An explorative research about the applicability of chatbots at the FM service desk in offices

(1)

Master Facility & Real Estate Management

Title assignment : Master Thesis Name module/course code : BUIL 1230

Name Tutor : Mr. M. van den Hoop

Name student : Mr. L. Cornax Full-time / Part-time : Full-time Greenwich student nr. : 001006520 Saxion student nr. : 434465

Academic year : 2018-2019

Date : 20-12-2018

(2)

(3)

An explorative research about the applicability of

chatbots at the FM service desk in offices

(4)

4

Management summary

A lot of trends and developments are happening within facility management (FM). From using cleaning and caring robots till changing the workplace to a wellness destination (Sodexo, 2017; CBRE, 2017). Another trend involves around buildings, which are becoming more intelligent than their users (ISS, 2013) with the use of smart sensors (Facto, 2018). Facility management is currently applying these trends to satisfy end-users, the customers of facility management. The general change that appears from all these trends is that personal service is changing into self-service with less human contact from an FM staff member (ISS, n.d.). A trend that started more than twenty years ago was the rise of Facility Management Information System (FMIS) software. This software provides service information, based on the offered facilities that help the operational processes and the supervision by the

management team (Maas & Pleunis, 2006). Nowadays it is possible to fill in Complaints, Wishes, Information, and Faults (CWIF) in FMIS-software on a computer, tablet or smartphone, in a web browser or in an application (Axxerion, 2014).

Another trend that has been acknowledged by many commercial organisations, is the use of chatbots for client contact. They are used by organisations such as, insurance companies (Nguyen, 2017) and web shops (Kokoszka, 2018). According to Allianz Worldwide Partners (2018), chatbots are used to enhance the customer experience and are active in social applications, such as Facebook Messenger or WhatsApp, and also in applications provided by hotels. According to a market study from the researcher, chatbots are currently not applied within facility management. The increasing amount of chatbots, the increasing use of self-service, and the increasing usage of mobiles phones could create cold robotic environments and experiences (Weissenberg & Langford, 2018). Nevertheless, the usage of mobile phones will likely increase even more in the near future and chatbots may give organisations the opportunity to further digitalise the way how services or information are requested and eventually the way how they are being delivered. From a facility management perspective, the combination of facility management and chatbots may improve the satisfaction of end-users by enabling them to give complaints, wishes, information, and faults via a chatbot, besides common communication channels. End-users do not have to be dependent on opening hours of offices, personnel costs could be lower since less personnel is needed, and the risk of communication faults can possibly be lower. Therefore, the following main research question was developed: “What is the applicability of chatbots in facility management within the service desk in offices?”. The research objective was to gain insight into the applicability of chatbots in facility management, especially within the service desk in offices. The data was collected by performing focus group interviews and semi-structured interviews with chatbot developers, managers from organisations who use chatbots in their communication with their external customers (consumers and organisations), with people who are active in facility management (FMIS-suppliers, consultancy, FM-suppliers), and with facility managers who could add a chatbot to their communication channels. The results show that chatbots can be added to the communication channels to receive and react 24/7 on CWIF-process requests and reports from end-users. Facility management related example situations, of when a chatbot can be used, relate to space & building management, catering, cleaning, and facility management related administration. The CWIF-process related requests and reports can be handled by a chatbot to as long as they involve a low number of steps, involve simple processes with a high frequency and do not involve emotionality, urgency, complexity, or individuality. For example, questions on FAQ lists or standardized forms for booking a meeting room. Adding a chatbot to communication channels can save costs, increase self-service, save time, and can increase the satisfaction of end-users. As a result, employees of the service desk can focus on CWIF-processes that need personal attention were the level of individuality, complexity, and emotionality are high, which increases the satisfaction of service desk employees. Also, chatbots can provide a more human conversation in comparison to self-service portals and can have a

connection to FMIS-software and other software that relates to for example HR or finance.

It is recommended to consider adding a chatbot to the already existing communication channels since (especially young) people want to communicate easily and quickly via their mobile phone. Secondly, it is recommended for organisations that want to use a chatbot for their end-users of their building to set-up a pilot with end-users to test and develop a chatbot that is able to receive and react on facility management related CWIF-process requests and reports. For future research is it recommended to test the findings of this research in practice by performing quantitative research. This can be done in an office environment where end-users can communicate requests and reports of facility management related CWIF-processes via a chatbot. The experiences of end-users can be measured by making use of a questionnaire.

(5)

Foreword

In April 2018 I started my Master Thesis in order to receive my degree in the Master of Facility & Real Estate (MFREM) from the University of Greenwich and Saxion University of Applied Science. During this research, I investigated the applicability of chatbots in facility management, while focusing on the service desk of the facility management department in offices.

I would like to thank all respondents from both the two focus groups and semi-structured interviews for their input, time, and effort. I would also like to thank Mr. M. van den Hoop for his input and helpful advice as an external examiner from Saxion University of Applied Sciences.

Bennekom, the Netherlands, 20th_{of December 2018} Laurens Cornax

(6)

6

1. Introduction

Nowadays a lot of trends and developments are happening within facility management. From using cleaning and caring robots till changing the workplace to a wellness destination (Sodexo, 2017; CBRE, 2017). Another trend is that buildings are becoming more intelligent than their users (ISS, 2013) with the use of smart sensors (Facto, 2018). Facility management is currently implying these trends to satisfy end-users, their customers. The general change that appears from all these trends is that personal service is changing into self-service with less human contact from a member of staff from the facility management department (ISS, n.d.). The involved change could be noticed across a number of services from the department. For example, soft services like reception, catering, cleaning, and security (BIFM, 2018a). According to Schmitter and Kofler (2018), these and other services could be delivered with the same type of service in a hospital environment. Services that they stress range from information management, signage, safety and security, to transport and logistics.

The trend of self-service will have an effect on the hospitality experience. The experience will change to a situation where end-users or guests can choose how they would like to receive the service, which sometimes enables the staff to surprise the users or guests by delivering a so-called WOW-effect (Van Soest, Hokkeling, & Buiteman, 2016). A trend that appeared more than 20 years ago was the rise of Facility Management Information System (FMIS) software which serves information, based on the offered facilities that help the operational processes and the supervision by the management team (Maas & Pleunis, 2006). Nowadays it is possible to fill in complaints, wishes, information, and faults in FMIS-software on a computer, tablet or smartphone in a web browser or in an application (Axxerion, 2014). The current trends in FMIS-software are the integration of assets that gather data based on usage, the upcoming platforms to share FMIS-software data with different stakeholders (end-users, suppliers, communicating assets, and the facility manager), and the seamless customer experience between physical and digital communication (Louws, 2017). Other trends that are recognised by a large FMIS-software supplier are the integration of building information modelling (BIM), Human Relations (HR), and finance software. The supplier also recognised the increasing usage of mobile phones and expects the next generation software will enable employees to always have access to all services that they need from their mobile devices. Sensor technology will support users to find and book rooms that fit their needs, and entrance and safety systems will be integrated into FMIS-software in the near future (Planon, n.d.).

Another trend that has been acknowledged by many commercial organisations, is the use of chatbots for client contact. They are used by organisations such as, insurance companies (Nguyen, 2017) and web shops (Kokoszka, 2018). According to Allianz Worldwide Partners (2018), chatbots are used to enhance the customer experience and are active in social applications, such as Facebook Messenger or WhatsApp, and also in applications provided by hotels. According to a market study from the researcher, chatbots are currently not applied within facility management. English Oxford Dictionaries (2018a) defines the word “chatbot” as: “A computer program designed to simulate conversation with human users, especially over the Internet”. Radde (2017) states that an intelligent chatbot is equipped with Artificial Intelligence (AI) to be able to have conversations with humans. English Oxford

Dictionaries (2018b) defines AI as: “The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages”.AI can, therefore, be seen as the brain of an intelligent chatbot. The possible impacts of chatbots are that the face-to-face communication between people could decrease and that they possibly value face-to-face communication higher than normally. Secondly, somefunctions will possibly be replaced by technology.

The increasing amount of chatbots, the increasing use of self-service, and the increasing usage of mobiles phones could create cold robotic environments and experiences (Weissenberg & Langford, 2018). Nevertheless, the usage of mobile phones will possibly increase further in the near future and chatbots could give organisations the opportunity to further digitalise the way how services or information are requested and eventually the way how they are being delivered. From a facility management perspective, the combination of facility management and chatbots may improve the satisfaction of end-users by enabling them to give complaints, wishes, information, and faults via a chatbot, besides common communication channels. End-users could be independent on opening hours of offices, personnel costs could be lower since less personnel is needed, and the risk of faults in communication could decrease.

(9)

Combining the trend of self-service and the trend of chatbots, it is interesting to research the applicability of intelligent chatbots within FM. The applicability could decrease the upcoming gap of personal service and it could decrease robotic environments that are being developed form the usage of mobile phones. Therefore, the main research question has been created to gain insight into the applicability of chatbots in facility management within the service desk in offices, since the service desk receives usually complaints, wishes, information requests, and faults and a lot of these processes could possibly also be sent via a chatbot besides common communication channels. The main research question is:

(10)

10

2. Literature review

This chapter can be seen as the backbone of this research, the theoretical foundation. The topics that are briefly mentioned in the discussion will be elaborated in the paragraphs below accordingly. The main topics in this chapter are: communication, facility management, and chatbots. The last-mentioned topic is mostly based on non-academic sources because academic sources could not be found. The conceptual model is discussed at the end of this chapter. The code trees that are based on the literature can be found in appendix 1A until 3B. An explanation about how these code trees are created can be found in paragraph 4.4.

2.1 Communication

English Oxford Dictionaries (2018c), defines communication as: “The imparting or exchanging of information by speaking, writing, or using some other medium”. While Merriam-Webster (2018) defines communication as: “a process by which information is exchanged between individuals through a common system of symbols, signs, or behaviour”. Communication has changed over the years, from the first printing press in 1450, the first telephone in 1876, the first e-mail in 1971, the first mobile phone and fax in 1973 to the first personal computer in 1975 (Northwest University Information Technology Office, 2018). Shannon and Weaver (1949) created one of the first models that describe the basic process of communication; the Transmission model, which is also known as the “Shannon and Weaver model of communication” (University of Twente, 2017).

The model includes different elements; the Information Source is from the sender who encodes the message to signals. Next, a channel is chosen by the sender to which signals are adapted. In this third element, noise can disrupt the message. Finally, the receiver decodes the message and tries to understand what the message is (Shannon & Weaver, 1949). The model is widely used in studies (Jabri, Allyson, & Boje, 2008; Catasús, Mårtensson, & Skoog, 2009; Wilks, 2016). However, the model has received critique about the fact that a distinction cannot be made between a successful

transmission of nonsensical words and a successful transmission of meaningful sentences (Bowman & Targowski, 1987). Although the high age of the model, it is still used in studies and more recent models are built on this model, said Hollnagel and Woods (2005). (Hollnagel & Woods, 2005). The researcher has chosen this model because it points only out the most important aspects of

communication, which is the main strength of this model. Figure 1 below shows the Transmission model.

Another model that describes the basic process of communication was developed by Lasswell (1948). Lasswell stated that to understand the process of communication, you need to question yourself the 5 W’s questions, which have their own element. All 5 W’s are placed in table 3 below.

The model from Lasswell provides a scientific method to distinguish different elements from the communication process. The focus point of this model is the so-called transmission vision, which focuses on transferring knowledge and information. The model is also known as “one-way model of communication”. Through the years, the model has received critique about its focus point towards the desired effect of the sender. Critics also stated that the model does not include the possibility of noise and only includes basic elements (Businesstopia, n.d.). This last point of critique can be seen as the strength of the model, which is also the main point of why the researcher has chosen to include this model in this research.

(11)

The difference between the models is that Lasswell’s model of communication does not include noise, the communicator is mentioned first, while the so-called transmitter is second in the Transmission model, and the Transmission model does not include the effect of the message that has been sent through a channel. While both of the models are essentially mentioning that a message can be sent with the use of a channel, which will be received and needs to be interpreted by the receiver to get the effect that the messenger wants. The sender will know that, once the receiver sends a message back, which is called the feedback loop (Daft, 2012). The ‘channel’ refers to communication channels. The models are chosen because they describe the basic process of communication, which helps to perform this research with an abstract focus on communication.

According to Daft (2012), Face-to-face communication is the richest channel because it permits immediate feedback and personal focus. The least rich channels are the reports or bulletins because of the one-way impersonal communication, which is shown in figure 2 below.

The different communication channels are shown in the blue coloured field. To sum up, these communication channels can be categorised into three basic sorts of communication: 1) verbal communication, which involves listening to a person to understand the meaning of the message, 2) non-verbal communication, which involves observing a person and induce a meaning, 3) written communication, which involves writing and reading a message (University of Minnesota, 2010). Unfortunately, criticism on this channel richness model from Daft could not be found.

Verbal communication is defined as: “..sharing of information between individuals by using speech” (Business Dictionary, n.d. a). Non-verbal communication is defined as: “Behaviour and elements of speech aside from the words themselves that transmit meaning” and includes: “pitch, speed, tone and volume of voice, gestures and facial expressions, body posture, stance, and proximity to the listener, eye movements and contact, and dress and appearance” (Business Dictionary, n.d. b). Lastly, written communication is defined as: “expresses thoughts and ideas clearly in writing using correct grammar, organization and structure” (Bowdoin College, n.d.).

Over the years, electronic messages have increased and include currently: Live-chat, WhatsApp, Facebook, chatbots, and applications. Live-chat is a relatively new communication channel and enables customers to directly contact an organisation, which can lead towards better client

relationships. When the customer is satisfied with the live-chat conversation, the chance that he will start a live-chat again increases. The most important condition for using live-chat is that the first response time needs to take no longer than 30 seconds. After 30 seconds of no reaction from the employee behind the live-chat, the customer stops the conversation. Other conditions are clear communication about where to find the live-chat on a website and clear opening hours (Oklopcic, 2016).

Question Element

1. Who? Communicator

2. Says What? Message

3. In Which Channel? Medium

4. To Whom? Audience

5. With What Effect? Effect

Table 3: Lasswell’s model of communication (Lasswell, 1948)

(12)

12

According to Quaden (2015), the most used communication channels in the workplace were in 2015 (from high to low popularity) e-mail, mobile phone conversations, and face-to-face communication. Due to digitalisation, chat will become a more mainstream communication channel because of the digitalisation. When introducing a new channel, the following five phases of product (as in channel) adoption need to be taken into account: 1) Awareness: people do know that the new product exists but do not have any information about it, 2) interest: people are curious enough to get information about the new product, 3) evaluation: people consider if the product has something that they could benefit from, 4) trial: people try the product on a small scale to come with a more grounded opinion, 5) adoption: people decide to try the product entirely and to use it often (Kotler, Armstrong, Saunders, & Wong, 2009). Within the introduction, people can be categorised to: 1) Innovators: people who like to try new ideas and are easy on taking risks, 2) Early adopters: people who know that their meaning is valued in their community, 3) Early majority: people who are doubters and not leaders. This category accepts new ideas fast, 4) Late majority: people who are sceptical and accepts the new idea when most of the people have tried it, 5) Laggards: people who are stragglers, who like tradition and are distrustful to change. This category accepts new ideas when these new ideas are becoming a tradition (Kotler, Armstrong, Saunders, & Wong, 2009).

Organisations use besides these above-mentioned communication channels also management information systems, which can enhance internal communication. Examples of these information systems are a Facility Management Information System (FMIS), employee information system or a financial administrative system (Centrum Facilitair, n.d. a). Information from these kinds of systems flows through-out organisations. The directions of communication within organisations are horizontal (literally) to a co-worker, vertical upwards to a supervisor, vertical downwards to a subordinate, and diagonally to a different department (University of Minnesota, 2010). The way how communication takes place is characterised by manners of how people communicate (in Dutch: Omgangsvormen), which are also called etiquettes. These manners are connected to culture and consist of verbal and non-verbal communication habits, which are sometimes unwritten. For example, the way how people greet each other or what it means when someone raises his/her eyebrow (Essink, 2016).

2.2 Facility Management

Atkin and Brooks (2014) define facility management as: “… the integrated process to support and improve the effectiveness of the primary activities of an organization by the management and delivery of agreed support services for the appropriate environment needed to achieve its changing

objectives”. BIFM (2018a) defines facility management as an “organisational function which integrates people, place, and process within the built environment with the purpose of improving the quality of life of people and the productivity of the core business.”

(13)

With the use of the Facility Management Model (see figure 3) from the Nederlands Normalisatie Instituut (2018), facility management will be further explained. The NEN-EN15221 model shows that facility management aligns the gap between the demand of an organisation and what facility

management supplies to an organisation. The supply is based on the demand that can be specified with the use of Service Level Agreements (SLAs). The supply is measured with the use of Key

Performance Indicators (KPIs). The provider of the facility services could be an internal and/or external body (BIFM, 2018b). The earlier mentioned FMIS software can help an organisation to manage the demand and supply. FMIS software can provide an organisation accurate and timely (Razali & Mana, 2007) and unambiguous information flow, which enables an organisation to manage its services efficiently and effectively (Centrum Facilitair, n.d. b). Additionally, the software can create a management dashboard and reports. The software also enables an organisation to collect and analyse the complaints, wishes, information requests, and faults (further abbreviated as: CWIF) (Facto, 2017a), which can be received by e-mail, phone, by visiting a physical desk, and even via social media (Eysink, 2016). Beckers and Roelofs (2010) describe complaints as those that relate to services or product that are provided by facility management. Wishes are described by the same writers as those to requests of products or services that are provided by facility management. Same counts for the description of information requests, which are requests about any kind of products or services that are provided by facility management. Lastly, faults are described as those who occur in facility management and that relate to every product or service that is provided by facility

management. The experience of a product or service by the end-user is important because it

influences the satisfaction about facility management. According to Williams and Naumann (2011) are the key drivers of customer satisfaction: “(1.) quality of relationship, (2.) establishing and maintaining fast, (3.) accurate two-way communication, (4.) innovativeness of products, (5.) account reps staying in touch, and (6.) performing work completely”.

According to Drion and Van Sprang (2012), attention points are important in the implementation phase. For example, the work processes need to be order and there needs to be support for FMIS software throughout the involved departments of an organisation. Besides facility management, more and more processes are being added to FMIS software. For example, HR or finance (Centrum Facilitair, n.d. b). According to Facto (2017b), FMIS software is changing to IWMS (Integrated Workplace Management System), which bundles the different processes into one environment. A principle that is widely used in facility management is the Click-Call-Face Principle, which can be implemented by an organisation to steer the CWIF-processes to the right communication channel that could be connected to the service desk of facility management. ‘Click’ often refers to the virtual service desk, like FMIS software. ‘Call’ refers to a phone call and ‘Face’ refers to a physical desk. The thought behind this principle is that people can do a lot by themselves by searching for information instead of calling or visiting a desk (Oud, 2013).

2.3 Chatbots

As mentioned in the introduction, chatbots are, according to English Oxford Dictionaries (2018a), defined as: “A computer program designed to simulate conversation with human users, especially over the Internet”. Radde (2017) states that a chatbot is equipped with Artificial Intelligence (AI) to be able to have conversations with humans. English Oxford Dictionaries (2018b) defines AI as: “The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.” According to Schlicht (2016) from Chatbots Magazine, a chatbot is a service, which is powered by rules and often artificial intelligence, that runs on a social chat product/application. For example, Facebook Messenger or Telegram. Examples of chatbots are: weather bot, grocery bot, news bot, scheduling bot, life advice bot, or personal finance bot (Schlicht, 2016). According to Boztas and Hadwick (2017), writers of the report “Are bots worth the bother?”, chatbots are applied in many businesses in the tourism sector, such as airlines, hotels, and airports. Speaking of hotels, according to Bhowmik (2017) are chatbots used for customer service by large hotel groups, such as Marriott, Hyatt, and Starwood. There are a number of positive impacts of chatbots for hotels. For example, hotel guests can stay connected via a chatbot from the moment of pre-arrival to post-departure, which increases the number of repeated visits and the so-called guest loyalty. Besides the tourism sector are chatbots also used by insurance companies for client contact (Kokoszka, 2018).

(14)

14

According to Godara (2018), the technology of chatbots began in 1966 by MIT professor Joseph Weizenbaum who created an early natural language processing computer program, Eliza.

Weizenbaum created the program to mimic human conversations by matching pre-scripted responses to user prompts. Eliza was the first ever chatbot that was able to interact with people. Weizenbaum used a script that was able to recognize certain keywords and patterns and generate a response sufficiently. In the years after Eliza, many chatbots have been developed. For example, Parry (1972) for persons with paranoid schizophrenia, Jabberwacky (1988) for entertainment, Alice (1995), the Microsoft paperclip Clippy (1996), Smarterchild (2001) for entertainment via MSN Messenger, Watson (2006) created by IBM for entertainment, Siri (2008) created by Apple, Cortana (2015) create by Microsoft, Alexa (2015) created by Amazon, and Tay (2016) created by Microsoft (WhosOn, 2018). Lastly, in 2016 Facebook made it for companies possible to use chatbots in Facebook Messenger. Eventually, more than 34,000 bots were available at the end of 2016 covering a wide range of use cases (Wizu, 2018). Schlicht (2016) stated that businesses can use chatbots to connect with potential customers by using social media applications because that is what people use the most on their mobile phone.

According to Schlicht (2016) are there basically two types of chatbots. The first type of chatbot is based on certain rules, which limits the bot. As a result, it can only respond to specific scripted commands. According to Boztas and Hadwick (2017) include these commands: keywords, e-mails, dates, and phone numbers. The answers are based on developed scripts. An advantage is that these chatbots are very precise and the developer is able to subtract rules to be able to handle new

situations. The second type uses machine learning. The “brain” of this type of bot is artificial intelligence (AI), which means that this type of chatbot can communicate because it understands language, speech recognition and it is able to make decisions and to translate between languages. It can get smarter because it learns from previous conversations (Schlicht, 2016). Building this type of chatbot takes more time and effort. Over time, the chatbot will be able to find suitable answers and it will be less confused by incomplete or missing information that the messenger sends (Boztas & Hadwick, 2017). Information about the common functional and technical requirements and the once that determine the user-friendliness of chatbots could not be found in publications, possibly because of the newness of chatbots. Merriam-Webster (n.d. b) defines user-friendly as: “easy to learn, use, understand, or deal with”. Technical IT information about how chatbots work will not be discussed because this information is out of the FREM (Facility & Real Estate Management) scope.

From the elaboration above, the benefits of chatbots are clearly that chatbots can service humans without a directly involved member of staff, can perform multiple services which involve keywords, e-mails, dates, and phone numbers. On the other hand, the disadvantages of chatbots are according to Dickinson (2017) is the complexity of language. He stated that words have different meanings in different contexts and situations. As a result, having an artificial intelligence chatbot that totally understands guests is very challenging. As an attention point, Dickinson stresses that guests need to understand that they need to use simple direct requests.

(15)

2.4 Conceptual model

The following model has been developed from the literature review. End-users do have a need for information; they want a solution for a complaint, communicate their wish, request information, or they want to report a fault. More information about CWIF-processes can be found in paragraph 2.2 “Facility Management”. End-users can communicate their complaints, wishes, requests for information, or faults through communication channels (telephone, e-mail, etc.), which influences the experience of the service provider (FM-department) of an organisation. This is the reason why the arrow from "Communication channels" points downwards on the other arrow line. The question is: will the

relationship between the end-user and the service provider increase, decrease or stay the same when a chatbot has been added to the communication channels. Any kind of answer will be from great value for the main research question: “What is the applicability of chatbots in facility management within the

service desk in offices?” The sub-questions to answer this question are explained in paragraph 3.3 on

the next page.

Figure 4: Conceptual model

Experience of the service provider from

the end-user Need for information

from the end-user

Communication channels

(16)

16

3. Objective and questions

This chapter discusses the research objective, the main research question, and the sub-questions to answer the main research question. This chapter ends with a break-down structure.

3.1 Research objective

Companies put a lot of effort into digitalising their processes and their systems. Chatbots are new to the common communication channels and are currently used by for example insurance companies and webshops for their communication with external customers. Information about the usage of chatbots for internal customers could not be found, especially not within facility management where complaints, wishes, information requests, and faults can be received by the service desk of facility management. Therefore, the objective of this research is to gain insight into the applicability of chatbots in facility management within the service desk in offices.

3.2 Main research question

What is the applicability of chatbots in facility management within the service desk in offices?

3.3 Sub-questions

The following sub-questions are created to help to answer the main research question.

Sub-question 1

What is the impact of chatbots on the communication between end-users and the service desk?

This sub-question is created to find out which impact chatbots can have on communication between end-users and the service desk. According to the literature review, chatbots can have a positive impact on client contact in other businesses, such as insurance companies and web shops.

Therefore, it is important to research the impact in the context of the applicability of chatbots in facility management. When looking at the conceptual model from paragraph 2.4, this question thrives to find an answer about the intervening variable “Communication channels” on the other two values of the model.

Sub-question 2

When is a chatbot user-friendly for the end-user?

Sub-question 2 is created to find out when a chatbot can be seen as user-friendly for the end-user in the field of facility management because information about the user-friendliness of chatbots could not be found by the researcher while writing the literature review. User-friendliness of chatbots is important because it possibly has an effect on the applicability and the success of chatbots. Same as sub-question 1, this sub-question thrives to find an answer about the intervening variable “Communication channels” on the other two values of the model.

Sub-question 3

Which CWIF-processes can be communicated via chatbots to the FM-department?

In order to apply chatbots in the communication of CWIF-processes between end-users and an FM-department, it is important to know for which of the many CWIF-processes a chatbot can be used. Therefore, this sub-question was created. When looking at the conceptual model from paragraph 2.4, this question thrives to find an answer about the variable “Need for information”.

3.4 Breakdown structure

Figure 5: Breakdown structure MRQ: What is the

applicability of chatbots in facility management within the service desk in offices?

SQ 1: What is the impact of chatbots on the communication between

end-users and the service desk?

SQ 2: When is a chatbot user-friendly for the end-user?

SQ 3: Which CWIF-processes can be communicated via chatbots to the

(17)

4. Research methodology

This chapter discusses the design of the research, methods of data collection, operationalisation, and finally the methods of data analysis.

4.1 Research design

This is a qualitative and an exploratory research because chatbots are currently not applied within facility management departments at this moment, according to a market study from the researcher. The reason for a qualitative study is because the research focusses on interpretation, experiences, and meanings about chatbots in facility management. According to Saunders, Lewis, and Thornhill (2007) is a qualitative research approach concentrated on measurements in words and include a specification of a certain situation. This research approach matches with the purpose of this research because the aim is to gain insight into the applicability of chatbots in facility management within the service desk in offices in order to make recommendations for the field. Since this research has an explorative character, a couple of models and theories are applied in this research. Some literature could not be found. Therefore, this research has both a deductive and inductive approach. Adrienn Eros (personal communication, 15-02-2018) stated in a lecture that regardless of the character of the research, a qualitative analysis can be deductive and inductive. How these two approaches have been used in this research is explained in paragraph 4.4.

All sub-questions have been answered by performing two focus group interviews and with ten semi-structured interviews. The focus groups interviews have been performed in order to gain insight into the thoughts and opinions from members about chatbots with the use of topics. An overview of the topics can be found in Appendix 5. The semi-structured interviews have been performed in order to gain insight into the thoughts and opinions from interviewees about chatbots. An overview of the questions can be found in Appendix 6. The members had different functions within different fields. More information about the respondents and the selection criteria can be found in the next paragraph. An overview of the respondents can be found in Appendix 4. The answers from both the

semi-structured interviews and focus group interviews have been combined to give an answer to each sub-question. More information about how data is analysed can be found in paragraph 4.4. This research design is partly the same as Kindström and Kowalkowski (2014) used for their study about “the nature and characteristics of business model elements required for successful service innovation”.

The researchers used data from several research projects and collected empirical data by performing interviews and focus groups interviews within a couple of firms. Their research method does not fully fit the topic of this research within the field. However, the methodology of performing interviews and focus group interviews does fit and has therefore been chosen to collect data from respondents with different functions and different fields because they do have experience in the applicability of chatbots from their diverse roles or functions, which increased the number of viewpoints. For example, chatbot developers. Other approaches would not fit since the field of facility management does currently not have experience in the applicability of chatbots.

(18)

18

4.2 Methods of data collection

The data collection techniques that have been used to answer all sub-questions are semi-structured interviews and focus group interviews. According to Alshenqueeti (2014), semi-structured interviews are based on a checklist with interview questions, which enables an in-depth conversation while keeping in alignment with the main/key concepts of the research. According to Hennink (2014), focus group interviews are based on topics. This type of interview enables an interactive discussion without having the pressure to reach a consensus (Krueger & Casey, 2000).

Recordings & locations

The data from semi-structured interviews have been collected by making use of an audio recording device and took face to face place in a small quiet space within the work environment of the

interviewee or on the phone. Interviewing in the work environment of the interviewee enables that the interviewee feels comfortable, which increases the chance of receiving honest answers (Saunders, Lewis, & Thornhill, 2009). Video conferencing with Skype was not used because the interviewees were not able to make Skype calls with externals from outside their organisation. The data from the focus group interviews have been collected by making use an audio recording device and a video recording device, which gave the researcher the opportunity to steer the group interview according to different topics instead of taking a lot of notes (Saunders, Lewis, & Thornhill, 2009). The focus group interviews took place in a nice-looking room with a low amount of echo in one of the buildings from of Saxion University of Applied Sciences. The way how the video and audio files have been analysed is described in paragraph 4.4.

Respondents

All respondents (interviewees of semi-structured interviews and members of focus group interviews) were selected on their function as a manager with a minimal duration of two years and needed to be working in the field of facility management, except from the chatbot developers who only need to have two years of work experience and needed to be specialised in chatbots within their organisation. The website LinkedIn.com has been used to find the respondents. The reason for these criteria was the fact that interviewees needed to have knowledge about facility management and needed to have enough experience to give sufficient information. The members of the focus group interviews had one extra criterion: all members needed to be interested in technology to increase the chance of delightful and enthusiastic group interviews. According to Krueger and Casey (2000) should this encourage the members to discuss their points of view alongside with the topics, which were based on the sub-questions.

The respondents were divided into three main categories: 1) Chatbot developer, 2) Chatbot users, and 3) Other. The last category contains different organisations with no experience with a chatbot.

The organisations from the category 3 have the following sub-labels: FMIS-supplier, Consultancy, Possible user, FM-supplier. Different types of organisations were chosen in order to receive opinions from different viewpoints/angles from the market of facility management, which had a positive influence on the reliability. Category 1 was chosen because they have the most experience with developing chatbots. The interviewees of this category were managers and client advisors who were specialised in chatbots. Category 2 was chosen because they have the most experience when it comes to working with chatbots, as a user. The interviewees from this category were responsible for the chatbot and the chatbot conversations in their organisation. The method of non-probability sampling has been used for the focus group interviews and semi-structured interviews because the total size of the populations of chatbot developers, organisations who use chatbots, organisations who are possible users of chatbots, FMIS-suppliers, consultancy organisations, FM-suppliers in the

Netherlands was not known. The total amount of respondents was aimed at twenty in order to draw conclusions. This is a small sample size, which was chosen because of the short research period. The sample size of the focus group interviews was aimed at twelve members spread across two interviews (six members per interview) but unfortunately, five members in total showed up. The

sample size of the semi-structured interviews was aimed at eight interviewees. According to Saunders, Lewis, and Thornhill (2009), focus group interviews require normally between four and seven

members. In general, interviews require between six and nine interviewees (A. Eros, personal communication, 15-02-2018). The number of people that have been interviewed is ten.

(19)

The content of the interviews

The aim of the first focus group interview was to gather thoughts and opinions by filling in two models; Empathy Map Canvas and Business Canvas model. The interviewees shared their opinions on chatbots in the communication between end-users and the service desk of facility management and gave feedback about the design of the focus group interview because not all elements from the models were filled in due to the time. A summary of focus group interview one can be found in Appendix 7. After performing the first focus group interview, the design of the focus group switched to topics that relate to the sub-questions and included small example questions. More information about the switch of the design of the focus group interview can be found in paragraph 6.3: “Constraints and limitations”. Example questions were used to discuss interesting subjects that correspond to the topics and to keep the interview going. These topics included communication between end-users and the service desk, usability, and CWIF-processes. An overview of these topics and example questions can be found in Appendix 5 and a summary of focus group interview two can be found in Appendix 8. The content of the semi-structured interviews consists of interview questions, which are partly based on the information that could not be found for chapter 2 “Literature review”. For example, the functional requirements of chatbot. The interviews questions were different across the three main categories of respondents because of the different levels of experience with chatbots. The following paragraph contains the literature that has been operationalised for an amount of the interview questions.

4.3 Operationalisation

According to Saunders, Lewis, and Thornhill (2009), operationalisation can be defined as: “the translation of concepts into tangible indicators of their existence”. Since this research is a qualitative research, it is partly based on meanings and interpretations from others and is supported by literature. This paragraph shows the operationalisation of each sub-question and is, if possible, supported by literature from chapter 2: “Literature review”.

Sub-question 1

The abstract concept “Impact of chatbots on the communication” can be described as the

organisational impact on communication. Such as, sorts of communication, communication channels, and messages. Table 4 below shows the keywords and indicators.

Table 4: Operationalisation of “Impact of chatbots on the communication”

Abstract concept Keywords Indicator

Impact of chatbots on the communication Impact Costs Self-service Time Communication channels Structure: Click-call-face principle Etiquettes Sorts of communication (University of Minnesota, 2010) Verbal communication Non-verbal communication Written communication Communication channels (Daft, 2012) Instant messaging Facebook WhatsApp Chatbots Live-chat Face-to-face Telephone E-mail Blog Twitter Messages

(Beckers & Roelofs, 2010)

Complaints Wishes

Information requests Faults

(20)

20

Sub-question 2

The abstract concept “User-friendly” is defined by Merriam-Webster (n.d. b) as “easy to learn, use, understand, or deal with”. It can be split into certain elements/requirements that might have an impact on the experience of the user. The abstract concept includes two types of concepts: functional requirements and technical requirements. Since both kind of requirements could not be found in literature and are not the same as requirements from other chat programs or applications, the researcher has thought them out by himself. Table 5 below shows the keywords and indicators. Table 5: Operationalisation of “User-friendliness”

Sub-question 3

The abstract concept “CWIF-processes” includes complaints, wishes, information requests, and faults. Since these elements can be anything, mentioning indicators would not be helpful. Therefore, the four elements are described below. Lastly, a definition of “FM-department” is mentioned.

Complaints

Complaints that occur within facility management are described as those that relate to services or product that are provided by facility management (Beckers & Roelofs, 2010)

Wishes

Wishes that occur within facility management relate to requests of products or services that are provided by facility management (Beckers & Roelofs, 2010).

Information requests

Information requests that occur within facility management are described as information requests about any kind of products or services that are provided by facility management (Beckers & Roelofs, 2010).

Faults

Faults that occur within facility management are described as those that relate to every product or service that is provided by facility management (Beckers & Roelofs, 2010).

FM-department

The word “FM-department” is used in this sub-question and has therefore been defined as the department who is responsible for all facility management related elements within an organisation (Beckers & Roelofs, 2010).

Abstract concept Keywords Indicator

User-friendly

Functional requirements

Correct answers Recognise emotion Give feeling reflections Trainable

Uses client data Ease of use

Technical requirements

Scalable Testable

Connections to different systems Support multiple languages Stable

Low response time Management dashboard

(21)

4.4 Methods of data analysis

The data from the semi-structured interviews have been analysed with both a deductive and inductive approach, based on the transcriptions of all semi-structured interviews. Besides the appendices in this master thesis, more research data can be found on the attached USB-stick. For example, the video and audio recording or the transcripts. The transcripts are placed on the USB-stick because some of the respondents shared information that could not be made publicly.

After performing and transcribing the semi-structured interviews, transcriptions have been coded with the codes from code trees that are based on information from the literature review, which is called the deductive approach. These code trees can be found in Appendix 1A until 3B. The second step was the inductive approach and included the action of placing codes by information in the transcriptions that are not included in the literature review, which is also called open coding. The coding itself was performed per interview question or per sub-question with the use of the programme Atlas.ti. According to Saunders, Lewis, and Thornhill (2009), coding ensures that the data is separated into fragments. The next step was to perform axial coding by grouping the codes from both the inductive and deductive approach. This was performed in order to find relationships between the fragments. From the moment the relationships were known, large fragments have been appointed, which is called selective coding. Axial coding and selective coding have been done with the use of a Microsoft Excel file that included all codes from coding, based on frequency. The file was created by Atlas.ti and shows the main code (including, head code, sub-code, and sub-sub-code) on the right side of the table. The rows on the left side are divided into head code and sub-code per code tree. After grouping the codes in the Excel file, codes have been grouped within Atlas.ti.

In order to write down the results per sub-question, all answers on the semi-structured interview questions have been combined into one Word document per sub-question and the Excel file has been used to find the most frequently used codes that could be used for writing the results. The main outcomes from the two focus group interviews have been summarised by using the audio and video recordings. The summaries include the main outcomes and quotations from members and can be found in Appendix 7 and Appendix 8. Eventually, the summaries have been compared to the findings from the semi-structured interviews after selective coding was performed.

Bases on the new codes from open coding, adjusted code trees have been created and can be found in Appendix 9A until 10. The new codes in the trees have the color green. Appendix 9A, 9B, and 9C include each a part of the code tree about chatbots. Appendix 10 includes the second adjusted code tree: “Other”, which includes codes about chatbots and communication and chatbots and facility management. Sub-sub-codes are not included in the adjusted code trees but are mentioned in the above-mentioned Excel file. This has been done because otherwise, the codes trees would include too much detailed information. A PDF-file has been created that includes all codes, including the transcriptions that have been coded. The file can be found on the attached USB-stick.

(22)

22

5. Results

This chapter discusses the results of the sub-questions, accordingly, as explained in paragraph 3.3. The results come from two focus group interviews and semi-structured interviews, as explained in chapter 4. Every sub-paragraph represents a sub-question. Since not every respondent category has got the same interview questions, not all respondent groups are mentioned in the sub-paragraphs. Wherever possible, results have been linked to literature, but due to newness was this sometimes a challenge. The code trees about chatbots and communication are adjusted based on the results. The adjusted code trees can be found in appendices 9A until 10. An overview of the credentials of all the respondents can be found in Appendix 4.

5.1 Impact of chatbots on the communication between end-users and the

service desk

This paragraph discusses the impact on communication channels, changes in the click-call-face principle, and the changes in manners. Further, the reasons to use a chatbot and the communication after the launch of a chatbot. Titles (sub-sub paragraphs) in this paragraph are based on interview questions that belong and help answer sub-question 1. Answers on interview questions that did not help to answer the sub-question are not mentioned. Lastly, this paragraph ends with secondary findings that are also based on interview questions belonging to sub-question 1.

5.1.1 Impact of chatbots on communication channels

When looking at the results from the interviews, the impact of chatbots on communication channels differ per organisation. The channel chatbot can be seen as an additional channel and is not a replacement for a self-service portal or other channels. Opinions of interviewees were quite diverse. Pham, Fitzpatrick, and Hiemstra considered that the usage frequency of other communication channels should decrease after a chatbot is launched. According to Hiemstra, whose organisation uses a chatbot in their communication between external customers, when the amount of chatbot conversations increases, the frequency of the other communication channels decreases. This effect is enhanced by the location of a chatbot on a website. After performing internal analyses, Hiemstra found that 70% of the customers who used the chatbot will not use other communication channels

afterwards, which means a significant high satisfaction score. According to Bakker, whose

organisation also uses a chatbot in their communication between external customers, the amount of e-mails has significantly decreased after the chatbot was launched. Numbers or facts about the effects on other communication channels were not available. Bakker found also that the position of the chatbot on a website has an influence on the usage of a chatbot. Van der Lee, Van der Leest and Wijbenga mentioned that e-mails will decrease after a chatbot has been added to communication channels.

On the other hand, Tromp and focus group two considered that communication channels (telephone, e-mail, etc.) will not be affected in their frequency of how much they are used after launching a chatbot. The reason could be the fact that especially communication channels ‘Electronic Messages’ (e-mail) and the telephone were most used in the workplace in 2015 (Quaden, 2015). The focus group stressed that the influence of a chatbot on internal communication will be minimal but there is a chance that an end-user will choose for a chatbot when the chatbot accommodates better than an employee from the service desk. When digitalisation increases, people will become more used to it. The channels that will remain are telephone and the electronic message type e-mail. Focus group one mentioned that the telephone will stay necessary for complex messages. For example, when an event planner wants to explain the feeling of an event to an employee of the service desk. Lastly, Van der Lee and focus group two expect that electronic messages via applications and websites will increase because of the general shift towards more digitalisation. To conclude, when an organisation

implements a chatbot in their communication channels between end- users and the service desk, written communication may increase when launched successfully. Other sorts of communication (verbal and non-verbal) may decrease.

(23)

When looking at the literature, the impact on communication channels belongs to the element

“Channel” of the Transmission model from Shannon and Weaver (1949), which can be explained with the help of the Channel Richness model from Daft (2012). According to the results from the interviews, chatbots can focus on personal contact that streams two-way. The interaction between the chatbot and the end-user goes fast and takes place briefly. Therefore, the channel chatbot has a certain level of high channel richness and can be added to the upper part of the blue coloured block of the model.

Changes in markets

According to interviewees Vermeulen and Vogels, the changes in communication channels after launching a chatbot might be different per organisation or market. For example, in the B2B market when an external organisation provides facility management services for their client and the provider wants to communicate with the employees of their client via a chatbot. It might happen that the employees do not accept the new communication channel and rather use the more common channels to throw their service-related message ‘over the fence’ because they are loyal to different habits. For example, sending an e-mail or by having a telephone conversation. It seems that these habits/ manners are formed by organisation culture, which then confirms Essink (2016) who stated that manners are formed by the culture of an organisation. Vermeulen and Vogels continued, chatbots might work perfectly in a B2C market because of the level and speed of adoption. When looking at the literature from Kotler, Saunders, and Wong (2009), people who are working in the B2B market and work for an organisation that has outsourced its facility services can be seen as the category Late majority or Laggards, since they seem to stick with the common communication channels. People from B2C market, on the other hand, can be seen as the category Innovators, Early adopters or Early majority because of the level and speed of adoption, said Vogels and Vermeulen.

Stress

Besides the impact on the communication channels, adding a chatbot can also create stress. Van Dam mentioned when you are creating choice stress at the moment of choosing which channel someone needs to use to contact the service desk and not every channel can handle the same CWIF-process related questions, then you are hurting yourself. End-users get stressed and therefore they should be free to choose which communication channel they would like to use in order to contact a service desk. Van Dam stated: “…at the end of the day it is all about that you made the easiest and simplest experience for the user…” of a chatbot. According to focus group two, the choice of which communication channel someone wants to use depends on the reason that a customer defines. The basic overall idea is that an end-user always wants that his/her needs are met and it does not matter which communication channel someone chooses.

5.1.2 Changes in the click-call-face principle

The expected impact of chatbots on communication channels is also visible in the results of the interview question about what might change within the click-call-face principle. The principle was not known across the interviewees, except from the consultants who are part of category 3. Therefore, the principle has been explained before continuing with the interview questions regarding the click-call-face principle. None of the interviewees had this principle applied within their organisation.

The interviewees considered that click will increase, due to increasing digitalisation, call will decrease, and face will not change because personal contact and showing compassion will stay necessary in the field of facility management.

According to interviewee Pham, patience for call and face might decrease when a chatbot has been added to communication channels because of the answer-speed of a chatbot because end-users might assume that these channels might be used less and they might receive answers faster. Vermeulen and Vogels expect that none of the elements of the principle will change. Lastly, focus group two expected when digitalisation increases, people will become more used to it. As a result, the members expected that call and face will decrease in frequency of usage and that it will be more and more important that an FM-department shows that it knows its customers and its added value for the customers.

(24)

24

5.1.3 Changes in manners

The expected changes in manners of communication varied across the interviewees. Manners of customers of Hiemstra’s employer were quite diverse within their chatbot conversations. For example, some customers did have a conversation with the chatbot in a robotic way by starting a conversation with a subject they would like to talk about. For example, return order. But there were also customers who use full sentences and send messages as if they had a conversation with a human. Changes in manners via other communication channels were not known but Hiemstra expected that it will take some time before manners will be the same across the average conversations that customers have with the employer’s chatbot.

Further, Hiemstra mentioned that a chatbot conversation can start with a more natural flow by sending a nice human looking invitation text before a conversation starts. Showing compassion is also very important to get a natural flow and this might steer the manners. For example, by answering with: “We are very sorry for the delivery delay”. Another important factor for a natural flow is showing empathy in personalised text messages. Unfortunately, chatbots are not able yet to send this type of messages due to technological complexity. In case of unclearness between the customer and the chatbot, a live-chat employee should take over the conversation to show compassion and empathy, which needs to be done seamlessly. As a matter of fact, the competences of live-chat customer service employees from a customer contact center or service desk will change to a more personal approach in order to be able to help customers with complex questions, while the simple questions (FAQs) will most of the time be answered by a chatbot or FAQ webpage. According to Bakker, the manners of customers within her employer’s chatbot conversations did not change after adding a chatbot to their commu-nication channels, while the employer maintained the same organisational commucommu-nication style throughout all channels. Bakker mentioned that manners normally differ per client group, organisational branding, and organisational service goals.

Van der Lee, Van der Leest, Vermeulen and Vogels, and Wijbenga, interviewees from organisations who might possibly use a chatbot in their business, expected that manners will stay the same. Focus group two shared this meaning and added two ideas that could enhance the manners of end-users: 1) the chatbot changes its text based on the word choice of the end-user and 2) the chatbot changes its messages and answers based on the age of the end-user.

5.1.4 Most common reasons to use a chatbot

Interviewees were asked to give reasons for companies to use a chatbot. Their answers can be seen as the expected impact. The most common reasons of organisations to add a chatbot to their com-munication channels were according to chatbot developers: cost reduction, increasing self-service, automation of FAQs, process enhancement to save time and to solve important customer questions instead of FAQs, scalability, and unavailability of employees, which can be seen as the main expected impacts after launching a chatbot. According to Tromp, other reasons to add a chatbot to

com-munication channels were that a chatbot enables a customer to get answers faster, it saves training, and organisations do not have to deal with personal human struggles, e.g. the Monday morning feeling. Lastly, Tromp added that employees are more satisfied because they do not have to answer the FAQs anymore and can focus on answering complex questions, which means more customization. Information can be received faster by organisations and customers and it is easier to analyse the con-versations from data than analysing telephone concon-versations, mentioned chatbot developer

FitzPatrick.

The reason to add a chatbot to the communication channels of the employer of Hiemstra was mainly scalability since the organisation was growing and the number of received questions increased. The employer continued the use of the chatbot be satisfaction across contact center employees increased slowly and an increasing number of FAQs were being answered by the chatbot. As a result, employ-yees could focus on valuable conversations and empathy and urgency became more important. The reason to add a chatbot to the communication channels of the employer of Bakker was mainly to automate the questions to unburden the contact center and the second main reason was scalability. Another reason was that employees from the center did not have to be trained to answer questions, which saved time and money. The box (1) below contains a summary of the above-mentioned text.