1 Introduction

Technology has overtaken the world through artificial intelligence (AI) in this digital era (Letheren, Russell-Bennett & Whittaker 2020). With rigorous developments in AI, chatbots have taken the centre stage transforming the customer service profession, and its interaction with the users (Wirtz et al. 2018). Currently, chatbots are utilised to provide 24/7 support service specifically in the fields of marketing, sales and support, where on average it has helped increase sales by 67% (Press 2019; Ashfaq et al. 2020). Przegalinska et al. (2019) discussed that chatbots are the form of AI technology, that interact and engage with humans verbally, or through texts. Interestingly, many firms and industries are utilizing, and depending heavily on using chatbots as virtual agents to assist them in their business. With the assistance of these virtual agents, users can acquire the required information, place an order, or purchase at their convenience (Sivaramakrishnan, Wan & Tang 2007).

With the inception of chatbots in different business sectors, the market size is growing exponentially, from $250 million (in 2017), to almost $1.34 billion, by the year 2024 (Luo et al. 2019). The focus of this study was on the banking sector, utilising chatbots, as their representatives to communicate with their customers. Trivedi (2019) discussed that the banking industry has been the early adopter to implement the use of chatbots for providing information and interacting with the customers. Offering relevant information and providing a sound virtual experience leave the customers satisfied (Price 2018).

The purpose of this research was to investigate and explore the influence of chatbots’ service quality on customer satisfaction in the banking sector. When talking about the various types and forms of AI, human-chatbot (virtual agents) interactions have been the most prominent ones. One such study published by Zarouali et al. (2018) stated that human-chatbot interaction needs further exploration, in order to have a better understanding about the chatbots’ service quality, and the customers’ satisfaction.

The research in point conducted an extensive investigation on this topic, which enables the readers to have a better understanding about the influence of chatbots on customer satisfaction, specifically in the banking sector of UAE, since there is a lack of published evidence in this area. Therefore, this research was the perfect platform to explore and learn about the influence of chatbots on customer satisfaction in the banking sector of United Arab Emirates (UAE).

Chatbots are also used extensively in other fields, such as hospitality, tourism, healthcare, and education, serving different purpose in each of these sectors. They differ from the banking sector in the way that chatbots are more customer-focused and sophisticated when compared to the other sectors, since the information provided by the banks is highly confidential. Thus, the study seeks to answer the following research questions:

  • What is the influence of chatbot service quality in enhancing the customer experience in the banking sector?

  • Can chatbot interaction be considered as the replacement for human interaction at the banks?

  • Do the banks have the required infrastructure for implementing chatbot?

  • Can chatbots be considered as good enough to connect with the customers on an emotional level?

  • Can the human-chatbot interaction be considered safer as compared to human-human interaction at the banks?

Furthermore, to have more augmented quality of data, qualitative inductive research method was applied, being exploratory in nature, as it assisted in providing enriched data, with in-depth information and feedback from the participants. Data saturation was achieved by around 25 participants, adopting a cross-sectional research design, investigating consumers’ responses and intentions towards interacting with chatbots, by conducting interviews over a period of 3–4 months.

The outcome of this research developed a conceptual framework, which identified the factors causing customer dissatisfaction once they interacted with chatbots. Overall, a total of eleven themes and sub-themes emerged from the gathered data. This research further added value to customer experience, voluntariness to interact with chatbots, subjective norms, and the actual system usage. Additionally, certain practical implications were also discussed for the banks to consider and implement.

2 Literature Review

2.1 Artificial Intelligence in the Banking Sector

Chatbots, digital assistants, voice-activated services, concierge robots, and virtual assistants are all different terminologies used for similar type of AI-enabled systems (Prentice & Nguyen 2020). In this research, the focus is on the chatbots which are utilised by the banks, as a customer service representative.

Various authors and researchers have defined AI in different ways. From the time it was first discovered to the present time, the definitions have evolved as well, keeping in mind the various perspectives. To start with, in 1950, Turing (P. 435) defined it as “such a human interaction between another human and the machine where an individual is unable to differentiate between the human and the machine, hence the machine is termed to be intelligent”. Haenlein and Kaplan (2019) stated that “such an ability of the system that can explain the external data accurately, that can learn from the same source, and by using and incorporating that data, specific tasks and goals can be achieved by being flexible”.

2.2 Chatbots in the Banking Sector

Chatbot technology is being applied in various businesses and industries like banking, hospitality, tourism, and healthcare. With the help of these conversational chatbots, different businesses can collect and store larger sets of data in terms of variety, while assisting them to become more cost effective and sustainable (Campbell et al. 2020; Um, Kim & Chung 2020).

With the help of appropriate coding, programming, and algorithms, chatbots are trained to identify similar customer queries and respond to them in a specific pattern (Campbell et al. 2020). Belanche et al. (2020) highlighted that robots (chatbots) are being considered as frontline employees for most of the businesses, as they can easily communicate with the customers, receive and store larger data of information, and are cost effective. Chatbots are well-equipped with intuitive and empathetic skills, which is helping them in replacing all kinds of jobs performed by humans in various industries, particularly those of customer service representatives (Belanche et al. 2020; Huang & Rust 2020a, b). However, Um, Kim and Chung (2020) contended albeit chatbots make work easier for organisations to utilise them, they still have drawbacks; failure to understand and provide appropriate service to the customer, incur additional costs; maintenance, installation and training.

2.3 Extant Literature on Chatbots

John McCarthy; the father of “artificial intelligence” coined this term first in 1956 during a summer research project (Haenlein & Kaplan 2019). Since the research in the field of AI-enabled systems (chatbots) is still in its infancy stage, the following Table 1 is devised based on the key information provided by various journal articles.

Table 1. Extant Research on Chatbots (Table Compiled by the Researcher)

2.4 Banking Sector of UAE

The world has witnessed substantial developments in AI-based innovations across the global banking sector, which serves customers that are using diverse banking networks (Kumar, Sujit & Charles 2018). There is still an impending need for research when it comes to the implementation of AI-based technologies in the banking sector of UAE. Hence, it was necessary to present a concise synopsis of UAE’s banking sector.

Over the years, UAE has become a strong proponent and fore-bearer in introducing the latest technologies to its various sectors. Talking about the banking sector, nearly all the banks based in UAE have now implemented AI-enabled systems (chatbots and virtual assistants) (Mehta & Bhavani 2017). Al-Marri, Ahmed and Zairi (2007) signified the fact that when it comes to the perception of customer service quality, UAE’s banking sector is large enough to cover this area with vast outreach and better outcome. Digitalisation of the banking sector backed by AI-enabled systems has given the UAE’s banking sector a growth of 13.9% in terms of profitability (KPMG 2020).

While banks in UAE are making extensive efforts to retain and satisfy their customers, it is imperative to primarily satisfy their needs employing the cutting-edge technological innovations that are in place (Kumar, Sujit & Charles 2018). Alhosani et al. (2019) discussed the benefits and drawbacks of implementing the latest technologies. One of the major drawbacks faced by UAE’s banking sector in implementing the latest innovations is that they have been the target of cybercrime attacks, resulting in data compromise of approximately 14 million records in the year 2018, whereas, UAE lost roughly $1.1 billion to cybercrime attacks in 2017 (Rosberg 2018; Alhosani et al. 2019). Another major reservation faced by the banks in UAE is the customer satisfaction, which is not earned easily in this part of the world (Sleimi, Musleh & Qubbaj 2020).

2.5 Customer Satisfaction

The word “satisfaction” is derived from Latin “satis” meaning “enough”, and “facere” meaning “make or to do”. Hence, the combined meaning of these words mean to provide the products and services which have the capacity to be “enough” (Oliver 2010). The meaning of the word satisfaction is also dependent on the context. For instance, if “satisfaction” is used in a marketing context, its meaning becomes more specific (Parker & Mathews 2001).

Söderlund (1998) defined customer satisfaction as, “the value one receives after purchasing or using a product or service”. It is worth mentioning here that the context in which customer satisfaction is mentioned is very important, since the meaning changes from one context to another.

2.6 Repurchase Intention

When someone mentions customer satisfaction, they cannot leave repurchase intention out of it, as they both go hand in hand. It will not be wrong to say that it is the customer satisfaction that leads to repurchase intention (Hellier et al. 2003). Repurchase intention is defined as, “the willingness to purchase something again” Dictionary (2021a).

Suhaily and Soelasih (2017) mentioned that customer satisfaction is a compulsory factor when it comes to the repurchase intention towards a product or service, based on the overall quality provided to the customer. Furthermore, Mittal and Kamakura (2001) elaborated that once the customer reaches threshold or the tolerance level towards the repurchase, it may vary from person to person to display the repurchase intention, as not everyone have similar characteristics when it comes to repurchasing.

2.7 Word of Mouth (WOM)

Word of mouth (WOM) is another important factor that plays a vital role when it comes to customer satisfaction and repurchase intention. According to the Cambridge Dictionary (2021), WOM is defined as “such a process in which the information is being passed on from one person to another through the medium of verbal communication”. WOM has developed into an important means of communication for the field of marketing, which is defined as “the conveyance of message from person to person using oral communication” (Matute, Polo-Redondo & Utrillas 2016).

Communication of information between customers about the products/services, or the company is considered crucial when it comes to the repurchase intention, as customer needs to be completely satisfied before making an informed decision to purchase a certain product/service (Litvin, Goldsmith & Pan 2006). WOM has also transformed over time into electronic word of mouth (eWOM) and algorithmic word of mouth (aWOM), where eWOM is electronic conveyance of message or information, and aWOM is the AI enabled word of mouth using all the sophisticated technology to communicate among people (Williams, Ferdinand & Bustard 2019).

2.8 Technology Acceptance Theories

Despite the ease and convenience provided by chatbots to the organisations, more attention needs to be paid to customer satisfaction, since chatbots are AI-enabled systems, programmed by humans, lacking flexibility (Prentice & Nguyen 2020). Various researchers have come up with different technology acceptance theories, where the most relevant ones are Technology Acceptance Model (TAM), Unified Theory of Acceptance and Use of Technology Model (UTAUT), Unified Theory of Acceptance and Use of Technology Model 2(UTAUT 2), and Technology Acceptance model 2 (TAM 2).

2.9 Technology Acceptance Model 2 (TAM 2)

Technology Acceptance Model 2 or TAM 2 is an extension of the Technology Acceptance model, developed by Venkatesh and Davis (2000). TAM 2 integrated additional external constructs to the original TAM model (Venkatesh & Davis 2000). The objective of this theory was not only to contemplate on user acceptance of technology, but also to understand the user adoption behaviour through social influence and cognitive instrumental processes (Venkatesh & Davis 2000).

One such study published by Ali et al. (2021) argued that the extension of TAM can also be adopted when it comes to the use of cardless banking system. They further suggested that perceived risk plays a big role when it comes to consumers’ intention of use, based on the extended TAM model (Ali et al. 2021). Hence, this theory was the perfect fit to be used and implemented for the purpose of this research.

3 Scope of the Study

Fig. 1.
figure 1

Scope of the Study

The “scope of the study” for this research is shown above. As illustrated in Fig. 1, the focus was on achieving customer satisfaction while an individual uses the chatbot services in the banking sector. This led the customer towards deciding about their responses and intentions related to the chatbots and their service quality. The focus of this research was, (and not limited to) “Going with the same bank”, “Changing bank”, “Complaining to the bank (Does the complaints reach the responsible person/department?)”, and “Word of Mouth (WOM)”, respectively.

As elaborated in the literature review, the outcome of customer satisfaction helps the customer to decide about their responses and intentions. Since chatbots are relatively new in the banking sector, there have been mixed responses about the interaction between the chatbots and an individual, as highlighted in the recent research in different business areas. However, the influence of chatbots on customer satisfaction in the banking sector is novel.

4 Data and Methodology

The target population for this research were all the individuals who interacted with the bank directly or indirectly. The sample size for this research was determined based on reaching data saturation. To target the right participants, this research implemented a non-probability, stratified purposeful sampling technique, based on the age category. For this research, a total of 25 participants were interviewed, achieving data saturation. An expert interview, semi-structured survey instrument was employed with open-ended questions guided by the underpinned theory and literature.

4.1 Data Collection and Sample

For the data collection of this research, a face-to-face, semi-structured expert interview with open-ended questions was implemented. Non-probability, purposeful stratified sampling technique was used to collect the required number of participants for this research (until data saturation was achieved). The participants were contacted through emails, WhatsApp, and over the phone. Due to Covid-19 pandemic, online platforms; Zoom and Google meets were used to interview the participants, in lieu of interviewing them in-person. Moreover, interview guide was established to stay on track with the purpose of this research, along with the questions for this interview, which were guided by the theory and literature (Kvale 2007; Given & Saumure 2008; Creswell 2009a, b). For collecting audio records, two separate high-quality instruments were used; iPhone 12 Max Pro, and HP Pavillion laptop. The researchers used iPhone’s “Notes” and Google Doc’s “Dictation” function to transcribe the data.

4.2 Coding Procedure and Thematic Analysis Framework

NVivo12 software was used to describe and classify data to interpret into codes and themes. With the help of NVivo12 software, a large set of data gathered from interviews was analysed in a systematic way. Initial coding was carried out manually, and then all the coded data and themes were transferred to NVivo12 for further organisation and analysis. It also assisted in describing the essence of phenomenology for this research. Analysing the data using NVivo12 further assisted in enhancing the consistency, and validity of this research. According to Creswell and Poth (2018), it is duty of the researcher to write a detailed description of the themes, that are developed from the study. Once all the data was transcribed into Word document, it was then run through NVivo12 for further analysis into themes, and sub-themes. For this research, the data was coded using open coding, axial coding, and selective coding.

5 Findings

A total of 25 participants were interviewed, which was clearly sufficient to achieve data saturation, using open-ended semi-structured expert interview questions. Out of the 25 participants, 12 participants were expatriates with various nationalities (USA, UK, Spain, Australia, Pakistan, India, South Africa, Macedonia, Bangladesh), while 13 were GCC nationals from UAE, Oman and KSA. 23 participants were Male, whereas only 2 Females participated to answer the questions of this research. All the participants were from various age groups ranging from 18 to 56 and above. As for the educational background, the participants were fairly educated, ranging from holding Diploma to PhD, respectively. The researchers implemented the Five-step process in Framework Analysis based on Ritchie and Spencer (1994), and Ritchie, Spencer and O’Connor (2003).

Once familiarising with the data collected from the participants was completed, the researchers were then able to identify the initial emerging themes and sub-themes. All these themes and sub-themes were gathered within a framework to develop an initial conceptual thematic framework, as shown below in Table 2.

Table 2. Conceptual Thematic Framework of Initial Themes and Sub-themes

A total of seven major themes and four sub-themes were developed that were consequently linked with three aggregate dimensions. The finalised thematic framework was developed using the individual interviews (containing the 1st order concepts, 2nd order themes and sub-themes, and aggregate dimensions). The aggregate dimensions deducted from the thematic framework were “Perceived Chatbot Service Quality”, “Customer Satisfaction”, and “Responses and Intentions”.

All the sub-themes that imitated related or parallel data gathered from the individual participants were recorded under the appropriate main themes. All the themes and sub-themes formulated the thematic framework based on the three-level framework suggested by Gioia, Corley and Hamilton (2012).

5.1 Aggregate Dimension One: Perceived Chatbot Service Quality

Perceived chatbot service quality was labelled as the first aggregate dimension. This dimension revealed how the customers perceived chatbot service quality of their bank. It comprised of the 1st major theme “chatbot efficiency” and its sub-theme “service information and understanding”, 2nd major theme “chatbot interaction” and its sub-theme “limited options and emotional aspects”, respectively. Almost all the participants agreed that chatbots are not very efficient when it comes to assisting or solving their problems, as depicted in the following excerpt from the interview of participant no. 12:

“I don’t know how it would be better, or how people would get used to it, but it needs a lot of time to be more efficient. It is not that efficient nowadays unfortunately, with most of the places that I experienced.” (P12)

Another participant number 10 mentioned that chatbots require lengthy security information before it can verify the customer and proceed further with their request:

“…Some entities tend to make it as a basic and an easy system for the customers, while some provide a very complicated system. Like in the banking sector, they tend to put some restrictions on securities, and this makes it difficult for the client to get what they want as easily as possible. I think that is the downside of using a chatbot in bank. It further needs to consider the type of service. Some services are easy to use with the chatbot, while some of them like banks is a bit difficult because of the security issues.” (P10)

5.2 Aggregate Dimension Two: Customer Satisfaction

The second aggregate dimension was labelled as Customer satisfaction. This dimension revealed how satisfied the customers felt while interacting with the chatbot of their bank. It comprised of 3rd major theme “experience” and its sub-theme “urgent actions and problem solving”, 4th major theme “participant’s feedback”, and the 5th major theme “human interaction” and its sub-theme “human representative and assistance”, respectively. A few examples are portrayed below, from the excerpts of participant’s interviews:

“…if you are in a hurry and you have some other work to do, it is extremely necessary for you to finish with the bank, but it takes 5 to 10 minutes, then that is very annoying. For me, I have been at both the ends, I mean I have experienced both; being satisfied with its service and annoyed with it at times.” (P22)

“…it was very interesting because this is something all of us have experienced at some point, happily and unhappily. So, I also had mixed kind of reaction while interacting with a chatbot. However, I distinctly remember my first interaction. I was intrigued that how this was going to work. I wouldn’t say it was too bad, but it was not the best experience that I had. I still remember my ATM card had gotten stuck in the ATM machine; it was not coming out. I was trying to reach the helpline, but I couldn’t reach any helpline. So, that was the first time I interacted with a chatbot. Now the thing with the chatbot is that it gives you almost the perfect information that one could possibly be looking for, and it gives you all the details that you might find necessary for your need of the hour, but also it is limited in a way that it is a programmed machine.” (P3)

5.3 Aggregate Dimension Three: Responses and Intentions

Responses and Intentions was labelled as the third aggregate dimension. This dimension revealed about the participants responses and intentions once they interacted with the chatbot of their bank. It comprised of 6th major theme “suggestions from experience” and 7th major theme “Word of Mouth (WOM)”, respectively. The following examples illustrate participant’s responses related to the aggregate dimension three:

“…the other one which I think is about the age factor especially the old people, I don’t think that they are very technical people. Maybe out of 100, 2 or maximum 3 people will know or are familiar with chatbot. Other than that, I am sure they are avoiding it and prefer to interact directly with a human representative or visit the branch itself.” (P1)

“I am a technology savvy person, but everyone is different than the other human, and let us not forget who hardly understand technology like aged people, labour class and uneducated people, for example. So, banks need to consider their situation as well to try and create a way for their easy access as well, because honestly, not everyone is willing to spend money on costly calls and end up with no result at the end.” (P20)

6 Discussion

Analysis of stories and experiences shared by the 25 participants revealed that customers are hardly influenced by chatbots in the banking sector, resulting in their dissatisfaction with it. As mentioned in the extant literature on chatbots, Um, Kim and Chung (2020) suggested that self-service technology is much more efficient than AI-enabled systems like chatbots and virtual assistants. Participants highlighted the fact that chatbots are good and highly efficient in other sectors like hotels, hospitals, airlines etc. They feel that banks are not quite ready when it comes to chatbot’s efficiency.

As emphasised by De Cicco, e Silva and Alparone (2020) that people need to connect on an emotional level when they interact with the bank. Unfortunately, that is not the case when it comes to interacting with the automated systems introduced by various banks. Sleimi, Musleh and Qubbaj (2020) also stated that not everyone is content with using the latest technology in the banking sector.

Additionally, the literature also predicted that customers need quality service based on their demands (Pakurár et al. 2019). Again, this is not what participants felt when they interacted with chatbot. It was due to poor chatbot quality in banks. Since no prior research has been conducted about the human interaction with chatbot, this research was able to discover that chatbots provide clear instructions on how to navigate or what options to choose, but for simpler requests only. On the contrary, when it comes to emergency situations, there is still no comparison with a human representative, as humans are much more understanding and cooperative.

Furthermore, in literature Moriuchi et al. (2020) also emphasised that organisations have started implementing chatbots as their customer representative, but they have neglected the customers’ side whether they like interacting with them or not. As for the sub-theme, emotional unavailability and limited options were one of the most important findings of this study and the most reported concern by the participants. One such research published by Elsholz, Chamberlain and Kruschwitz (2019) mentioned that chatbots are limited when it comes to commands, languages and emotions.

Additionally, participants further revealed that human representatives are better at assisting and navigating through the offers and services, in a cost-effective manner. It was not the same case with chatbots, as there are different categories and sub-categories of options that take too much time to go through. As per Huang and Rust (2020a, b), AI-enabled systems are outperforming humans, hence a grave threat to their jobs.

Alternatively, a few participants had positive WOM about the influence of chatbots. Such experiences make up for the overall theme. However, a few participants were not bothered by others’ opinion. They preferred visiting the bank personally to get their work done. According to the literature reviewed, negative WOM exerts far greater influence than the positive WOM, creating switching intentions (Um, Kim & Chung 2020).

7 Implications

Theoretically, this research supported the marketing literature in the fields of bank marketing, marketing communication, consumer behaviour, product development and relationship management, respectively. Furthermore, the findings of this research could easily be associated with the underpinned Technology Acceptance Model 2 (TAM 2), as demonstrated in Fig. 2. All the black arrows indicate the original model suggested by Venkatesh and Davis (2000), whereas the orange arrows and boxes indicate the theoretical contributions of this research.

Fig. 2.
figure 2

Theoretical Contributions

The findings of this research could be considered as the variables of a new quantitative study. Hence, it could be used as an initial point of future studies associated with the influence of artificial intelligence (AI) on customer satisfaction. Moreover, the findings of this research further add value to customer experience, which matters the most when it comes to being satisfied with their interaction with chatbot. Based on what participants went through while interacting with chatbots, it was their experience which created a certain image of using such a system implemented by banks.

On the other hand, voluntariness of interacting with chatbots had an impact on the service quality. The more willing and volunteering the customers are to interact with chatbots, the better will be the service quality. There was also an addition of 5 new variables to TAM 2 model, namely emotionally draining, unwieldy interface/cumbersome navigation, misinterpretation/delusional, backend operational malfunction, and customer privacy and confidentiality. All these factors fall under the classification of customer satisfaction.

This research recognized and developed an understanding of the influence of chatbots on customer satisfaction in the banking sector of UAE. Therefore, based on the findings of this research, it could be linked to many aspects of banks’ strategies related to customer satisfaction, responses and intentions, bank marketing, marketing communication, consumer behaviour, product development and relationship management, respectively.

Reflecting on the findings of this research, banks should develop new strategies to satisfy the customers by providing them easy to use services and improving chatbot service quality. Hence, this research would provide the banks with the knowledge of customer satisfaction, responses and intentions, drawbacks of chatbots’ service quality, upgrading it to newer easier system, and devising new strategies to make the best out of implementing chatbots in banks in the United Arab Emirates.

The word “AI” or “chatbot” meant something very intelligent in people’s perspective, which changed drastically once they interacted with it. Hence, it calls into account the banks’ policies and strategies towards consumer behaviour and relationship management. By doing so, bankers will be able to understand better what is causing their customers’ dissatisfaction, and how they need to adjust their strategies according to customers’ demand. Therefore, the banks and marketing professionals need to understand that although banks are trying to adapt to new norms, technologies and policies followed globally, they need to devise a proper marketing strategy for it.

8 Limitations and Future Research

Firstly, the research was conducted within UAE during the time when COVID-19 was at its peak. Due to the nature of this research, the authors had to conduct semi-structured, open-ended, expert interviews. Inviting participants was not a big issue but agreeing to meet and sit for an interview was quite challenging.

Secondly, the time constraint for data collection was also very crucial, as the researchers had to reschedule many of the online interviews due to the pandemic situation. Hence it took 3–4 months to gather all the required data. Another limitation faced during the completion of this research was not to interview labour class individuals, as they are not able to withdraw their salary from the ATM machine, so interacting with chatbots was completely out of the question for them.

Considering the findings as the foundation to further enhance and develop studies accordingly, it is highly recommended for the future researchers to explore quantitatively the influence of chatbots on customer satisfaction in the banking sector in UAE, using the factors and the proposed conceptual model in this research. For example, if there is influence of hybrid systems on customer satisfaction, how would the customers perceive service quality of hybrid systems? Does interacting with hybrid systems impact customer loyalty? To what extent do customer experiences with hybrid systems influence their intentions?

Moreover, future researchers need to focus more on the ways marketing could minimise the negative consequences on the influence of chatbots on customer satisfaction in the banking sector in UAE, knowing how to manage different marketing and consumer behaviour aspects in predicting their behaviour. Also, cultural, and other environmental factors affecting the influence of chatbots on customer satisfaction in the marketing perspective need to be studied.