Electronic Commerce Research and Applications 65 (2024) 101380

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications

journal homepage: www.elsevier.com/locate/elerap

Understanding virtual agents’ service quality in the context of customer

service: A fit-viability perspective
Yanping Zhang a, Changyong Liang a, Xiaodong Li b, *
a
School of Management, Hefei University of Technology, Hefei 230009, PR China
b
School of Economics and Management, Anhui Polytechnic University, Wuhu 241000, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: Virtual agents, a prevalent application of artificial intelligence, significantly enhance e-commerce customer
Virtual agent service (i.e., online interactions that address customers’ concerns). This creates an innovative service model
Service quality aiming to meet diverse customer demands and create value for enterprises. However, understanding the factors
E-commerce customer service
driving the performance of virtual agents, especially in terms of service quality, remains a gap. Expanding the fit-
Fit-viability model
viability model (FVM), this study investigates the drivers and mechanisms influencing virtual agents’ service
quality, considering the matching of technology, service scenarios, and individual readiness. Surveying 399
respondents experienced in virtual agent service within e-commerce, the findings reveal that personalization
negatively impacts response speed but positively influences response accuracy. Technology readiness positively
affects convenience and accessibility. Overall, service quality is positively influenced by response speed, response
accuracy, convenience, and accessibility. This study extends the fit-viability model to virtual agent service
literature, providing practitioners with a fresh perspective to enhance e-commerce customer service.

1. Introduction hold inconsistent assessments of the service quality of virtual agents, and
most of them are dissatisfied with these agents (UJET, 2022). In fact,
With the rapid development of artificial intelligence, virtual agents service quality is the core of performing e-commerce customer service
are widely used to reform the business operations model of e-commerce and determines the true success of virtual agents. High service quality
customer service in various contexts, such as the tourist industry, directly enhances consumer experience, which helps improve consumer
banking, and retailing (Das et al., 2023; DATA CONOMY, 2022). Spe­ satisfaction and loyalty (Wang et al., 2019; Kasiri et al., 2017). Moving
cifically, virtual agents are programs capable of providing customer into the context of the e-commerce customer service delivered by virtual
service through simulated conversation, leveraging artificial intelli­ agents, the positive relationships are also verified (Behera et al., 2021;
gence (e.g., natural language processing and self-learning) to achieve Ashfaq et al., 2020). Thus, how to enhance service quality of virtual
automation and intelligence. They can handle multiple tasks simulta­ agents has become a key focus and challenge for practitioners.
neously and provide personalized and real-time information and solu­ Though service quality is important, less attention was paid to
tions to customers, thus improving service efficiency and consumers’ exploring its antecedents. Very few studies examined the drivers of
experience, strengthening customer relationships, and reducing enter­ service quality from the perspective of technology features (Chung et al.,
prise operating costs (Radziwill and Benton, 2017; Wirtz et al., 2018). 2020; Haugeland et al., 2022). However, virtual agents providing
Virtual agents have proved to have the potential to bring new value to customer service are a process of human–computer interaction. Con­
business organizations. For example, Servion (2020) forecasted that sumers’ perception of virtual agents can be affected not only by tech­
virtual agents would drive 95% of consumer interactions within firms by nical characteristics but also by environmental support factors, such as
2025. Furthermore, Statista (2021) predicted that the market value of individual-related factors (Rapp et al., 2021) and availability (Ashfaq
virtual agents in banking, retailing, financial services, and insurance will et al., 2020). Therefore, it is crucial to understand the drivers of virtual
reach 6.83 billion U.S. dollars by 2030. agents’ service quality in e-commerce customer service from a
Although virtual agents have significant market potential, consumers comprehensive perspective.

* Corresponding author.
E-mail address: [email protected] (X. Li).

https://doi.org/10.1016/j.elerap.2024.101380
Received 12 July 2023; Received in revised form 21 February 2024; Accepted 14 March 2024
Available online 17 March 2024
1567-4223/© 2024 Elsevier B.V. All rights reserved.
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Even though previous studies have explored the comprehensive and learning consumer data (e.g., history information and preferences).
factors (e.g., service per se, technical characteristics, and personal Third, virtual agents can help companies reduce customer service
ability) that influenced service quality provided by traditional agents (e. operation costs. For example, virtual agents can process multiple session
g., human employees, websites, and self-service) (Brady and Cronin, windows at the same time without limitations of location or knowl­
2001; Blut, 2016; Kim and Chen, 2023), we cannot directly apply the edgeable staff (Wu et al., 2020). Finally, virtual agents do not have
results to the context of virtual agents. This is because virtual agents, as negative human emotions like impatience or fatigue. Time after time,
new agents, are different from traditional agents. Virtual agents inte­ they can engage in friendly conversation based dynamically on con­
grate human attributes and technical capacity and replace humans to sumer inputs (Adam et al., 2021). Thus, practitioners and researchers
provide service information and solve questions. They possess the abil­ are interested in how to enhance the performance of customer service
ities of big data analysis and deep learning and surpass traditional virtual agents.
agents, enabling them to offer personalized services that cater to the Previous studies have paid attention to acceptance (Zhang et al.,
diverse needs and preferences of consumers. It is essential to focus on the 2021; Rese et al., 2020), customer experience and satisfaction (Araujo,
extent to which virtual agents meet consumer needs in customer service 2018; Pizzi et al., 2021; Sands et al., 2021), and engagement (Kull et al.,
to understand how these factors influence service quality in the context 2021; Schuetzler et al., 2020) of virtual agents in the e-commerce
of virtual agents. customer service context. Meanwhile, some studies also examined the
Given the practical and research gaps, we have posed the following consequences (e.g., customer experience and satisfaction) of virtual
research questions (RQs): RQ1: What factors drive service quality of agents’ service quality (Behera et al., 2021). For example, Ashfaq et al.
virtual agents from the perspective of the matching of technology, ser­ (2020) identified that service quality positively influenced consumers’
vice, and consumer? RQ2: How do these drivers affect virtual agents’ satisfaction and predicted continuance intention. Very few studies
service quality? In order to fill the research questions, based on the fit- explored the drivers of virtual agents’ service quality in e-commerce
viability model (FVM), we propose the drivers of service quality from fit customer service (Baabdullah et al., 2022). For example, Haugeland
which measures the match between technical characteristics and service et al. (2022) focused on the roles of two interaction design features (i.e.,
requirements of tasks (i.e., personalization, response speed, and topic-led conversations and task-led conversations, free text interaction
response accuracy), and viability which measures the match between and button interaction) in enhancing the service quality of customer
application of technology and its associated consumer (i.e., technology service virtual agents. These studies illustrated the importance of service
readiness, accessibility, and convenience). We also examine the influ­ quality from the perspective of technology features, neglecting the de­
ence of personalization (technology readiness) on response speed and gree to which technological features meet the customer service tasks,
response accuracy (accessibility and convenience). and the support from environmental factors. Therefore, we need to
This study makes several contributions. First, it identifies several identify the drivers of virtual agents’ service quality through the fit-
antecedents of virtual agents’ service quality in the context of e-com­ viability model.
merce customer service by considering the roles of technology, service
scenario, and individual readiness. It provides a comprehensive angle 2.2. Service quality and its antecedents
for understanding how virtual agents create value for customers in e-
commerce and extends the understanding of virtual agent services. Service quality was formally proposed by Grönroos (1984) and has
Second, the FVM is applied in this new scenario to understand consumer been recognized as an important construct to measure consumers’
perceptions of service quality and examine the correlation of variables perception of service performance. Previous research has examined
within the fit and viability dimensions, extending the application scope different service agents in the context of marketing and identified the
of the model. Furthermore, this study provides practitioners with sug­ drivers of service quality from various perspectives (see Table 1). The
gestions on how to design and improve virtual agents’ performance, earliest research focused on human employee service in face-to-face
enhancing the development of e-commerce. encounters, which identified service per se, human employee’s exper­
The rest of this paper is structured as follows. Section 2 reviews the tise, and environment as vital factors driving service quality. For
literature on virtual customer service agents, service quality and its instance, Brady and Cronin (2001) proposed a classical model of service
antecedents, and the FVM, then establishes the proposed model. Section quality with three dimensions: outcome, interaction, and environmental
3 presents the hypotheses. Section 4 reports the research method. Sec­ quality. This model demonstrated and verified that attitude, behavior,
tion 5 details the results. Finally, we discuss the findings, contributions, expertise, ambient conditions, design, social factors, waiting time, tan­
additional implications, and limitations. gibles, and valence play important roles in assessing the quality of
human employee service.
2. Literature review With the development of information technology, customers can
access services through websites. Studies have investigated factors
2.1. Virtual agents for e-commerce customer service influencing service quality by considering technology-related factors,
such as website availability, and how websites meet customer service
Virtual agents, which have changed the platform business model by requirements. For instance, the e-service quality model (eTailQ model)
creating more value for customers and enterprises, have been widely is commonly used to predict web service quality based on dimensions
used for e-commerce customer service. As conversational service tools, such as website design, customer service, fulfillment, and security/pri­
they provide customers with recommendations and decision aids to vacy (Wolfinbarger and Gilly, 2003; Blut, 2016). Huang et al. (2019)
meet customers’ needs by artificial intelligence technology (Chen et al., found that accessibility and ease of use are important antecedents of
2021b). Virtual agents play a vital role in facilitating customer online customer service quality.
engagement, enhancing the user experience through real-time and more Moreover, with the rise of e-commerce, consumers can complete the
human-like interactions (Chung et al., 2020; Scherer et al., 2015). Vir­ entire service process online by interacting with self-service technology,
tual agents have numerous advantages over traditional agents. First, placing them in a dominant position in the service. Individual initiative
they can provide 24/7 and real-time online customer service, which can can significantly affect the user experience; therefore, individuals’
improve service providers’ responsiveness and product conversion rate. ability (e.g., attitude and control) to use technology is also considered.
Radziwill and Benton (2017) verified that virtual agents can provide For example, Pooya et al. (2020) illustrated that technology readiness
uninterrupted and time-to-response customer service, increasing has a positive significant effect on consumers’ perception of service
customer satisfaction and engagement. Second, virtual agents have quality toward self-service banking.
unique capabilities and can provide personalized service by analyzing Although previous studies on traditional agents have enhanced our

2
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Table 1
Literature on drivers of service quality in different contexts of agents.
Types of agents Sources Research perspectives Drivers

Human (Brady and Cronin, Outcome quality, interaction quality, Attitude, behavior, expertise, ambient conditions, design, social factors, waiting time,
employee 2001) environmental quality tangibles, valence
(Kasiri et al., 2017) - Standardization, customization
Online human (Wang et al., 2004) SERVQUAL model Tangibles, reliability, responsiveness, assurance, and empathy, network quality
employee
Website (Wolfinbarger and E-service quality model (eTailQ Website design, customer service, fulfillment/reliability, security/privacy
Gilly, 2003) model)
(Lee and Lin, 2005) SERVQUAL model Website design, reliability, responsiveness, trust
(Fassnacht and - Reliability, functional benefit, emotional benefit, attractiveness of selection, information
Koese, 2006) quality, ease of use, technical quality, graphic quality, clarity of layout
(Blut, 2016) eTailQ model Website design: information quality, website aesthetics, purchase process, website
convenience, product selection, merchandise availability, price offerings, website
personalization, and system availability;
Fulfillment: delivery timeliness, order accuracy, and delivery condition;
Customer service: service level and return handling/policies;
Security/Privacy: security and privacy
Self-service (Dabholkar et al., Attribute-based model, overall affect Speed of delivery, reliability, enjoyment, ease of use, control, attitude toward using
technology 1996) model technological products, need for interaction with service employee
(Shamdasani et al., Self-service quality framework, the Speed of service, ease of use, reliability, enjoyment, control
2008) technology acceptance model
(Kallweit et al., 2014) Technology acceptance model Adequacy of information, usefulness of content, perceived ease of use, attitude towards usage
(Pooya et al., 2020) - Technology readiness
(Kim and Chen, - Consumer empowerment
2023)
Virtual agent (Baabdullah et al., Technology interactivity model Personalisation, responsiveness, ubiquitous connectivity, readability, transparency
2022)
(Haugeland et al., Interaction design features Interaction style: topic-led conversations, task-led conversations; conversation types: free text
2022) interaction, button interaction
(Chung et al., 2020). Technical features Interaction, entertainment, trendiness, customization, problem-solving
(Song et al., 2022) Type of service agent Type of service agent (chatbot/human)

understanding of antecedents of service quality by considering service- There are three critical reasons for adopting the FVM as our theo­
related, technology-related, and individual-related factors, it is not retical framework. First, the FVM is utilized to assess the performance of
appropriate to translate them directly to the context of virtual agent new technology applications, and virtual agents fall into this category.
services. This is because the existing antecedents are based on specific In this study, our focus is on exploring the performance, specifically the
encounter types, such as human employee service, website service, or service quality, of virtual agents in customer service. Thus, the FVM
self-service. Virtual agent services have some unique interaction styles. aligns with our research goals. Second, service quality is influenced not
Firstly, unlike human employees, virtual agents provide service through only by technical attributes satisfying consumer needs (Grönroos, 1984)
an artificial intelligence (e.g., big data analysis and natural language but also by the way consumers obtain services (Zeithaml et al., 2002). As
processing) system. Secondly, compared with website service or self- mentioned above, in the context of virtual agents, we should explore
service, virtual agent service is based on interactive conversations. drivers of service quality by considering the matching of technology,
These unique interaction styles determine that the key to the successful service, and consumer. These aspects align with the two dimensions of
provision of customer service by virtual agents lies in virtual agents’ the FVM: fit and viability. Third, the FVM has been applied in the
ability to match service and consumer. Currently, only a limited amount assessment from the perspective of individual consumers. For example,
of research has begun to explore the service quality determinants of Zhang et al. (2020) applied the FVM to assess how mobility affects social
virtual agents from the perspective of technical characteristics. Hence, a media advertising effectiveness from the consumer standpoint. It thus
new model is needed to comprehensively understand the virtual agent can be equally applicable to our research context.
service context adequately by combining the match between technol­
ogy, service, and consumer. 2.4. Proposed model

2.3. Fit-viability model 2.4.1. Contextualization of the fit dimension

In the fit dimension, we contextualize and identify personalization,
The fit-viability model was developed by Liang and Wei (2004) to response speed, and response accuracy as the drivers of service quality.
assess the success of mobile commerce applications from an organiza­ Consistent with the fit dimension of the FVM, in this study, fit measures
tional perspective. It originated from a two-dimensional matrix (i.e., fit the alignment between the technical characteristics of virtual agents and
and viability) (Tjan, 2001) and then became a guideline to evaluate the customer service requirements (Liang et al., 2007). First, efficiency and
performance of adopting technologies via fit and viability dimensions automation are primary reasons organizations use virtual agents for
(Liang et al., 2007). Fit measures the match between technologies and customer service tasks (Li et al., 2020; Selamat and Windasari, 2021). In
the requirements of tasks (Liang et al., 2007); viability refers to the other words, virtual agents can provide consumers with real-time and
extent to which the environment or organization is ready for a given accurate services, such as information and solutions. Additionally, vir­
application (Liang et al., 2007), which generally has three aspects: tual agents possess big data analysis and deep learning capabilities,
economic feasibility, technical infrastructure, and users’ readiness to enabling them to offer personalized services (Chung et al., 2020). Thus,
use. This model has already been adopted to explain the performance of providing real-time, accurate, and personalized customer services are
new technology applications in many contexts (Larosiliere and Carter, technical features of virtual agents. Second, speed and accuracy of
2016; Liang et al., 2021; San Martin et al., 2012). For example, Lar­ performing tasks are considered two fundamental requirements of on­
osiliere and Carter (2016) used the FVM to evaluate antecedents of e- line customer service. Previous research has identified that responsive­
government maturity at the country level. ness and reliability are vital factors affecting service performance

3
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

(Palese and Usai, 2018; Parasuraman et al., 1988). Considering the the viability dimension (i.e., technology readiness, convenience, and
match between technical features of virtual agents and the requirements accessibility) in the “Contextualization of the viability dimension” sec­
of customer service (Goodhue and Thompson, 1995), we argue that tion. We propose a model to elucidate the driving mechanism of virtual
personalization, response speed, and response accuracy influence ser­ agents’ service quality in customer service settings (see Fig. 1). In the
vice quality. model, we examine the relationship between fit, viability, and service
Furthermore, virtual agents, powered by big data and artificial in­ quality. Furthermore, we also propose the internal relationships be­
telligence, can provide unique services tailored to the needs of different tween personalization (technology readiness) and response speed/
customers based on preferences, demand backgrounds, and contextual response accuracy (convenience/accessibility). Then, we regard service
information. Because of this personalized service, virtual agents can quality as the measure of virtual agents’ performance.
respond faster and identify consumer needs more accurately during
service interactions (Baabdullah et al., 2022). Therefore, we posit that 3. Hypotheses
personalization influences response speed and response accuracy.
3.1. Response speed
2.4.2. Contextualization of the viability dimension
Within the FVM, viability reflects the extent to which the organiza­ Response speed refers to the time consumers perceive it takes to
tional environment is ready for the application, encompassing general actively complete services via virtual agents (Collier and Sherrell, 2010).
economic feasibility, technical infrastructure, and social readiness Response speed is one of the major considerations when introducing
(Liang et al., 2007). We abstract convenience, accessibility, and tech­ virtual agents into customer service settings (Binder et al., 2017; Chen
nology readiness from the viability dimension after analyzing environ­ et al., 2021a). For instance, in an online shopping context, when a
mental support for the process of consumer interaction with virtual consumer needs more information than the web display provides, they
agents. First, economic feasibility includes economic costs and benefits try to use the dialog window to consult a sales agent for more infor­
(Liang and Wei, 2004). Compared with using other agents for customer mation. Once the consult button is triggered, virtual agents need to
service, consumers do not need to cover costs and efforts associated with respond to the consumer’s requirements quickly (Cheng and Jiang,
choosing the place or time. Conversely, consumers can initiate an 2021), and their quick response speed can enhance the customer’s
interaction with virtual agents at any time and anywhere. This provides experience and perception of service quality. Accordingly, Brandtzaeg
convenient conditions for consumers to encounter services. Thus, we and Folstad (2017) found that more consumers tend to use virtual agents
regard convenience as the general economic feasibility of using a virtual if virtual agents help consumers obtain information quickly, and a sur­
agent. Second, technical infrastructure denotes technical environment vey found that more than half of baby boomers and millennials state that
support, providing the necessary foundation to support technological virtual agents should provide instant responses to their questions
operations. Virtual agents facilitate the accessibility of information or (Sweezey, 2018). In the context of customer service provided by virtual
services for the consumer. When a new virtual agent is easy to access, agents, we hypothesize that:
this indicates that the new service agent is ready and able to perform H1: Response speed has a positive influence on virtual agents’ service
service delivery (Cheng and Jiang, 2021); that is, it indicates that the quality.
technical environment of the service encounter is feasible. Thus, we
abstract accessibility to measure technical infrastructure in the context 3.2. Response accuracy
of virtual agent services. Finally, the readiness of the individual is
another crucial supporting factor for the success of the system. Con­ In the context of virtual agents for customer service, response ac­
sumers’ technology readiness for virtual agents, involving experiences, curacy is defined as the extent to which a consumer perceives a virtual
beliefs, and technical abilities, is a key factor affecting the consumer agent service provides precise information or services that suit con­
experience (Roy et al., 2018) and service quality (Zeithaml et al., 2002). sumers’ needs (Lee and Benbasat, 2011; Chan et al., 2021). If virtual
Thus, we abstract technology readiness as another variable in the agents can offer accurate information or services that satisfy consumers’
viability dimension. needs, consumers will perceive less uncertainty and obtain required
Moreover, consumers’ abilities and attitudes toward technology in­ services or solve their problems efficiently, which would increase con­
fluence their perceptions of using the technology. Consumers with some sumers’ perceived value. Consistent with this view, reliability, reflecting
degree of knowledge and a positive attitude toward new technologies the ability to perform the promised service dependably and accurately,
are more likely to adapt to the interface and functionality of virtual is one of the most important constructs for predicting an individual’s
agents, making it easier for them to navigate and interact with virtual perception of service quality (Parasuraman et al., 1988). In the domain
agents seamlessly. Additionally, if consumers are ready to use virtual of e-government service, Chan et al. (2021) also identified accuracy as
agents, they are more likely to initiate services as needed, rather than an important factor influencing service quality. Thus, we hypothesize:
waiting for human employee services, making the overall experience H2: Response accuracy has a positive influence on virtual agents’
more convenient. Thus, we argue that technology readiness influences service quality.
convenience and accessibility.
3.3. Personalization
2.4.3. Research model
By comparing virtual agents to traditional agents, we discovered that Personalization reflects the degree to which the virtual agents’ of­
virtual agents possess unique interaction styles, which determine that ferings are tailored to meet heterogeneous customers’ needs and pref­
factors influencing service quality should consider the matching of erences (Coelho and Henseler, 2012). A synonym for personalization is
technology, service, and consumer. Therefore, we proposed the FVM, customization. Personalization is the most important advantage of the
which evaluates technology performance by examining the match be­ virtual agent in customer service. Alt et al. (2019) indicated that during
tween technology, service scenario, and user readiness, as a suitable customer service, the supply side (virtual agents) and demand side
framework for exploring service quality. With the assistance of the FVM, (customers) should be matched, projecting customer-oriented service as
we integrated the technical characteristics of virtual agents with the future pathway for improving service. In contrast with human em­
customer service scenarios to abstract the fit dimension (i.e., personal­ ployees, virtual agents use natural language processing technology to
ization, response speed, and accuracy) in the “Contextualization of the provide personalized services based on the user’s input information or
fit dimension” section. Additionally, we considered the match between personal portrait. These personalized services can help consumers find
the application of virtual agents and its associated consumer to abstract the proper products or information in a timely manner, increasing their

4
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Fig. 1. Research model.

perception of speed. As Radziwill and Benton (2017) indicated, virtual that accessibility is a vital factor that facilitates service perception and
agents can meet diverse needs and have been used by companies to has a positive effect on service quality in e-government services. Thus,
reduce time-to-response. Thus, we speculate that personalization is we hypothesize that:
related to the response speed of virtual agents. Therefore, we H5: Accessibility has a positive influence on virtual agents’ service
hypothesize: quality.
H3a: Personalization has a positive influence on response speed.
Personalization afforded by virtual agents provides consumers with 3.6. Technology readiness
more valuable information or services (Cheng and Jiang, 2021). For
instance, using consumers’ browsing history or purchase history, virtual In this study, technology readiness not only reflects the consumer’s
agents can access individuals’ preferences and recommend products or ability to use the virtual agent but also includes the consumer’s beliefs or
after-sale services to them. Previous studies have consistently found that attitudes about virtual agents. We conceptualize technology readiness as
such personalization of virtual agents has a positive effect on the accu­ the extent to which a consumer has the ability and propensity to use
racy of communication (Chung et al., 2020). Therefore, we hypothesize: virtual agents for customer services (Parasuraman, 2000; Venkatesh and
H3b: Personalization has a positive influence on response accuracy. Bala, 2012). It also reflects an individual’s feelings or beliefs about high-
tech products and services and their ability to use them. Technology
3.4. Convenience readiness is conducive to a consumer’s use of a virtual agent to achieve a
task, and it offers them convenience. For example, if a consumer is ready
In this study, convenience refers to consumers’ perceived time and to use a virtual agent and views it as an efficient tool to gain information
effort required to find and facilitate the use of virtual agents for or solve a problem, they are more inclined to choose a virtual agent.
customer service (Collier and Sherrell, 2010). It also reflects the cus­ They can then communicate with the virtual agent at a convenient time
tomer’s ability to dictate the service experience. Virtual agents usually and place, which can save them time and effort and lead them to
have preset procedures and Q&As contributing to solving customers’ perceive that the customer service they received was convenient. Pre­
problems. For example, through operation procedures or recommen­ vious research has revealed that technology readiness influence con­
dations, virtual agents can deliver information or services that aid sumers’ cognitive capability, leading them to take advantage of new
consumers in decision-making. The convenience of virtual agents thus technologies to acquire necessary services conveniently (Ferreira et al.,
enhances the customer’s experience of obtaining a helpful response. 2014; Roy et al., 2018). Previous research has also demonstrated that
Previous research has found that convenience significantly influences technology readiness (i.e., personal computer skills) plays a vital role in
service quality in the e-business environment (Srinivasan et al., 2002; service convenience (Udo et al., 2010). Thus, we hypothesize:
Udo et al., 2010). Convenience is also regarded as a critical driver of a H6a: Technology readiness has a positive influence on convenience.
virtual agent’s performance (Jang et al., 2021). Thus, it is hypothesized When a consumer is ready to adopt new technology, they can deal
that: with challenges and discomforts that may arise during technology use.
H4: Convenience has a positive influence on virtual agents’ service Thus, high levels of technology readiness make it easier for consumers to
quality. access technology (Jin, 2020). When considering service quality de­
livery through websites, technology readiness was identified as one
3.5. Accessibility customer-specific construct to understand customer behaviors (Zeithaml
et al., 2002). In the context of virtual agents, technology readiness af­
Accessibility refers to the ease with which information and services fords consumers the capability of understanding the functions of virtual
can be accessed through virtual agents in the customer service context agents and how a virtual agent can help them accomplish their shopping
(Wixom and Todd, 2005). Accessibility is a key driver of the successful goals and tasks. This capability makes it easier for consumers to access
use of a new system or technology. If the system or technology is easy to virtual agents for information or services (Roy et al., 2018). Thus, it is
access, individuals are likely to use it successfully. When a virtual agent hypothesized that:
with a simple operation process increases service accessibility, the H6b: Technology readiness has a positive influence on accessibility.
consumer has a positive communication and interaction experience.
Previous research has indicated that the accessibility of a virtual agent is
a key dimension of its communication quality and thus influences its
performance (Cheng and Jiang, 2021). Chan et al. (2021) also revealed

5
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

4. Methodology (Richardson et al., 2009). Two models were constructed in Mplus soft­
ware: the baseline model and the ULMC model. The findings showed
4.1. Samples and data collection that fit indices of the ULMC model were not significantly different
compared to the baseline model (Δχ2 = 4.886, Δdf = 8, p > 0.05,
As our study focuses on virtual agents in customer service, we chose a ΔRMSEM = - 0.009, ΔCFI = 0.008, ΔTLI = 0.012), indicating that there
context in which virtual agents provide customer service on Chinese e- is no serious common method bias.
commerce platforms (e.g., Taobao, JD.com, Suning). First, the re­
spondents on those platforms are representative, as the number of online 4.3. Development of measurement instrument
shopping platform users in China reached 840 million as of June 2022
(CNNIC, 2022). Second, the respondents are familiar with the research All survey items were measured using a 7-point Likert scale ranging
context, as 42.3% of Chinese e-commerce platform users used virtual from “strongly disagree (1)” to “strongly agree (7).” All measures were
agents for customer service in this context in 2021 (iiMedia Research, based on existing validated scales and were modified to fit the research
2021). context. The measurement of response speed was derived from Collier
An online survey was performed to collect data through Sojump and Sherrell’s (2010) research. A four-item scale taken from Lee and
(http://www.sojump.com). We targeted people who have experienced Benbasat (2011) was used to measure response accuracy. Personaliza­
customer services through virtual agents. To encourage all respondents tion was assessed with a three-item scale extracted from Gattiker and
to actively and seriously participate in the investigation, we provided Goodhue (2005) and Coelho and Henseler (2012). Convenience was
those who successfully completed the questionnaire with an opportunity tested using a scale proposed by Collier and Sherrell (2010). A three-
to win a power bank worth RMB 49. Data were collected in China from item scale proposed by Wixom and Todd (2005) was adopted to mea­
March 18 to April 15, 2021, and 427 completed responses were ob­ sure accessibility. The two items for technology readiness were adapted
tained. After we eliminated questionnaires completed in under 60 sec­ from Venkatesh and Bala (2012), and another item that reflected con­
onds and failed to recognize the reverse items, 399 valid responses were sumers’ technology propensity was added. The measurement of service
retained for use in the analysis. Sample demographics are shown in quality of virtual agents consisted of four items from Taylor and Baker
Table 2. Of the respondents, 219 (54.89%) were female and 180 (1994).
(45.11%) were male. The largest age group was 25-34 years (177, To ensure the reliability and validity of the measure, we conducted
44.36%), followed by the 35-45 age group (103, 25.81%). The gender multiple rounds of assessments before the formal survey. First, we asked
and age distributions are similar to that of CNNIC (2022). In terms of two professors, two Ph.D. candidates, and three potential respondents to
frequency of virtual agent use, 38.35% (153) of the respondents used evaluate the understandability and clarity of the modified items. Then,
virtual agents 1-2 times per month. all items were translated from English to Chinese using the back-
translation technique (McGorry Susan, 2000), and minor adjustments
4.2. Response bias and common variance bias were made to ensure they would be easily and accurately understood.
Second, a pilot test was conducted with thirty students experienced in
The time trend method was used to check for non-response bias. virtual agent services. After analyzing the data, the results demonstrated
Thus, we used the first 50 and the last 50 observations to assess the that the scales obtained acceptable reliability and validity. During this
difference. The results showed that the significant value in all variables process, we also assessed the respondents’ process of answering the
exceeds 0.05, which means there is no statistically significant difference survey, checking whether there were any steps or questions that might
between early and late participants and indicates that the issue of non- have caused confusion or distress. The items are shown in Appendix 1.
response bias is not a serious concern.
Considering that our data were self-reported by individuals and 5. Results
gathered at the same point in time, two methods were utilized to check
for common method variance (CMV). First, Harman’s single-factor test 5.1. Measurement model
was used to examine common method bias (Malhotra et al., 2006). The
results showed that all the scale items are classified into seven factors Confirmatory factor analysis was used to establish a seven-factor
with eigenvalues greater than 1.0, and the cumulative variance expla­ correlated measurement model by using MPLUS 7.0. An inspection of
nation rate is 76.66%. Unrotated principal components analysis showed the model showed good overall fit indexes (χ2 (384.384)/df (254) =
the variance explanation rate of the first factor is 27.31%, which is far 1.513, CFI = 0.978, TLI = 0.974, RMSEA = 0.036 and SRMR = 0.045),
less than the threshold of 50% (Podsakoff et al., 2003). These results indicating an acceptable fit to the data. The internal reliability,
showed that common method bias is also not a serious concern in this convergent validity, and discriminant validity were demonstrated by the
study. Second, we employed the unmeasured latent method construct measurement model. As shown in Table 3, Cronbach’s α of each scale
(ULMC) approach to examine the potential impact of CMV on the and the composite reliability (CV) of variables exceed the suggested
observed relationships among the indicators of the measurement models threshold of 0.7 (Hair et al., 2012), indicating that all scales are inter­
nally consistent and reliable. The convergent validity was assessed by
Table 2 examining the standardized factor loadings and average variance
Sample characteristics. extracted (AVE). Table 3 indicates that each factor loading of the items
Variable Value Frequency (N= Percentage
with the underlying constructs is greater than 0.6 and significant at the
399) (%) 0.001 level. The values of AVE from each construct are all greater than
the recommended 0.5 (Fornell and Larcker, 1981; Sparkman, 1979).
Gender Female 219 54.89
Male 180 45.11 Thus, convergent validity is established among these constructs. In
Age Below 25 99 24.81 addition, to examine discriminant validity, we compared the root of AVE
25–34 177 44.36 values and correlation analysis results. As Table 4 shows, the discrimi­
35–45 103 25.81 nant validity is confirmed, as the square root of AVE for each construct
More than 45 20 5.02
Frequency of using virtual [1, 2] times 153 38.35
exceeds the correlations between the construct and other constructs
agents (per month) [3, 5] times 133 33.33 (Fornell and Larcker, 1981).
[6, 10] times 89 22.31
More than 10 24 6.01
times

6
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Table 3
Cronbach’s alpha, composite reliability, factor loading, and average variance extracted.
Construct Item Mean S.D. Factor loadinga Average variance extracted Cronbach’s α Composite reliability

Response speed (RS) RS1 3.125 1.389 0.898 0.769 0.929 0.930
RS2 3.143 1.383 0.903
RS3 3.130 1.419 0.890
RS4 3.241 1.294 0.814
Response accuracy (RA) RA1 3.985 1.354 0.714 0.507 0.802 0.804
RA2 4.271 1.281 0.744
RA3 3.466 1.295 0.722
RA4 3.576 1.442 0.665
Personalization (PER) PER1 4.697 1.455 0.791 0.699 0.873 0.875
PER2 4.900 1.468 0.860
PER3 4.960 1.431 0.856
Convenience (CON) CON1 5.574 1.470 0.782 0.649 0.880 0.881
CON2 4.982 1.395 0.761
CON3 5.271 1.417 0.830
CON4 5.165 1.413 0.847
Accessibility (ACC) ACC1 4.506 1.274 0.732 0.673 0.857 0.860
ACC2 4.594 1.362 0.887
ACC3 4.556 1.344 0.834
Technology readiness (TR) TR1 4.732 1.296 0.833 0.729 0.890 0.890
TR2 4.779 1.321 0.846
TR3 4.987 1.337 0.881
Service quality (SQ) SQ1 4.341 1.475 0.871 0.668 0.887 0.889
SQ2 4.386 1.487 0.855
SQ3 4.025 1.553 0.809
SQ4 4.436 1.430 0.726

Note: a. χ2 (384.384) / df (254) =1.513, CFI = 0.978, TLI = 0.974, and RMSEA = 0.036, SRMR = 0.045;b. ***p <0.001

Table 4
Discriminant validity.
Construct 1 2 3 4 5 6 7

Response speed 0.877

Response accuracy -0.120 0.712
Personalization -0.347 0.374 0.836
Convenience -0.236 0.232 0.315 0.806
Accessibility -0.253 0.181 0.255 0.265 0.820
Technology readiness -0.257 0.359 0.320 0.441 0.346 0.854
Service quality 0.069 0.471 0.273 0.406 0.307 0.282 0.817

Note(s): The figures in the subdiagonal are correlation coefficients, which except for the italic figures are all significant at the 0.010 level, and the bold figures in the
diagonal represent the square root of the average variance extracted (AVE).

5.2. Structural model assessment 5.3. Mediation examinations

A structural equation model was applied to test the hypotheses in To analyze the mediating role of response speed, response accuracy,
MPLUS 7.0. It generated a satisfactory model fit with χ2 (449.217)/df convenience, and accessibility, we followed a procedure proposed by
(266) = 1.689, CFI = 0.969, TLI = 0.965, RMSEA = 0.042, and SRMR = Baron and Kenny (1986) and built parallel multiple mediator models
0.081. The fit index, explanatory power, standardized path coefficients, using MPLUS 7.0 (Hayes, 2018) (see Table 5). First, the results of model
and associated T-values of the structural model are reported in Fig. 2. 1 indicate the significant effects of personalization (technology readi­
We observe that response speed (β = 0.240, p < 0.001) and response ness) (independent variable (IV)) on service quality (dependent variable
accuracy (β = 0.402, p < 0.001) are positively associated with service (DV)). After adding response speed and response accuracy (convenience
quality. Thus, H1 and H2 are supported. Personalization significantly and accessibility) (mediating variable (MV)), the coefficient value be­
and positively influences response accuracy (β = 0.391, p < 0.001) while tween personalization (technology readiness) and service quality (DV)
it significantly and negatively influences response speed (β = - 0.351, p decreases and is not significant (see model 2), while the 95% confidence
< 0.001); thus, H3a is not supported and H3b is supported. Convenience intervals of indirect effects based on bootstrap samples do not include
(β = 0.326, p < 0.001) and accessibility (β = 0.223, p < 0.001) are zero (see model 3). Therefore, the relationship between personalization
positively associated with service quality, supporting H4 and H5. (technology readiness) and service quality is fully mediated by response
Finally, technology readiness is positively related to both convenience (β speed and response accuracy (convenience and accessibility).
= 0.452, p < 0.001) and accessibility (β = 0.358, p < 0.001), indicating
that H6a and H6b are supported. 6. Discussion
To test the robustness of the results, we added control variables (i.e.,
gender, age, and frequency of using virtual agents) to the model. The 6.1. Major findings
subsequent analysis indicated that gender, age, and frequency of using
virtual agents had no statistically significant influences on service This study examined the antecedents and mechanisms of virtual
quality, and the identified hypothesized relationships remained signifi­ agents’ service quality in the context of e-commerce customer service,
cant. It strengthened the overall credibility and generalizability of our thus enhancing the understanding of the new service model afforded by
findings. virtual agents. Our findings showed that response speed, response ac­
curacy, convenience, and accessibility have direct influences on service

7
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Fig. 2. Results of the structural model.

Table 5
Results of the mediation effects.
Construct Model 1 Model 2 Indirect effect of IV→MV→DV

IV→DV IV+MV→DV Index BootLLCI BootULCI

Personalization 0.201*** 0.080 Fully mediated

Technology readiness 0.217*** -0.018 Fully mediated
Response speed 0.263*** -0.093*** -0.133 -0.052
Response accuracy 0.369*** 0.142*** 0.083 0.201
Convenience 0.309*** 0.143*** 0.091 0.195
Accessibility 0.211*** 0.078** 0.037 0.119
χ2 32.087 384.384 446.970
df 32 254 264
χ2/df 1.003 1.513 1.693
CFI 0.999 0.978 0.969
TLI 0.999 0.974 0.965
RMSEA 0.003 0.036 0.042
SRMR 0.024 0.045 0.080

Note(s): a. IV=independent variable, MV=mediating variable, DV=dependent variable; b. ***p < 0.001, **p < 0.01; c. Indirect effect is generated by Model 3 with IND
command; d. Number of bootstrap samples for bias corrected bootstrap confidence intervals is 5,000.

quality, and personalization and technology readiness are indirectly context of retail marketing (Chung et al., 2020). It also suggests that
related to service quality. These findings provide empirical evidence to virtual agents utilizing data analysis and deep learning technology have
verify that it is appropriate to understand the antecedents of service the potential to rival human employees in providing personalized ser­
quality from the dimensions of fit and viability. First, two variables vices, thereby increasing response accuracy (Song et al., 2022). How­
(response speed and response accuracy) from the fit dimension have ever, the hypothesis regarding the positive relationship between
positive influences on service quality of virtual agents. It means that personalization and response speed is not supported; instead, person­
when virtual agents satisfy consumers’ service requirements quickly and alization has a significant and negative influence on response speed. A
accurately, consumers will think highly of such services provided by possible reason for this may be that current technology has not yet
virtual agents. These results are close to those of Baabdullah et al. reached the stage of “strong artificial intelligence” (Huang and Rust,
(2022), who proved the significant role of responsiveness in customers’ 2018). In order to better satisfy consumer needs, virtual agents may
experience with chatbots. These findings indicate that service speed and require more time to analyze and match data to provide personalized
accuracy are the same key factors of the new service model as the solutions, consequently reducing consumers’ perception of speed. These
traditional service model (especially driven by human staff) (Luo et al., findings reveal that personalization implemented by virtual agents is a
2019). double-edged sword, requiring more attention when increasing the level
Secondly, the findings demonstrate that personalization is positively of personalization.
associated with response accuracy, while it exerts reversed influences on Thirdly, the findings elucidate the direct and indirect relationships
response speed. Specifically, the positive relationship between person­ between constructs in the viability dimension and service quality. Spe­
alization and response accuracy indicates that personalized agents cifically, convenience and accessibility exhibit direct positive influences
enhance communication accuracy, supporting previous research in the on the service quality of virtual agents in e-commerce customer service.

8
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

This emphasizes the critical role of environmental factors in promoting agents’ service quality in the context of e-commerce. First, managers and
service operability, affirming that service availability contributes to operators should pay attention to response speed, response accuracy,
service quality, whether conducted by new agents (i.e., virtual agents) or convenience, and accessibility when assessing virtual agents’ perfor­
traditional agents (Stamenkov and Dika, 2019; Zeithaml et al., 2002). mance, especially the service quality of the new business model. They
Moreover, the findings indicate that technology readiness influences should take care of customers’ experience by reducing consumer waiting
accessibility and convenience. This implies that customers who actively time, improving the accuracy of services, and providing a convenient
embrace new technology and possess the capability to adopt it are more and accessible operating environment when embedding virtual agents
likely to leverage the feasible conditions offered by virtual agents. These into the service delivery process. Second, organizations and developers
findings align closely with those of Roy et al. (2018), indicating that should note the complex effects of personalized service strategies. On
consumers, when ready and capable of using networks of smart or one hand, service providers can leverage more advanced artificial in­
intelligent objects and devices for retail services, perceive greater ease of telligence technologies to enhance the responsiveness of virtual agent
use. Additionally, mediation effects of convenience and accessibility services. On the other hand, personalization may reduce consumers’
suggest that individuals’ technology readiness influences service quality perception of service response speed to avoid misleading customers. A
through their perceptions of operational conditions, extending the trade-off between speed and accuracy should be made when increasing
research conducted by Goutam et al. (2022), which only explored the the level of personalization. Third, consumers’ readiness for virtual
direct relationship between technology readiness and service quality. agents should be emphasized. Operators and developers can provide
more guiding or explanatory language for virtual agents to increase
6.2. Theoretical contributions consumers’ levels of optimism and familiarity toward the new agents.
Thus, virtual agents can become more readily accepted in customer
This study offers several significant theoretical contributions. First, it service settings.
is one of the earlier studies that pioneered the idea of the service quality
of virtual agents as a new kind of service principal in the context of e- 6.4. Limitations and future research
commerce. Diverging from prior studies that predominantly focused on
examining antecedents of acceptance, customer experience, satisfaction, Despite this study’s numerous significant contributions, it is essential
and engagement (Zhang et al., 2021; Araujo, 2018; Schuetzler et al., to recognize its limitations, offering avenues for consideration in future
2020), this study significantly contributes by placing a greater emphasis research. The first limitation is the low level of external validity due to
on service quality with virtual agents. This enriches the literature on the sample composition (i.e., it comprises only consumers in China).
virtual agents and enhances the understanding and optimization of the Further research is necessary to verify the results obtained in different
new service model afforded by virtual agents. This study also opens up a cultures and countries. Second, this study only obtained cross-sectional
novel avenue for researchers to advance the effectiveness of virtual data. Future longitudinal study designs can lead to a better under­
agent applications. standing of the influencing factors of service quality in different use
Secondly, this study enhances the understanding of the service stages. Third, the antecedents in this study explain 38.4% of the variance
quality of customer service virtual agents from a new perspective. In in service quality. Future research could examine more potential vari­
contrast to a few studies that solely examined drivers of service quality ables (e.g., anthropomorphism, customer interaction) that may drive the
from the perspective of technical features, neglecting process factors and service quality of virtual agents in the customer service context.
available resources (Chung et al., 2020; Haugeland et al., 2022), this
study identifies the antecedents of service quality by integrating tech­ Funding
nology, service scenario, and individual readiness in the context of
customer service virtual agents. It provides a vital supplement to pro­ This work was supported by the Ministry of Education of China; the
mote the understanding of service quality. Humanities and Social Sciences Project (No. 22YJC630068); the Anhui
Thirdly, the exploration of internal relationships among service Philosophy and Social Science Planning Project (No. AHSKQ2022D066);
quality antecedents contributes to a deeper understanding of theoretical the University Excellent Young Scholar Project of Anhui Province
mechanisms and pathways of service quality. Previous research has (2022AH030096) and the National Natural Science Foundation of China
primarily focused on the parallel effects of agent service characteristics, (No.72131006).
technological availability, and consumer individual factors on service
quality, neglecting to delve into the internal relationships. This study CRediT authorship contribution statement
explores the relationship between personalization and response speed
and accuracy. Additionally, it emphasizes the influence of technology Yanping Zhang: Writing – original draft, Visualization, Software,
readiness on convenience and accessibility. By revealing the complex Investigation, Formal analysis, Data curation. Changyong Liang: Su­
relationships among antecedent variables, this study provides insights pervision, Project administration, Conceptualization. Xiaodong Li:
into service quality theory and advances theoretical development. Writing – review & editing, Validation, Methodology, Investigation,
Fourthly, this study extends the application scope of FVM by Funding acquisition, Conceptualization.
exploring the antecedents of service quality of virtual agents in e-com­
merce customer service. By applying FVM as the theoretical foundation, Declaration of competing interest
this study contextualizes the fit and viability dimensions to elucidate the
drivers of performance (i.e., service quality) (Liang et al., 2007; Zhang The authors declare that they have no known competing financial
et al., 2020). The findings indicate that the FVM can be a suitable interests or personal relationships that could have appeared to influence
framework to verify the importance of the degree to which the technical the work reported in this paper.
characteristics meet customer service tasks and the individual environ­
ment is ready for applications in the customer service context. Similar to Data availability
the research of Zhang et al. (2020), this study also enriches the appli­
cation of FVM at the consumer level. Data will be made available on request.

6.3. Practical implications

This study is conducive to understanding and improving virtual

9
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Appendix 1. . Factor items

Construct Item References

Response speed 1. I interact with virtual agents without spending too much time. (Collier and Sherrell, 2010)
2. Virtual agents respond to my request quickly.
3. Virtual agents complete an interaction quickly.
4. It is efficient to interact with virtual agents.
Response accuracy 1. The answers/services provided by virtual agents suit my requirements. (Lee and Benbasat, 2011)
2. The answers/services provided by virtual agents best match my needs.
3. The answers/services provided by virtual agents do not match my requests. (R)
4. I agree with and choose answers/services provided by virtual agents.
Personalization 1. Virtual agents can offer me the information/services that satisfy my specific needs. (Gattiker and Goodhue, 2005), (Coelho and Henseler,
2. Virtual agents are altered to improve their fit with me. 2012)
3. Virtual agents can’t provide personalized information or services for my personal needs.
(R)
Convenience 1. Virtual agents allow me to initiate an interaction whenever I choose. (Collier and Sherrell, 2010)
2. Virtual agents allow me to initiate an interaction at a convenient place.
3. I value the ability to initiate an interaction that is comfortable to me.
4. I like the ability to initiate an interaction conveniently.
Accessibility 1. Virtual agents allow information/services to be readily accessible to me. (Wixom and Todd, 2005)
2. Virtual agents make information/services very accessible.
3. Virtual agents make information/services easy to access.
Technology 1. I am ready to embrace interaction with virtual agents. (Venkatesh and Bala, 2012)
readiness 2. I have the IT infrastructure that I need to interact with virtual agents.
3. I have enough expertise to interact with virtual agents.
Service quality 1. I believe that the general service quality of virtual agents is high. (Taylor and Baker, 1994)
2. Overall, I consider virtual agents’ services to be excellent.
3. The service quality of virtual agents is generally excellent.
4.The service quality of virtual agents is generally poor (R).

References Cheng, Y., Jiang, H., 2021. Customer-brand relationship in the era of artificial
intelligence: understanding the role of chatbot marketing efforts. J. Prod. Brand
Manag. 31 (2), 252–264. https://doi.org/10.1108/JPBM-05-2020-2907.
Adam, M., Wessel, M., Benlian, A., 2021. AI-based chatbots in customer service and their
Chung, M., Ko, E., Joung, H. and Kim, S.J. (2020), “Chatbot e-service and customer
effects on user compliance. Electron. Mark. 31 (2), 427–445. https://doi.org/
satisfaction regarding luxury brands”, Journal of Business Research, Vol. 117, pp. 587-
10.1007/s12525-020-00414-7.
595, https://doi. org/10.1016/j.jbusres.2018.10.004.
Alt, R., Ehmke, J., Haux, R., Henke, T., Mattfeld, D., Oberweis, A., Paech, B., Winter, A.,
CNNIC (2022), “CNNIC: The 50th statistical report on the development of internet in
2019. Towards customer-induced service orchestration - requirements for the next
China”, China Internet Network Information Center, available at: http://www.cnnic.
step of customer orientation. Electron. Mark. 29, 79–91. https://doi.org/10.1007/
net.cn/n4/2022/0914/c88-10226.html (Accessed 13 October 2022).
s12525-019-00340-3.
Coelho, P.S., Henseler, J., 2012. Creating customer loyalty through service
Araujo, T., 2018. Living up to the chatbot hype: The influence of anthropomorphic
customization. Eur. J. Mark. 46 (3/4), 331–356. https://doi.org/10.1108/
design cues and communicative agency framing on conversational agent and
03090561211202503.
company perceptions. Comput. Hum. Behav. 85, 183–189. https://doi.org/10.1016/
Collier, J.E., Sherrell, D.L., 2010. Examining the influence of control and convenience in
j.chb.2018.03.051.
a self-service setting. J. Acad. Mark. Sci. 38 (4), 490–509. https://doi.org/10.1007/
Ashfaq, M., Yun, J., Yu, S.B., Loureiro, S.M.C., 2020. I, Chatbot: Modeling the
s11747-009-0179-4.
determinants of users’ satisfaction and continuance intention of AI-powered service
Dabholkar, P.A., Thorpe, D.I., Rentz, J.O., 1996. A measure of service quality for retail
agents. Telematics Inform. 54, 101473 https://doi.org/10.1016/j.tele.2020.101473.
stores: Scale development and validation. J. Acad. Mark. Sci. 24 (1), 3–16. https://
Baabdullah, A.M., Alalwan, A.A., Algharabat, R.S., Metri, B., Rana, N.P., 2022. Virtual
doi.org/10.1177/009207039602400101.
agents and flow experience: An empirical examination of AI-powered chatbots.
Das A.C., Phalin G., Patidar I.L., Gomes M., Sawhney R. and Thomas R. (2023), “The next
Technol. Forecast. Soc. Chang. 181, 121772 https://doi.org/10.1016/j.
frontier of customer engagement: AI-enabled customer service”, available at:
techfore.2022.121772.
https://www.mckinsey.com/capabilities/operations/our-insights/the-next-frontier-
Baron, R.M., Kenny, D.A., 1986. The moderator mediator variable distinction in social
of-customer-engagement-ai-enabled-customer-service (Accessed 1 April 2023).
psychological research: Conceptual, strategic, and statistical considerations. J. Pers.
DATA CONOMY (2022), “The rise of artificial intelligence customer services”, available
Soc. Psychol. 51 (6), 1173–1182. https://doi.org/10.1037/0022-3514.51.6.1173.
at: https://dataconomy.com/2022/07/artificial-intelligence-customer-service/
Behera, R.K., Bala, P.K., Ray, A., 2021. Cognitive Chatbot for Personalised Contextual
(Accessed 12/8 2022).
Customer Service: Behind the Scene and beyond the Hype. Inf. Syst. Front.
Fassnacht, M., Koese, I., 2006. Quality of electronic services: Conceptualizing and testing
Binder, R., P. Pellegrini, and G. Gujral (2017), “Intelligence-powered human services”,
a hierarchical model. J. Serv. Res. 9 (1), 19–37. https://doi.org/10.1177/
available at: https://www.accenture.com/_acnmedia/pdf-66/accenture-ai-in-social-
1094670506289.
services-pov-final.pdf (Accessed 19 September 2022).
Ferreira, J.B., da Rocha, A., da Silva, J.F., 2014. Impacts of technology readiness on
Blut, M., 2016. E-Service Quality: Development of a Hierarchical Model. J. Retail. 92 (4),
emotions and cognition in Brazil. J. Bus. Res. 67 (5), 865–873. https://doi.org/
500–517. https://doi.org/10.1016/j.jretai.2016.09.002.
10.1016/j.jbusres.2013.07.005.
Brady, M.K., Cronin, J.J., 2001. Some New Thoughts on Conceptualizing Perceived
Fornell, C., Larcker, D.F., 1981. Evaluating structural equation models with unobservable
Service Quality: A Hierarchical Approach. J. Mark. 65 (3), 34–49. https://doi.org/
variables and measurement error. J. Mark. Res. 18 (1), 39–50. https://doi.org/
10.1509/jmkg.65.3.34.18334.
10.1177/002224378101800104.
Brandtzaeg, P.B., Folstad, A., 2017. Why people use chatbots. In: 4th International
Gattiker, T.F., Goodhue, D.L., 2005. What happens after ERP implementation -
Conference on Internet Science (INSCI). Thessaloniki, Greece, pp. 377–392.
Understanding the impact of interdependence and differentiation on plant-level
Chan, F.K.Y., Thong, J.Y.L., Brown, S.A., Venkatesh, V., 2021. Service design and citizen
outcomes, 777–777 MIS Q. 29 (4). https://doi.org/10.2307/25148695.
satisfaction with e-government services: A multidimensional perspective. Public
Goodhue, D.L., Thompson, R.L., 1995. Task-technology fit and individual performance.
Adm. Rev. 81 (5), 874–894. https://doi.org/10.1111/puar.13308.
MIS Q. 19 (2), 213–236. https://doi.org/10.2307/249689.
Chen, J.V., Le, H.T., Tran, S.T.T., 2021a. Understanding automated conversational agent
Goutam, D., Ganguli, S., Gopalakrishna, B.V., 2022. Technology readiness and e-service
as a decision aid: Matching agent’s conversation with customer’s shopping task.
quality – impact on purchase intention and loyalty. Mark. Intell. Plan. 40 (2),
Internet Res. 31 (4), 1376–1404. https://doi.org/10.1108/intr-11-2019-0447.
242–255. https://doi.org/10.1108/MIP-06-2021-0196.
Chen, J.S., Le, T.T.Y., Florence, D., 2021b. Usability and responsiveness of artificial
Grönroos, C., 1984. A service quality model and its marketing implications. Eur. J. Mark.
intelligence chatbot on online customer experience in e-retailing. Int. J. Retail
18 (4), 36–44. https://doi.org/10.1108/EUM0000000004784.
Distrib. Manag. 49 (11), 1512–1531. https://doi.org/10.1108/ijrdm-08-2020-0312.
Hair, J.F., Sarstedt, M., Ringle, C.M., Mena, J.A., 2012. An assessment of the use of
partial least squares structural equation modeling in marketing research. J. Acad.
Mark. Sci. 40 (3), 414–433. https://doi.org/10.1007/s11747-011-0261-6.

10
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Haugeland, I.K.F., Følstad, A., Taylor, C., Bjørkli, C.A., 2022. Understanding the user Radziwill, N.M. and Benton, M.C. (2017), “Evaluating quality of chatbots and intelligent
experience of customer service chatbots: An experimental study of chatbot conversational agents. ”, available at: https://ui.adsabs.harvard.edu/abs/
interaction design. Int. J. Hum Comput Stud. 161, 102788 https://doi.org/10.1016/ 2017arXiv170404579R (Accessed 23 November 2021).
j.ijhcs.2022.102788. Rapp, A., Curti, L., Boldi, A., 2021. The human side of human-chatbot interaction: A
Hayes, A.F., 2018. Introduction to mediation, moderation, and conditional process systematic literature review of ten years of research on text-based chatbots. Int. J.
analysis: A regression-based approach. The Guilford Press, New York and London. Hum Comput Stud. 151, 102630 https://doi.org/10.1016/j.ijhcs.2021.102630.
Huang, Z.W., Luo, Y., Wang, D., 2019. Online customer service quality of online Rese, A., Ganster, L., Baier, D., 2020. Chatbots in retailers’ customer communication:
shopping: Evidence from dangdang.Com. Cluster Computing-the Journal of How to measure their acceptance? J. Retail. Consum. Serv. 56, 102176 https://doi.
Networks Software Tools and Applications 22, 15285–15293. https://doi.org/ org/10.1016/j.jretconser.2020.102176.
10.1007/s10586-018-2565-5. Richardson, H.A., Simmering, M.J., Sturman, M.C., 2009. A Tale of Three Perspectives:
Huang, M.-H., Rust, R.T., 2018. Artificial Intelligence in Service. J. Serv. Res. 21 (2), Examining Post Hoc Statistical Techniques for Detection and Correction of Common
155–172. https://doi.org/10.1177/1094670517752459. Method Variance. Organ. Res. Methods 12 (4), 762–800. https://doi.org/10.1177/
iiMedia Research (2021), “Data analysis of intelligent customer service industry: In 2021, 1094428109332834.
62.4% of users in China used intelligent customer service when handling telephone Roy, S.K., Balaji, M.S., Quazi, A., Quaddus, M., 2018. Predictors of customer acceptance
charges and other services”, available at: https://www.iimedia.cn/c1061/82108. of and resistance to smart technologies in the retail sector. J. Retail. Consum. Serv.
html (Accessed 10 January 2022). 42, 147–160. https://doi.org/10.1016/j.jretconser.2018.02.005.
Jang, S.M., Hong, Y.J., Lee, K., Kim, S., Van Chien, B., Kim, J., 2021. Assessment of user San Martin, S., Lopez-Catalan, B., Ramon-Jeronimo, M.A., 2012. Factors determining
needs for telemedicine robots in a developing nation hospital setting. Telemedicine firms’ perceived performance of mobile commerce. Ind. Manag. Data Syst. 112 (5-6),
and E-Health 27 (6), 670–678. https://doi.org/10.1089/tmj.2020.0215. 946–963. https://doi.org/10.1108/02635571211238536.
Jin, C.-H., 2020. Predicting the use of brand application based on a tram. International Sands, S., Ferraro, C., Campbell, C., Tsao, H.Y., 2021. Managing the human-chatbot
Journal of Human-Computer Interaction 36 (2), 156–171. https://doi.org/10.1080/ divide: how service scripts influence service experience. J. Serv. Manag. 32 (2),
10447318.2019.1609227. 246–264. https://doi.org/10.1108/JOSM-06-2019-0203.
Kallweit, K., Spreer, P., Toporowski, W., 2014. Why do customers use self-service Scherer, A., Wünderlich, N.V. and Wangenheim, F.V. (2015), “The value of self-service:
information technologies in retail? The mediating effect of perceived service quality. Long-term effects of technology-based self-service usage on customer retention”, MIS
J. Retail. Consum. Serv. 21 (3), 268–276. https://doi.org/10.1016/j. quarterly, Vol. 39 No.1, pp. 177-200, 10.25300/misq/2015/39.1.08.
jretconser.2014.02.002. Schuetzler, R.M., Grimes, G.M., Scott Giboney, J., 2020. The impact of chatbot
Kasiri, L.A., Cheng, K.T.G., Sambasivan, M., Sidin, S.M., 2017. Integration of conversational skill on engagement and perceived humanness. J. Manag. Inf. Syst. 37
standardization and customization: Impact on service quality, customer satisfaction, (3), 875–900. https://doi.org/10.1080/07421222.2020.1790204.
and loyalty. J. Retail. Consum. Serv. 35, 91–97. https://doi.org/10.1016/j. Selamat, M.A., Windasari, N.A., 2021. Chatbot for SMEs: Integrating customer and
jretconser.2016.11.007. business owner perspectives. Technol. Soc. 66, 101685 https://doi.org/10.1016/j.
Kim, B., Chen, Y., 2023. Empowering consumers through self-service technology: a techsoc.2021.101685.
comparative analysis. J. Hosp. Tour. Res. https://doi.org/10.1177/ Servion (2020), “AI will power 95% of customer interactions by 2025”, available at:
10963480231182959. https://www.financedigest.com/ai-will-power-95-of-customer-interactions-by-
Kull, A.J., Romero, M., Monahan, L., 2021. How may I help you? Driving brand 2025.html. (Accessed 15 December 2021).
engagement through the warmth of an initial chatbot message. J. Bus. Res. 135, Shamdasani, P., Mukherjee, A., Malhotra, N., 2008. Antecedents and consequences of
840–850. https://doi.org/10.1016/j.jbusres.2021.03.005. service quality in consumer evaluation of self-service internet technologies. Serv.
Larosiliere, G.D., Carter, L.D., 2016. Using a fit-viability approach to explore the Ind. J. 28 (1), 117–138. https://doi.org/10.1080/02642060701725669.
determinants of e-government maturity. J. Comput. Inf. Syst. 56 (4), 271–279. Song, M., Xing, X., Duan, Y., Cohen, J., Mou, J., 2022. Will artificial intelligence replace
https://doi.org/10.1080/08874417.2016.1163995. human customer service? The impact of communication quality and privacy risks on
Lee, Y.E., Benbasat, I., 2011. Research note-the influence of trade-off difficulty caused by adoption intention. J. Retail. Consum. Serv. 66, 102900 https://doi.org/10.1016/j.
preference elicitation methods on user acceptance of recommendation agents across jretconser.2021.102900.
loss and gain conditions. Inf. Syst. Res. 22 (4), 867–884. https://doi.org/10.1287/ Sparkman Jr, R.M., 1979. Book review: Multivariate data analysis with readings. SAGE
isre.1100. 0334. Publications Sage CA, Los Angeles, CA.
Lee, G.G., Lin, H.F., 2005. Customer perceptions of e-service quality in online shopping. Srinivasan, S.S., Anderson, R., Ponnavolu, K., 2002. Customer loyalty in e-commerce: An
Int. J. Retail Distrib. Manag. 33 (2), 161–176. https://doi.org/10.1108/ exploration of its antecedents and consequences. J. Retail. 78 (1), 41–50. https://
09590550510581485. doi.org/10.1016/S0022-4359(01)00065-3.
Li, L., Lee, K.Y., Emokpae, E., Yang, S.B., 2020. What makes you continuously use Stamenkov, G., Dika, Z., 2019. Quo vadis, (e-)service quality? Towards a sustainability
chatbot services? Evidence from chinese online travel agencies. Electron. Mark. 31 paradigm. Total Qual. Manag. Bus. Excell. 30 (7-8), 792–807. https://doi.org/
(3), 575–595. https://doi.org/10.1007/s12525-020-00454-z. 10.1080/14783363.2017.1338521.
Liang, T.-P., Huang, C.-W., Yeh, Y.-H., Lin, B., 2007. Adoption of mobile technology in Statista (2021), “Size of chatbot in banking, financial services and insurance (bfsi)
business: A fit-viability model. Ind. Manag. Data Syst. 107 (8), 1154–1169. https:// market worldwide in 2019 and 2030”, available at: https://www.statista.com/
doi.org/10.1108/02635570710822796. statistics/1256242/worldwide-chabot-in-bfsi-revenues/ (Accessed 24/3 2022).
Liang, T.P., Kohli, R., Huang, H.C., Li, Z.L., 2021. What drives the adoption of the Sweezey, M. (2018), “The value of chatbots for today’s consumers”, Forbes
blockchain technology? A fit-viability perspective. J. Manag. Inf. Syst. 38 (2), Communications Council, available at: https://www.forbes.com/sites/
314–337. https://doi.org/10.1080/07421222.2021.1912915. forbescommunicationscouncil/2018/02/13/the-value-of-chatbots-for-todays-
Liang, T.-P., Wei, C.-P., 2004. Introduction to the special issue: Mobile commerce consumers/?sh=5fd927b72918 (Accessed 1 January 2022).
applications. Int. J. Electron. Commer. 8 (3), 7–17. https://doi.org/10.1080/ Taylor, S.A., Baker, T.L., 1994. An assessment of the relationship between service quality
10864415.2004.11044303. and customer satisfaction in the formation of consumers’ purchase intentions.
Luo, X.M., Tong, S.L., Fang, Z., Qu, Z., 2019. Frontiers: Machines vs. Humans: The impact J. Retail. 70 (2), 163–178. https://doi.org/10.1016/0022-4359(94)90013-2.
of artificial intelligence chatbot disclosure on customer purchases. Mark. Sci. 38 (6), Tjan, A.K., 2001. Finally, a way to put your internet portfolio in order. Harv. Bus. Rev. 79
937–947. https://doi.org/10.1287/mksc.2019.1192. (2), pp. 76–85, 156.
Malhotra, N.K., Kim, S.S., Patil, A., 2006. Common method variance in is research: A Udo, G.J., Bagchi, K.K., Kirs, P.J., 2010. An assessment of customers’ e-service quality
comparison of alternative approaches and a reanalysis of past research. Mark. Sci. 52 perception, satisfaction and intention. Int. J. Inf. Manag. 30 (6), 481–492. https://
(12), 1865–1883. https://doi.org/10.1287/mnsc.1060.0597. doi.org/10.1016/j.ijinfomgt.2010.03.005.
McGorry Susan, Y., 2000. Measurement in a cross-cultural environment: Survey UJET (2022), “Critical State of Automation in Customer Experience”, available at:
translation issues. Qual. Mark. Res. Int. J. 3 (2), 74–81. https://doi.org/10.1108/ https://ujet.cx/resources/reports/critical-state-of-automation-customer-experience-
13522750010322070. 2022-report-lp (Accessed 10 December 2023).
Palese, B., Usai, A., 2018. The relative importance of service quality dimensions in e- Venkatesh, V., Bala, H., 2012. Adoption and impacts of interorganizational business
commerce experiences. Int. J. Inf. Manag. 40, 132–140. https://doi.org/10.1016/j. process standards: Role of partnering synergy. Inf. Syst. Res. 23 (4), 1131–1157.
ijinfomgt.2018.02.001. https://doi.org/10.1287/isre.1110.0404.
Parasuraman, A., 2000. Technology readiness index (Tri): A multiple-item scale to Wang, Y., Lo, H.-P., Yang, Y., 2004. An integrated framework for service quality,
measure readiness to embrace new technologies. J. Serv. Res. 2 (4), 307–320. customer value, satisfaction: evidence from China’s telecommunication industry. Inf.
https://doi.org/10.1177/109467050024001. Syst. Front. 6, 325–340. https://doi.org/10.1023/B:ISFI.0000046375.72726.67.
Parasuraman, A.P., Zeithaml, V., Berry, L., 1988. SERVQUAL: A multiple- item scale for Wang, W.-T., Ou, W.-M., Chen, W.-Y., 2019. The impact of inertia and user satisfaction
measuring consumer perceptions of service quality. J. Retail. 64 (1), 12–40. on the continuance intentions to use mobile communication applications: a mobile
Pizzi, G., Scarpi, D., Pantano, E., 2021. Artificial intelligence and the new forms of service quality perspective. Int. J. Inf. Manag. 44, 178–193. https://doi.org/
interaction: Who has the control when interacting with a chatbot? J. Bus. Res. 129, 10.1016/j.ijinfomgt.2018.10.011.
878–890. https://doi.org/10.1016/j.jbusres.2020.11.006. Wirtz, J., Patterson, P.G., Kunz, W.H., Gruber, T., Lu, V.N., Paluch, S., Martins, A., 2018.
Podsakoff, P.M., MacKenzie, S.B., Lee, J.Y., Podsakoff, N.P., 2003. Common method Brave new world: service robots in the frontline. J. Serv. Manag. 29 (5), 907–931.
biases in behavioral research: a critical review of the literature and recommended https://doi.org/10.1108/JOSM-04-2018-0119.
remedies. J. Appl. Psychol. 88 (5), 879–903. https://doi.org/10.1037/0021- Wixom, B.H., Todd, P.A., 2005. A theoretical integration of user satisfaction and
9010.88.5.879. technology acceptance. Inf. Syst. Res. 16 (1), 85–102. https://doi.org/10.1287/
Pooya, A., Khorasani, M.A., Ghouzhdi, S.G., 2020. Investigating the effect of perceived isre.1050.0042.
quality of self-service banking on customer satisfaction. Int. J. Islam. Middle East. Wolfinbarger, M., Gilly, M.C., 2003. eTailQ: dimensionalizing, measuring and predicting
Financ. Manag. 13 (2), 263–280. https://doi.org/10.1108/imefm-12-2018-0440. etail quality. J. Retail. 79 (3), 183–198. https://doi.org/10.1016/S0022-4359(03)
00034-4.

11
Y. Zhang et al. Electronic Commerce Research and Applications 65 (2024) 101380

Wu, F.C.T., Hong, O.N.J., Trappey, A.J.C. and Trappey, C.V. (2020), “VR-enabled chatbot Zhang, Y., Li, X., Hamari, J., 2020. How does mobility affect social media advertising
system supporting transformer mass-customization services”,in 27th ISTE effectiveness? A study in WeChat. Ind. Manag. Data Syst. 120 (11), 2081–2101.
International Conference on Transdisciplinary Engineering (TE) - Transdisciplinary https://doi.org/10.1108/IMDS-05-2020-0268.
Engineering for Complex Socio-Technical Systems in Perspective of Real-Life Application, Zhang, S., Meng, Z., Chen, B., Yang, X., Zhao, X., 2021. Motivation, social emotion, and
Electr Network, pp. 291-300. the acceptance of artificial intelligence virtual assistants—trust-based mediating
Zeithaml, V., Parasuraman, A.P., Malhotra, A., 2002. Service quality delivery through effects. Front. Psychol. 12, 728495 https://doi.org/10.3389/fpsyg.2021.728495.
web sites: a critical review of extant knowledge. J. Acad. Mark. Sci. 30 (4), 362–375.
https://doi.org/10.1177/009207002236911.

12