aim of this research.
3.6 Sampling
Sampling is the subset of population on interest for collecting data and making observations.
We cannot study the whole population because of time and cost constraints that is why the
sample is selected which represent the population of interest which will further be used for
analysis (Research Methods for the Social Sciences, 2019). Whether the research is
qualitative or quantitative, the sample is taken from the population. There are many ways of
sampling including random sampling, simple sampling, stratified sampling, cluster sampling,
systematic sampling, quota sampling and convenience sampling (Biggam, 2018). For the
purpose of this research simple random sampling is used because in simple random sampling,
every sample has an equal chance to be selected which reduces the selection bias
(Shantikumar, 2018).
This dissertation adopts simple random samples of claims professionals from insurance
industry who are using Artificial Intelligence in their claims management process. The reason
to choose this approach is because in simple random sampling every participant has an equal
chance of being selected. This small random portion of claims professionals represents the
entire population as each member has equal chance to be chosen as a part of sample. The
reason behind this approach is that it reduces the biases during selection and should represent
the entire population. As all participants chosen are claims professionals and have same
nature of work. Also, they all are already using Artificial Intelligence in managing insurance
claims. Moreover, this approach is also cost effective and less time consuming (Gravetter and
Forzano, 2018).
3.7 Data Collection method
3.7.1 Interview
Interview is considered as the most personal form of data collection as the interviewee is in
direct contact with the researcher. Since, the qualitative study focuses on understanding the
research problem in more humanistic way to find a solution (Kalra, Pathak and Jena, 2013).

For qualitative approach, interview is a common method to collect data. For the purpose of
this research one to one interview method is used. However, considering the time limits and
other resource limitations, use of technology is considered the most appropriate way to do the
interviews. Salmons (2016) elaborated the use of technology for data collection and
conducting online interviews due to geographical and financial constraints. The evolution of
technology has made the communication rapidly available for everyone and the method of
data collection has also been evolving. The answers of the research questions can be
generated in audio, video or in written form. Now online life is similar to offline life as the
online life is as real as offline life (Salmons, 2016).
Online interviews refer to the interviews conducted by Information and Communication
technologies (ICT) which could be synchronously or asynchronously. Synchronous interview
focuses on Skype, video conferencing and instant messenger (Janghorban, Roudsari and
Taghipour, 2014). Synchronous method is typically mixed with verbal and non-verbal
mediums like telephone or chatrooms or skype (Salmons, 2016).
Asynchronous method is another method used for qualitative research. Here the information
is shared via email. It is cost effective and reduces the time limitations. This method is
arguably less expensive than any other online method (Ratislavová and Ratislav, 2014). It is
also very effective in the situation where researcher and participant cannot agree on the same
time to meet online and there are limitations in planning the same time for researcher and
participant. This method is also very effective where there is important to maintain social
distancing, or the participant is reluctant or shy to come forward for face to face interview.
For this dissertation asynchronous interview method is used as this research is conducted
during the time when there were global health emergency and people must have stay home
and work from home and also maintain social distancing. Initially the strategy was to conduct
face to face interviews but the new situation has forced to redesign the data collection method
as it was too late to do face to face interview and participants were also reluctant to
communicate at the same time of participant due to their health conditions and new routines
of work. However, the challenge with asynchronous interview is that there are less chances to
build trust and environment of openness with the participant since there is no face to face
interaction and this makes difficult for researcher to ask follow-up questions (Ratislavová and
Ratislav, 2014).
For this research five email interviews were conducted from claims management
professionals. Mr.Khawaja Ghulam Wajahat is working at AXA Insurance in London.
Currently he is Underwriting Governance Manager but previously has been working in

claims management. He is highly skilled professional with fifteen years of experience in
claims and underwriting. Other two participants are Mr. Anthony Byars and his colleague
both are Claims Handler at Zurich Insurance company in Glasgow Scotland from almost two
years where they are using Artificial Intelligence in handling motor claims. Fourth participant
is Mr. Martin Nilaus Jørgensen. He is Director at Sedgwick Leif Hansen A / S. The company
provides risk assessment, claims handling, loss adjusting services especially for complex
claims like industry, cyber, auto, property etc. Fifth member Mr. Stefan Bo Gravang is also
from Sedgwick Leif Hansen A / S. Currently he is working as a Chief Commercial Officer of
the company and has been leading the Building Claims and Claims procurement.
The emails were sent to the mentioned participants on 27th March 2020 and fortunately
received prompt and positive responses. The new approach of asynchronous email interview
actually impacted the original method of face to face interview in a more positive way as the
participants were very prompt in responding the emails and in depth information is acquired
because the participant responded at his own convenient time and answered the question with
more details. Since the interview was not face to face and time bound, the participants were
more focused, relaxed and less reserved while answering the interview questions which
increased the richness and quality of data. The email interview was not time bound so they
had more time to think and consider their answer. Therefore, more clarity and in-depth
information was received as compared to the original approach of face to face interview.
3.8 Data Analysis
There is no systematic method to analyse qualitative data and however analysing qualitative
data is the most important part of any research. Qualitative data analysis begins as soon as the
first piece of information is received (Kruskal and Maxwell, 1963). For this research data
received from interviews was analysed. Data analysis includes description of data and then
interpretation of data by comparing it with literature review and contrasts comparison with
other interview questions. Patterns and themes were found in the responses which were
organised to create logic. To do that a framework approach is used to analyse the interview
data. Moreover, framework analysis is mostly suitable for the research with limited time
frame (Srivastava and Bruce, 2009). Therefore, it is suitable approach for this dissertation.

S/No Themes Source of Theme Research Objective
1 High operational cost of AI  Section 5, 5.1,
5.2 of
Literature
review
 Interview data
Barriers and challenges
to implement Artificial
intelligence in Claims
management process
by comparing the
theoretical potential to
its actual potential
2 Inconsistent service delivery with
robots
3 Security issues with AI
4 Compliance government regulations
5 Requires high expertise and training
6 Human factor is important  Section 6, 6.1,
6.2, 6.3 of
Literature
review
 Interview data
Critically explore the
potential of Artificial
Intelligence
7 AI can improve customer satisfaction
8 AI can expedite whole claim process
9 Quality of AI technology
10 AI can reduce human errors
Table 3.1: Linking generated themes with source of evidence and research objectives
3.8.1 Thematic Coded Analysis
Thematic coding is considered as one the most important for of qualitative analysis which
includes identifying or recording the text passages that are linked by a common idea or
theme. This method provides opportunity for the researcher to index the collected data from
qualitative source into categories, whereas proper framework can be established based in the
thematic ideas generated from the text (Joffe, 2012). This method is commonly adopted with
research choice which includes large texts such as interview answers, open ended
questionnaire, and others. Moreover, thematic analysis can be adopted for different types of
approaches such as framework analysis, template analysis, interpretative phenomenological
analysis, and Grounded theory (Gibbs, 2007).
There are two key factors importantly considered with this particular analytical method such
as social reality and subjective meanings of participants. Moreover, the collected data from
qualitative source cannot be simply considered as the containers of meaning. There may be
multiple meanings of texts but researcher needs to identify the true meaning of data within
the research (Nowell et al., 2017). The true meaning of data can be presented in the form of
themes or sub themes depending on the understanding of researcher. However, both

categorization and theme creation shows some sort of variations during social science
researches (Vaismoradi et al., 2016).
There are six steps involved with thematic coding analysis such as familiarisation, coding,
generating themes, reviewing themes, defining themes, and writing up (Gibbs, 2007). This
research thesis has collected qualitative data and text based data from five respondents. The
face to face interviews were not possible due to Corona virus pandemic and lock downs.
Therefore, interview questions were converted into open ended questionnaire to collect the
data.
3.9 Collected data
This section adopts the first step of data familiarisation used with thematic framework,
whereas coding from qualitative data was done within open ended questionnaires (See
Appendix B). The open ended questionnaire approach helped in collecting the data from five
respondents from different insurance companies. The collected data was in very raw format
which included large textual answers from respondents. There were 7 open ended questions
with sub questions were asked from respondents which were set based on research objectives
and aim.
3.10 Ethical Consideration
It is important to adhere with the ethical principles during the research. Ethical approval is
required in research where there is human participation involved. Ethical approval ensures
that the safety, rights and wellbeing of the participants are considered as an important factor
during the research. Ethical issues that can arise during the research should be the most
important concern of the researcher (Bell, Bryman and Harley, 2019). Ethical consideration
is an integral part of any research which ensures the researcher that ethical risks are
minimised as failure to meet ethical requirements can place the researcher in vulnerable
situation (Bell, Bryman and Harley, 2019).
For the purpose of this research, anonymity has been offered to the participants and
participants are asked to fill the consent form which described that their participation is
voluntary, and they can withdraw from the research at any time. Moreover, according to Data
Protection Act 2018, the entire participant provided data has been kept confidential
(Government Digital Service, 2011)

3.11 Limitation
One main limitation to this research was time constraint. Due to time limitation in-depth
interviews could not be done to analyse the value of AI in achieving improved claims
management.
Another major limitation was this research was conducted during the times when there was
global health emergency and people were advised to keep social distancing and work from
home. So it was difficult to conduct in-depth interviews as participants were not approachable
due to their new routines to work from home.
3.12 Conclusion
This chapter provides the explanation of the research methodology in terms of research
approach and strategy for this study. The chapter explains the data collection method that was
carried out through the interviews and also highlights the limitations of this empirical
research.