“How chatbot e-services motivate communication credibility and lead to customer satisfaction: The perspective of Thai consumers in the apparel retailing context”

Many apparel retailing brands use e-service marketing tools such as a chatbot (a sys- tem that is available 24 hours a day, 7 days a week) to increase their competitive advantage in today’s world of digitalization. During the COVID-19 pandemic, chatbots gained more power to serve as a communication tool that provides information and maintains customer experience. Therefore, this study is conducted to investigate the influence of chatbot e-service agents’ marketing efforts (involving interaction, entertainment, trendiness, and problem-solving) on Thai customers’ perceived communica- tion credibility and satisfaction in apparel retailing, as research in this area is limited. In order to test the hypotheses, the paper employed structural equation modeling us- ing Amos. In addition, an online survey of 400 Thai consumers who had previously used chatbots in the apparel retailing industry was conducted. The results showed that chatbot e-service marketing efforts, including interaction, trendiness, and problem-solving, affected customer satisfaction without entertainment elements. Beyond this, a chatbot, viewing interaction and entertainment, was found to have an insignificant effect on communication credibility. Thus, the coefficient value proved that informa- tion regarding communication credibility is more dominant in customer satisfaction. Therefore, the chatbot e-service marketing effort is essential in motivating communica- tion credibility in customer satisfaction. These findings delivered managerial implica-tions for understanding consumers in the field of digitalization.


INTRODUCTION
Artificial intelligence (AI) refers to tools that support organizations in designing a service interaction line with customers. Innovative AIbased technologies have garnered significant interest in recent years, both in theoretical study and practice (Tran et al., 2021). AI is used in a wide variety of applications, including autonomous vehicles (e.g., Tesla), intelligent recommendation tools (e.g., Salesforce Einstein), and natural language treatment (e.g., chatbots). However, customers' preference for AI changes according to the service task difficulty, as they tend to be content with using AI for low-complexity tasks but prefer to deal with human agents during matters of higher complexity (Xu et al., 2020;Tran et al., 2021). In this study, which focuses on natural language processing, the chatbot is derived from a grouping of the two words "chat" (as in online communication) and "bot" (as in a robot) (Rese et al., 2020). Indeed, any software application that converses with a human using natural language is referred to as a conversational agent. Chatbots show a natural language interface that can "understand natural language and respond in natural language to a user request" (Lester & Piore, 2004, p. 220; Rese et al., 2020).
Customers had to rely on social media tools such as chatbots to obtain information and build buying choices, particularly during the COVID-19 pandemic, when consumers were quarantined at home, and human agents were unavailable. In the US market, Cheng and Jiang (2022) reported that more than 30,000 chatbots have been launched on messaging social media like the Messenger application of Facebook and Viber, while approximately 2 billion messages are sent via these services a month. Due to continued advancements in AIs' ability to mimic conversational language, chatbots are being built to connect with humans or even to take the place of human agents in digital marketing (Kumar et al., 2016;Cheng & Jiang, 2022).
Many scholars have studied AI chatbots to find ways to prove what chatbot support is when it comes to marketing efforts in luxury brands (Chung et al., 2020;Kim & Ko, 2012). Despite this, a few studies have focused on exploring the concept of chatbot e-service agents' marketing efforts, perceived communication credibility, and satisfaction. There are limited studies concerning this concept in apparel retailing and chatbot e-service marketing in Thailand. Therefore, this paper focuses on bridging the existing knowledge gap by setting two essential research objectives. They are to examine the effect of e-service agents' marketing efforts delivered by chatbots in Thai apparel retailing, and to identify how chatbots influence Thai people regarding communication credibility and customer satisfaction.

E-service agents and marketing e-service
Today, since many brands are becoming globalized, AI and digital marketing are changing service agent roles (Jansom & Pongsakornrungsilp, 2021). E-service agents are continuously accessible personal helpers that support creating vital buyer relations, maximizing customer time, and providing an improved understanding of product performance ( The present study fills the above-mentioned gap by measuring e-service agents' interaction, entertainment, trendiness, innovativeness, and problem-solving, thereby enabling a more significant examination of consumer satisfaction. Furthermore, although AI chatbots are an increasing platform in the apparel retail business, research on their use in Thailand is limited.
Therefore, the paper assesses how e-service agents can affect communication quality and customer satisfaction, as well as how this affects purchase intention for apparel retail fashion brands that use chatbots for e-service.

Interaction
At physical stores, interaction is one of the dimensions used to increase the number of brand associates. Indeed, Dabholkar et al. (1996)  Based on the current knowledge gap, interaction factors have been studied for interactive representatives both online and offline in various industries. However, perceived communication credibility and customer satisfaction with e-service agents such as chatbots have not been extensively studied, especially in the apparel retailing industry. Therefore, the purpose of this paper is to examine whether chatbot e-services can generate positive interactions in communication credibility and customer satisfaction.

Entertainment
Given that social media stimulates human entertainment, retailing brands are attempting to build customer social media experiences to increase competitive advantage (Godey et al., 2016). Successful brands understand the benefits of combining fun and amusement into their daily work (Redman & Mathews, 2002 . Although the factor of online entertainment with brands has been explored in the preceding literature, the impacts of perceived communication credibility and customer satisfaction with the entertainment of chatbot e-service agents have not been thoroughly studied. As a result, the purpose of this study is to investigate how chatbot e-services might increase communication credibility and customer satisfaction.

Trendiness
Social media platforms deliver news stories and trending debate topics (Naaman et al., 2011) while serving as a primary product search. Trendiness can build the perception of a brand since numerous consumers require up-to-date brand and product information to be reassured that products appropriately reflect their fashionable lifestyles (2020) state that trending information on social media is consumed for four distinct reasons: surveillance, knowledge, pre-purchase, and motivation. Surveillance is the act of observing and maintaining knowledge of one's social environment. Knowledge indicates brand-related communication that customers obtain to benefit from the knowledge and expertise of other consumers, gathering further information about a product or brand. In order to make well-informed purchasing decisions, consumers often conduct pre-purchase research by reading online reviews of products or threads on brand communities. Finally, inspiration relates to customers acquiring new ideas regarding brand-related information, with that information thus serving as a source of stimulation. For example, when customers look at the clothes worn by other people, they develop ideas regarding what they want to wear.
As per the above argument, trendiness is stated in this study in terms of the transmission of the most current and trending communication from the apparel context. However, in-store salespeople were originally the primary sources of fashion trends in the past, while technological advances are now allowing online and in-store encounters to coexist (Chung et al., 2020). Consequently, the purpose of this paper is to examine extensively whether trendiness of chatbot e-services can promote communication credibility and customer satisfaction, especially in apparel retailing.

Problem-solving
When consumers perceive service quality, they might be concerned with the answers provided by salespeople in a shop. According to Kim et al. (2016), service quality includes personal interaction, physical aspect, policy, problem-solving, and reliability, which are precursors to a variety of consumer emotions. Furthermore, each customer has different shopping behaviors based on motives and tends to focus on different aspects of service (Dawson et al., 1990;Kim et al., 2016). Hence, a customer evaluates a given quality measurement in a different way.
Previous studies employed problem-solving to identify the impact of chatbot e-services on consumers' experiences by assessing communication and satisfaction on social media sites. However, customers disappointed with the quality of a service or product may experience resentment and even humiliation due to feeling constrained (Izard, 1977;Chung et al., 2020). Since the way in which customers cogitate in retail service is based on how quickly and honestly, they deal with customer problems, complaints, returns, and exchanges, a retail brand's associates are often supposed to teach recruits how to deal with these issues right away sincerely (

Hypotheses
Based on the literature review, this study aims to examine the impact of chatbot e-service agents' marketing efforts in terms of interaction, entertainment, trendiness, and problem-solving on Thai customers' perceptions of communication credibility and satisfaction with apparel retailing particularly in the Thailand context. The following hypotheses are therefore proposed ( Figure 1): H1a: Chatbot e-services can provide positive interactions that evoke communication credibility.
H1b: Chatbot e-services can provide positive interactions that evoke customer satisfaction.
H2a: Chatbot e-services can provide positive entertainment that evokes communication credibility.
H2b: Chatbot e-services can provide positive entertainment that evokes customer satisfaction.
H3a: Chatbot e-services can provide positive trendiness that evokes communication credibility.
H3b: Chatbot e-services can provide positive trendiness that evokes customer satisfaction.
H4a: Chatbot e-services can provide positive problem-solving that evokes communication credibility.
H4b: Chatbot e-services can provide positive problem-solving that evokes customer satisfaction.
H5: Communication credibility of chatbots can provide positive customer satisfaction.

METHODOLOGY
The study implemented a deductive approach to test and validate the conceptual model, which included delivering a survey to gather data to investigate the hypotheses generated by the liter- AI chatbot e-service agents ature review. The data were collected in March 2022 after the study had been approved by the Ethics Committee of the Human Research at Walaialk University (WUEC-22-050-01). Validated scales from earlier literature were adapted to the current situation to create the questionnaire, and the study was carried out via an online survey utilizing Google Forms. A total of 33 questions were translated into Thai and then retranslated by marketing academics to ensure the authenticity and dependability of the information they included. The questionnaire consisted of three main parts: the first included screening questions (Is the person aged 18 or over, has he/she ever used chatbots and/or platforms that employ chatbots?), while the second covered socio-demographic characteristics of the respondents. The last part concentrated on the question items associated with the construct proposed in the model, which was used in prior studies to measure marketing efforts according to interaction, entertainment, trendiness, and problem-solving ( All items were measured using a five-point Likert scale (1 = strongly disagree to 5 = strongly agree), which helped to quantify the participants' responses. Social media channels, including Facebook groups, LinkedIn, and the Line application, were used to gather data for the study. Non-probabilistic (accidental sampling) sampling was used, with 410 Thai respondents who had personal experience interacting with chatbots in the apparel retailing industry participating in the study. After excluding invalid responses, the paper obtained a final sample of 400 surveys for the investigation. Re-reading the screening question on the platforms where respondents use chatbots, it is clear that more than 85.50% of the respondents were using one via Messenger on Facebook, and 10.50% were using chatbots through the Line applica-tion, while 4% engaged with chatbots via other platforms. As shown in Table 1, most of the respondents (58.75%) were female and aged 26-33 (37.75%). A Bachelor's degree was the highest educational level of the participants (49.25%), and the average monthly income was 10,001-20,000 Thai baht (36.75%). The respondents were mostly purchasing online apparel 3-5 times per month (38.75%) and using chatbots 2-4 times per month. Amongst the sample, 98.75% of Thai customers considered a chatbot to be a tool for motivating satisfaction with brands. An overall measure of sampling adequacy (MSA) and Bartlett's test of sphericity were used to determine whether or not the observed variables had good correlations before moving further with the factor analysis. The test findings are summarized in Table A1; the aggregate MSA calculated by KMO was 0.946, exceeding the 0.50 cut-off figure (Hair et al., 2006). This revealed that the observed variables were significantly linked in aggregate and hence suitable for factor analysis.
An Exploratory Factor Analysis (EFA) using SPSS 24.0 was performed simultaneously on all of the six constructs. The purpose was to identify how scale items were interrelated and to interpret these interrelationships using the principal axis factor analysis with a varimax rotation. The study used 23 items for factor extraction, but they were also utilized to assess the underlying dimensions of the given items. The initial EFA results showed that two items had either low factor loadings or cross-loadings; therefore, they were eliminated from the analysis (Bilgin et al., 2015, p. 209). The items were loaded under six factors (Appendix A), which explained a total of 69% of the variance. Eigenvalues for each construct were higher than 1.0, which was used as the threshold value to retain the factors. The factor loadings of all 21 items were above 0.5, ranging from 0.584 to 0.834. These examples demonstrate that the requirements for factor interpretation outlined above were met, and as a result, the factors derived from this study were appropriate. After EFA, the scales for each of the extracted factors were tested for internal consistency through Cronbach's alpha coefficient, with a suggested threshold of 0.7. The re-sults of the reliability tests are presented in Table 3. Cronbach's alphas for each construct surpassed 0.7 and ranged from 0.829 to 0.903.

Descriptive statistics
Descriptive statistics are used to analyze the model's variables in order to demonstrate how chatbots motivate communication credibility and lead to customer satisfaction, as presented in Table 3. The standard deviations were less than 1.5 with 30% of the mean; thus, the statistics were not broadly dispersed from the mean, with a range of 3.760-3.990 and standard deviations of 0.770-0.939.

Confirmatory factor analysis (CFA)
The CFA test began with a measuring model comprising all 21-scale items corresponding to the six components identified in previous EFA analyses.  CFA is used to ascertain the factor appropriateness of items and the number of dimensions in an empirical model (Nunnally & Bernstein, 1994;Jansom & Pongsakornrungsilp, 2021), as well as to identify dependent variables. CFA determines the data's fit to the empirical investigation (Bollen, 1989). The paper chose a theoretical framework for this research that included six variables: interaction, entertainment, trendiness, problem-solving, communication credibility, and customer satisfaction. The results of this model's multi-factor confirmatory study indicate that the acceptable threshold levels are compatible with those proposed by Hair et al. (1998) and Bollen (1989). Table 3 provides the CFA results in terms of how chatbots motivate communication credibility and lead to customer satisfaction. The average variance extracted (AVE) measures the variation calmed by the indicators relative to measurement error. This study discovered a range from 0.538 to 0.713, higher than 0.50. The composite reliabilities (CR) for all model concepts were over the threshold value of 0.70 (Teo et al., 2008), although the range was also more than 0.60 (0.725 to 0.909). All variable measurements were found to be acceptable, which powerfully suggests that the item set presents a single fundamental concept and provides indication of discriminatory validity.

Structural equation modeling (SEM)
SEM is a popular data analysis method used by scholars across various disciplines for maximum likelihood estimation. Therefore, the study analyzed the constructed model using SEM analysis to test the relationships. Once the overall model fit had been approved, the significance of individual path coefficients was examined, as this was the input for hypothesis testing. Commonly, the chi-square value is used to evaluate the overall fit of a model and to estimate the magnitude of the disagreement between the sample and the fitted covariance matrices (Hu & Bentler, 1999). Barrett (2007), Hair et al. (1998), and Bollen (1989) Table 4 shows the findings of the hypothesis testing for the fit confirmation of the model describing how chatbots motivate communication credibility and lead to customer satisfaction, summarizing the path coefficients and the hypotheses in the SEM analysis.

DISCUSSION
H1 displayed that the chatbot e-service can provide interaction in apparel retailing, which is H1a, a chatbot with positive interaction evokes satisfaction with a regression weight estimate of standardized coefficients of 0.406, t-value 3.576, and Sig.0.000 < 0.05 (H1a is supported, Sig = 0.000*). The result matches with Holzwarth et al. (2006). They studied an avatar on web-based retail sales, indicating that customers will be satisfied with interactions when they make decisions to purchase and save time with an adviser while looking for products. Moreover, Kim and Ko (2010) also supported the notion that chatbot technologies can provoke positive interactions with customers. Meanwhile, H1b states that chatbot  (2021), who discovered that chatbot interactions motivate communication credibility among customers. However, the study showed that Thai respondents do not believe in information communication provided via e-services, as they instead like to interact with salespeople face-to-face. Dabholkar et al. (1996) supported the finding that brand associates in a physical store can build customer interaction and make people see the brand as courteous, helpful, and trustworthy. Thus, although Thai respondents use chatbots for interaction without wanting information credibility while communicating, interaction with chatbots can lead to satisfaction.
Interestingly, the finding for H2 showed that chatbot e-services with entertainment elements in apparel retailing show that both H2a and H2b are unsupported on communication credibility and customer satisfaction. Thai respondents did not experience entertainment of the chatbot that provides information when using e-service. The above discussion emphasizes the importance of studying chatbot e-services in apparel retailing. When Thai customers engage with communication credibility and customer satisfaction, they are concerned about many e-service marketing efforts, such as using chatbots for interaction, trendiness, and problem-solving without any entertainment factor. Therefore, the study achieved its primary objective by examining and identifying the effect of e-service agents on communication credibility and customers' satisfaction.

CONCLUSION
During the COVID-19 pandemic, many brands redesigned their services to heighten online experiences. The conclusion sheds light on how chatbot e-service agents' marketing motivates consumers to perceive communication credibility and satisfaction in apparel retailing. The results can serve as a guideline for apparel retailing brands to certify appropriate marketing efforts via e-services. Chatbots utilize ways to save time interacting with brands, together with trendiness, which can build positive perceptions of fashionable or up-to-date brands. Companies may now exceed customer expectations while also achieving corporate goals and creating value using new technology tools. E-service agents are continuously accessible personal helpers that help create vital client relations, allow extra effective use of customer time, and offer greater empathy regarding product performance. Additionally, as accuracy increases, users will be able to engage in intelligent social interactions with virtual agents. One of the most essential features that a chatbot should have is the ability to engage in problem-solving with customers.
Furthermore, communication credibility can enhance satisfaction. The study investigated the elements of trendiness and problem-solving in chatbots in regard to communication credibility, finding that both are important concerning this aspect. While interactive chatbots can provide customer satisfaction without credibility, the manager or marketer can design first impressions with friendly or informal information. Moreover, chatbots with entertainment components are unimportant to focus on during communication and may not result in customer satisfaction.
Despite the fact that the study's purpose was accomplished, some limitations were discovered. This study did not include the mediating and moderating variables perceived to be related to the evaluation of chatbot e-service marketing efforts, communication credibility, and customer satisfaction. Additionally, the paper did not spatially study online communication platforms such as Facebook, websites, or the Line application via e-service marketing efforts. Therefore, it is suggested that future studies should add value perception and consumer experience as mediating variables in obtaining a deeper understanding of consumers. ENT2 I was engrossed in the chatbot service agent's response. 0.781 ENT3 I was excited to talk with the chatbot service agent.

AUTHOR CONTRIBUTIONS
0.776 ENT1 It is amusing and pleasurable to converse with the chatbot service agent.
0.749 ENT4 I appreciate selecting things more when they are suggested by a chatbot service agent than when I select them myself.
0.741 SAT3 Overall, I am pleased with my experience with the chatbot. 0.834 SAT4 I would advise people to utilize the chatbot.
0.783 SAT2 I am delighted with the pre-purchase experience I had with the chatbot (e.g., product search, quality of product or service information, product comparison). Note: * means factor loading > 0.5 (removed from statistical analysis).