Major Research Project [MS44601]

"Adoption of AI chatbots in E-Commerce a Behavioral and

Technological perspective"
A project report submitted in partial fulfilment of
the requirement for the degree of

Masters of Business Administration

Session 2024-25
Semester IV
School of Management Studies
Motilal Nehru National Institute of Technology Allahabad

Submitted to:
Prof (Dr.) Rakesh Kumar

Submitted by:

Mohd Daniyal Ali

2023MB63

1
 Undertaking
I declare that the work presented in this report titled “Adoption of AIchatbots in E-Commerce a
Behavioral and Technological perspective” submitted to the School of Management Studies,
Motilal Nehru National Institute of Technology Allahabad, for the partial fulfilment of the degree
of Master of Business Administration (MBA) is my original work. It is not submitted anywhere
else for the award of any other degree.

Date: 07/05/2025 Mohd Daniyal Ali (2023MB63)

Place: Prayagraj

2
 Certificate
This is to certify that the project titled “Adoption of AI chatbots in E-Commerce a Behavioral and
Technological perspective”, is being submitted by Mohd Daniyal Ali (2023MB63) to School of
Management Studies, Motilal Nehru National Institute of Technology Allahabad, for partial
fulfilment of the requirements for the degree of Master of Business Administration (MBA), is a
record of original work carried out by him under my supervision and guidance.

Date: 07/05/2025 Prof (Dr.) Rakesh Kumar

Supervisor

3
 Acknowledgment

I extend my heartfelt gratitude to Prof (Dr.) Rakesh for providing me with the invaluable
opportunity to research on the topic of “Adoption of AI chatbots in E-Commerce a Behavioral and
Technological perspective”. His guidance, suggestions, and assistance were instrumental in the
completion of this project. Without his help it wouldn’t have been possible for me to complete the
project.

I’m also thankful to my family and friends for their unwavering support and encouragement
throughout this journey. Additionally, I am thankful to all those individuals who contributed their
expertise and assistance, enabling me to successfully accomplish this project.

Date: 07/05/2025 Mohd Daniyal Ali (2023MB63)

Place: Prayagraj

4
 Abstract

Purpose – Guided by the Technology Acceptance Model (TAM) and Behavioral Reasoning Theory
(BRT), this study aims to investigate consumers’ acceptance of AI-based chatbots in the context of e-
commerce, specifically focusing on apparel shopping. The research develops and tests a conceptual
model that explores how various "reasons for" and "reasons against" influence perceived ease of use
(PEU), perceived usefulness (PUF), attitude, and intention to use AI chatbots.
Design/methodology/approach – Primary data was collected from 156 valid responses via a
structured questionnaire. Structural Equation Modeling (SEM) was employed using JASP software to
analyze the hypothesized relationships among constructs such as optimism, innovativeness, relative
advantage, discomfort, insecurity, and complexity, as mediated by PEU and PUF.
Findings – The findings suggest that optimism, innovativeness, and relative advantage significantly
and positively influence PEU and PUF, establishing them as key enablers. Complexity negatively
impacts PEU, making it a significant reason against adoption. Discomfort and insecurity show no
significant effect. PEU positively affects PUF, and both PEU and PUF significantly influence
consumer attitudes, which in turn affect the intention to adopt AI chatbots in apparel shopping.
Originality/value – By integrating TAM and BRT, this study provides a comprehensive
understanding of the enabling and inhibiting factors that shape consumer attitudes and intentions
towards adopting AI chatbots in online apparel retail. It extends existing literature by identifying
psychological and technological determinants critical to e-commerce adoption behavior.

5
 Table of Content
Chapter 1: INTRODUCTION ...................................................................................................................... 8
1.1 Introduction.............................................................................................................................. 8
1.2 Relevance of the Study............................................................................................................ 11
1.3 Objective of Study .................................................................................................................. 12
1.4 Scope of the Study .................................................................................................................. 12
Chapter 2: LITERATURE REVIEW ......................................................................................................... 13
Literature Review ......................................................................................................................... 13
2.1 Review ..................................................................................................................................... 13
Chapter 3: RESEARCH METHODOLOGY ............................................................................................. 20
Research Methodology ................................................................................................................. 20
3.1 Introduction ............................................................................................................................ 20
3.2 Research Questions ................................................................................................................ 20
3.3 Population and Sample.......................................................................................................... 20
3.4 Sampling ................................................................................................................................. 20
3.5 Questionnaire Design ............................................................................................................. 21
3.6 Data Collection ....................................................................................................................... 21
Chapter 4: DATA ANALYSIS & FINDINGS ............................................................................................. 22
Data Analysis & Findings ............................................................................................................ 22
Chapter 5: CONCLUSION......................................................................................................................... 43
5.1 Summary of key findings ....................................................................................................... 43
5.2 Limitations .............................................................................................................................. 44
5.3 Direction for future research ................................................................................................ 46
Chapter 6: RECOMMENDATION............................................................................................................ 49
6.1 Implications............................................................................................................................. 49
6.1 Suggestion ............................................................................................................................... 50
REFERENCES .......................................................................................................................................... 53

6
 List of Tables and Figures

Table 1: Demographic Analysis of Respondents. ....................................................................................... 22

Table 2: Outer Loadings.............................................................................................................................. 24
Table 3: Cronbach’s Alpha Table................................................................................................................ 26
Table 4: Average Variance Extracted (AVE) Table..................................................................................... 27
Table 5: Fornell-Larcker Criterion s ............................................................................................................. 29
Table 6: Heterotrait-Monotrait Ratio (HTMT) ............................................................................................ 30
Table 7: Path Coefficients and Significance Table. .................................................................................... 31
Table 8: Explained Variance (R² Values) Table .......................................................................................... 33
Table 9: Model Fit table.............................................................................................................................. 34
Table 10: Hypothesis Testing Table............................................................................................................ 36
Figure 1: Final PLS-SEM Path Model of AI Chatbot Adoption Framework.................................................. 39

7
 CHAPTER 1: INTRODUCTION
1.1 Introduction

In today’s digital-first marketplace, the integration of emerging technologies has become central to
how consumers and retailers interact throughout the buying journey. The proliferation of artificial
intelligence (AI), particularly in the form of AI-powered chatbots, is a key development reshaping
customer experiences across numerous industries. These chatbots—digital agents that use natural
language processing (NLP) and machine learning (ML) to simulate human-like conversations—are
increasingly employed to deliver fast, efficient, and personalized services in real-time. Industries such
as tourism, banking, and healthcare have widely adopted chatbot technologies, but their impact is
especially evident in the retail sector, where customer interaction plays a pivotal role in driving
engagement, satisfaction, and brand loyalty (De Cicco et al., 2020; Rese et al., 2020).

In the retail space, AI chatbots provide a wide range of services, including responding to customer
inquiries, assisting with product discovery, offering personalized suggestions, and streamlining
transaction processes. Retailers are turning to chatbot solutions not only to improve operational
efficiency but also to enhance customer relationships through seamless and scalable digital
interactions. As online shopping becomes more data-driven, consumers increasingly expect real-time
assistance that is contextually relevant, emotionally intelligent, and aligned with their personal
preferences. In this context, chatbots serve as virtual assistants capable of learning from past
interactions to offer more refined and customized shopping experiences, making them an essential
asset for brands navigating today’s competitive digital retail landscape (Cheng et al., 2023; Krishnan
et al., 2022).

The fashion and apparel retail industry has shown a particular affinity for AI chatbot integration due
to its strong reliance on customer engagement and personalization. In this sector, chatbots act as virtual
fashion stylists, helping users find apparel based on their stated preferences regarding color, style, and
occasion. Notable fashion brands such as Tommy Hilfiger, H&M, and Levi’s have adopted chatbot
services via platforms like Facebook Messenger, Kik, and other messaging apps, using AI to improve
the discovery and selection of fashion products (Landim et al., 2022; Davenport et al., 2020). These
bots can guide consumers through a curated journey, suggest outfits, and direct them to purchase
pages—enhancing not just convenience but also the emotional connection between the consumer and
8
the brand.

Due to their ability to deliver personalized, efficient, and around-the-clock service, AI chatbots are
increasingly viewed as essential tools in fashion retail. As reported by Walk-Morris (2020), retail
spending through chatbots reached $2.8 billion in 2019 and is expected to rise sharply to $142 billion
by 2024. This rapid growth reflects a broader shift in consumer behavior and expectations, particularly
among digitally savvy shoppers who seek instant responses, consistent service quality, and tailored
experiences. As the fashion industry moves further into digital and omnichannel strategies, AI chatbots
are positioned to become critical to customer engagement, acquisition, and retention.
Despite the clear commercial and technological benefits, scholarly exploration into AI chatbot
adoption within the apparel industry remains limited. Existing studies have mostly focused on
functionality and performance, such as how chatbots improve customer satisfaction and usability
(Chung et al., 2020; Rese et al., 2020). For instance, Chung et al. (2020) analyzed how chatbot
communication quality influences consumer satisfaction in the luxury fashion segment, revealing that
well-designed chatbot interactions can significantly enhance the perceived value of the brand.
Meanwhile, Rese et al. (2020) identified conversation authenticity, perceived enjoyment, and
usefulness as major determinants of chatbot acceptance, while privacy concerns and immaturity served
as barriers.
However, these investigations often fail to address the deeper psychological, emotional, and
behavioral mechanisms that influence consumer adoption and resistance. Given the multifaceted
nature of consumer behavior, there is a growing need to move beyond purely functional evaluations
and explore the underlying factors that drive user intentions and attitudes toward AI chatbots in fashion
retail. Understanding these dynamics is critical not only for academic inquiry but also for practical
implementation, as brands strive to align chatbot design with evolving consumer expectations.
To address this research gap, the present study adopts a dual-theory approach that integrates the
Technology Acceptance Model (TAM) (Davis, 1989) with the Behavioral Reasoning Theory (BRT)
(Westaby, 2005). TAM has been widely used to explain the adoption of new technologies,
emphasizing two primary factors—perceived ease of use (PEU) and perceived usefulness (PUF)—as
key determinants of user acceptance (Celik & Yilmaz, 2011; Hsu & Lu, 2004). These constructs have
been validated across a variety of digital contexts, making TAM a robust foundation for studying
chatbot adoption in fashion e-commerce.
However, TAM alone may not fully capture the complexity of consumer reasoning, particularly the
9
coexistence of both acceptance and resistance factors in the adoption process. To enrich this
perspective, this study incorporates BRT, which highlights the role of motivational beliefs—“reasons
for” and “reasons against”—in shaping behavior (Chatzidakis & Lee, 2013; Claudy et al., 2015).
Importantly, BRT challenges the notion that resistance is simply the inverse of acceptance, proposing
instead that these are distinct, independent forces influencing decision-making. This integrated model
thus enables a more holistic exploration of the psychological variables at play.
In this study, “reasons for” adoption are operationalized through constructs such as optimism,
innovativeness, and relative advantage, which represent consumers’ positive evaluations of chatbot
technologies. On the other hand, “reasons against” adoption include discomfort, insecurity, and
complexity—factors that contribute to apprehension or resistance toward chatbot usage. These
variables are positioned as antecedents to PEU and PUF in the TAM framework, providing a
comprehensive model that captures both enablers and inhibitors of adoption. This allows for a deeper
understanding of how consumers form attitudes toward AI chatbot usage and how these attitudes
ultimately influence behavioral intentions.

The integration of BRT and TAM also supports the inclusion of external factors as predictors of PEU
and PUF, aligning with Wu and Wang’s (2005) argument that TAM becomes more robust when
external variables are considered. By examining the relationship between external motivational drivers
and TAM constructs, this research offers nuanced insights into the consumer decision-making process.
Specifically, the model seeks to understand how positive and negative beliefs influence perceptions of
chatbot usability and utility, and how these perceptions shape attitudes and intention to adopt.
Accordingly, this study aims to examine consumer behavior related to AI chatbot adoption in the
apparel retail context by empirically testing a conceptual model informed by TAM and BRT. The
specific objectives of the study are fourfold: (1) to investigate the influence of external factors (reasons
for and against) on PEU and PUF; (2) to examine the relationship between PEU and PUF; (3) to assess
the impact of PEU and PUF on attitudes toward AI chatbot usage; and (4) to explore whether these
attitudes predict consumers’ intention to use AI chatbots for apparel shopping. Through this approach,
the study aims to fill existing gaps in the literature and provide actionable insights for fashion retailers.
In conclusion, AI chatbot technology represents a significant advancement in digital retail, with the
potential to transform the way consumers interact with apparel brands. While current trends indicate
growing adoption and user interest, a deeper understanding of the psychological mechanisms driving
consumer acceptance and resistance remains essential. By adopting a dual-theoretical lens that
10
incorporates both TAM and BRT, this study provides a comprehensive framework for examining the
complex relationship between chatbot attributes, user perceptions, and adoption behavior. The
findings are expected to inform both academic discourse and strategic practice, enabling retailers to
design more effective, customer-centric chatbot solutions in an increasingly AI-driven retail
environment.

1.2 Relevance of the study

The relevance of this study lies in its ability to bridge theoretical models with practical insights to
understand consumer behavior regarding the adoption of AI chatbots in online shopping. As the e-
commerce landscape continues to evolve, AI-powered chatbots have emerged as critical tools for
enhancing customer engagement, providing real-time assistance, and personalizing the shopping
experience. This paper, along with other recent research and industry developments, highlights how
consumer perceptions of ease of use, usefulness, trust, and convenience significantly impact chatbot
adoption.
By integrating the Technology Acceptance Model (TAM) and Behavioral Reasoning Theory (BRT),
the study by ThuMyin and Watchravesringkan offers a comprehensive framework that goes beyond
surface-level acceptance to explore both motivators and barriers to adoption. These insights are
particularly valuable in today's digital marketplace, where chatbots serve not just as service agents but
as virtual shopping companions influencing purchase decisions.
Furthermore, external sources such as industry reports from Gartner and Juniper Research forecast a
substantial increase in chatbot-driven sales and customer interactions, reinforcing the practical
importance of this topic. For instance, the global chatbot market is expected to reach over $10 billion
by 2026, with e-commerce being one of the leading sectors driving this growth. Consumer
expectations around 24/7 service, instant support, and seamless user interfaces are shaping the way
businesses design and deploy chatbot systems.This research is therefore crucial in laying the
groundwork for a deeper understanding of consumer readiness, technological trust, and strategic
implementation of AI chatbots in the e-commerce sector. It not only informs academic inquiry but
also guides businesses in crafting effective, user-centric digital experiences.

11
1.3 Objectives of the study
• To identify and examine the factors (optimism, innovativeness, and relative advantage) that
positively influence consumers’ perceived ease of use and perceived usefulness of AI chatbots
in online shopping.
• To analyze the impact of barriers (discomfort, insecurity, and complexity) on consumers’
perceived ease of use and perceived usefulness of AI chatbots in e-commerce platforms.
• To evaluate the relationship between perceived ease of use and perceived usefulness in the
context of AI chatbot adoption in online shopping.
• To investigate how perceived ease of use and perceived usefulness influence consumer
attitudes, and how these attitudes, in turn, affect the intention to use AI chatbots in online
shopping.
• To provide strategic insights for e-commerce businesses to enhance chatbot adoption by
improving user experience and addressing consumer concerns.

1.4 Scope of the study

The scope of this study encompasses the investigation of consumer adoption and usage behavior of
AI-powered chatbots within the online shopping ecosystem. It focuses on understanding how factors
such as perceived ease of use, usefulness, trust, responsiveness, personalization, and complexity
influence consumer attitudes and intentions toward using AI chatbots for e-commerce purposes.
This study is limited to the context of online shopping platforms that utilize AI chatbots for tasks such
as customer support, product recommendations, order tracking, and virtual assistance. The target
population includes online shoppers who have interacted with chatbots during their shopping journey.
Geographically, the study may be focused on a specific region or country (e.g., India or global online
users, based on your preference), and is conducted within a specified time frame. The findings will be
based on primary data collected through surveys or interviews, supported by secondary data from
academic literature and industry reports.This research does not include other forms of AI in e-
commerce such as visual search engines, recommendation algorithms (excluding chatbot-assisted
ones), or augmented reality tools. It also does not cover chatbot applications in non-retail sectors like
healthcare or education.By narrowing its focus to chatbot technology in e-commerce, the study aims
to provide actionable insights for online retailers, developers, and digital marketers to improve chatbot
design, usability, and adoption strategies.
12
 CHAPTER 2: LITERATURE REVIEW
2.1 Introduction
Artificial Intelligence (AI) has emerged as a transformative technology that continues to reshape
multiple industries, including e-commerce. Among its many applications, AI-powered chatbots have
become a central tool in revolutionizing customer service and user experience. A chatbot, defined as
an AI program capable of simulating human conversation through Natural Language Processing
(NLP) and sentiment analysis, is a fundamental part of the Human-Computer Interaction (HCI)
paradigm (Bansal & Khan, 2018). In e-commerce, AI chatbots act as digital assistants that facilitate
real-time interactions, assist users in product selection, handle queries, and provide personalized
shopping experiences. These capabilities not only enhance operational efficiency but also contribute
significantly to customer satisfaction and engagement (Paliwal et al., 2020).
With the rapid expansion of digital marketplaces, the integration of AI-driven chatbots has become
increasingly critical. Traditional customer service models relying on human agents often faced
limitations in scalability, response time, and personalization (Rojas, 2024). In contrast, AI chatbots
offer round-the-clock support, simultaneously manage thousands of interactions, and deliver
consistent, customized responses. Research indicates that AI chatbots greatly enhance product
discovery by narrowing down choices based on customer preferences, browsing history, and
purchasing behavior (Lee, 2020; Krishnan et al., 2022). This advanced recommendation ability plays
a pivotal role in improving customer decision-making and increasing cross-selling and up-selling
opportunities.
Moreover, chatbots have proven instrumental in improving the overall user experience by offering
seamless, intuitive, and interactive shopping journeys. The capability of AI to process vast amounts
of customer data in real-time allows chatbots to adapt their responses dynamically, making interactions
more engaging and personalized (Fida, 2019; Cheng et al., 2023). Notably, AI chatbots also address
critical business challenges by reducing operational costs, minimizing the burden on human support
teams, and maintaining service quality during peak periods such as holiday sales seasons. These
operational benefits translate into enhanced customer retention rates and stronger consumer trust
(Ebaietaka, 2024; Savastano et al., 2024).
However, despite the visible advantages, the adoption of AI chatbots by consumers is not guaranteed.
Multiple factors influence users' willingness to accept and regularly use chatbot services, including
perceived ease of use, perceived usefulness, trust, privacy concerns, and prior technology experience.

13
 To understand these adoption behaviors comprehensively, the Technology Acceptance Model (TAM),
introduced by Davis (1989), provides an effective theoretical foundation. TAM posits that users’
perceptions regarding the ease of use and usefulness of technology directly impact their attitudes and
intentions toward using it. Studies such as Kasilingam (2020) and Chung et al. (2020) validate that
these constructs are highly relevant in predicting chatbot adoption within e-commerce settings.
Therefore, this research aims to investigate the adoption of AI chatbots in e-commerce platforms using
the Technology Acceptance Model (TAM) as a guiding framework. It will focus on analyzing how
perceived ease of use and perceived usefulness shape customer attitudes and intentions, while also
considering additional dimensions such as trust, satisfaction, and engagement. The findings are
expected to provide valuable insights for e-commerce businesses seeking to maximize the
effectiveness of their chatbot implementations and enhance customer experience.

2.2 Theoretical Background

The Technology Acceptance Model (TAM) was developed by Davis (1989) as an adaptation of the
Theory of Reasoned Action (TRA) by Fishbein and Ajzen (1975). It remains one of the most influential
models for predicting user acceptance and usage behavior toward new technologies. TAM proposes that
two key beliefs — Perceived Ease of Use (PEU) and Perceived Usefulness (PUF) — are primary
determinants of users’ attitudes toward using a system, which subsequently influence their behavioral
intention and actual usage behavior.
Perceived Ease of Use (PEU) refers to the degree to which an individual believes that using a
particular system would be free from effort. In the context of AI chatbots, it reflects how intuitively
customers can interact with the chatbot without technical difficulties.
Perceived Usefulness (PUF) is the degree to which an individual believes that using a system will
enhance their performance. In e-commerce, this would imply the effectiveness of chatbots in helping
customers make faster, easier, and better purchasing decisions.
According to Davis (1989), if users perceive a technology as both easy to use and useful, they are
more likely to develop a positive attitude towards it and intend to use it regularly. Over the years, TAM
has been widely validated across numerous studies and applied in various domains, such as online
banking, mobile commerce, healthcare technologies, and e-learning platforms (Venkatesh & Davis,
2000; King & He, 2006).
In the specific context of AI chatbots in e-commerce, TAM is highly applicable. Chatbots must be
perceived as user-friendly and genuinely helpful for customers to trust and adopt them. Research by
14
Kasilingam (2020) demonstrated that both PEU and PUF significantly influenced customer attitudes
toward mobile shopping chatbots. Similarly, Chung et al. (2020) found that the ease and usefulness of
chatbot interactions increased luxury brand customers’ satisfaction and loyalty.
Moreover, recent studies such as ThuMyin and Watchravesringkan (2024) extended the basic TAM
model to integrate external factors like optimism, complexity, and trust, highlighting how user perceptions
can be shaped by both system characteristics and psychological factors. These adaptations confirm TAM’s
flexibility and relevance in studying AI-driven innovations like chatbots.
Thus, by utilizing the TAM framework, this research aims to investigate how perceived ease of use
and perceived usefulness influence customer attitudes and behavioral intentions toward adopting AI
chatbots in e-commerce settings.

2.3 Past Studies

2.3.1 Influence of Perceived Ease of Use and Usefulness on Chatbot Adoption
Kasilingam (2020) explored consumer adoption of smartphone-based AI chatbots for online shopping
using the Technology Acceptance Model (TAM) integrated with elements of the Diffusion of Innovations
(DOI) theory. The study aimed to examine how perceived ease of use (PEOU) and perceived usefulness
(PU), among other variables, shape consumer attitudes and intentions toward chatbot use. A Facebook-
based shopping chatbot was developed for this research, and data were collected from 350 participants
who interacted with the bot before completing a structured questionnaire. The findings, analyzed using
Partial Least Squares Structural Equation Modeling (PLS-SEM), revealed that both PEOU and PU had a
significant positive influence on user attitudes, reinforcing the foundational TAM assumption that ease
and usefulness shape technology acceptance. Additionally, perceived enjoyment, price consciousness,
perceived risk, and personal innovativeness were also found to impact attitudes. However, the intention
to use the chatbot was influenced primarily by trust, personal innovativeness, and the users’ attitude itself,
suggesting that while usability and usefulness contribute to forming favorable attitudes, they do not
directly determine behavioral intention without trust. The study further identified attitude as a full
mediator between PEOU/PU and intention to use, underlining its critical role in the adoption process.
These insights stress the importance for e-commerce firms to not only design user-friendly and efficient
chatbot systems but also to cultivate trust and innovation-friendly environments to drive actual usage. The
research provides a comprehensive perspective on how TAM constructs operate in conjunction with user
perceptions and traits in the context of AI chatbot adoption for mobile shopping.

15
 2.3.2 Service Quality of Chatbots in Luxury Retail
Chung et al. (2020) investigated the role of AI chatbots in delivering e-service quality and customer
satisfaction in the context of luxury fashion retail. Recognizing the growing reliance on digital tools in
high-end retail, the study examined whether chatbots could replicate the personalized, engaging service
typically associated with in-store luxury experiences. The researchers developed a five-dimensional model
to evaluate chatbot service based on interaction, entertainment, trendiness, customization, and
problem-solving. Using data collected from 157 young luxury consumers in South Korea who had prior
experience with chatbot-assisted online shopping, the study utilized structural equation modeling (SEM)
to test the hypothesized relationships. The results indicated that these five marketing effort dimensions
significantly impacted customers’ perceptions of communication quality—specifically in terms of
accuracy, credibility, and communication competence. Communication quality, in turn, was shown to
strongly influence customer satisfaction. The study found that although users perceived chatbot
communication as credible and accurate, they were less convinced of its competence compared to human
service agents. Nonetheless, customers still reported overall satisfaction with chatbot interactions when
the service was timely, personalized, and trustworthy. The findings suggest that even in the emotionally
driven luxury sector, AI chatbots can effectively meet customer expectations and enhance satisfaction—
provided they are designed to mirror the values of exclusivity, attentiveness, and personalization. The
research supports the notion that AI chatbots, when properly implemented, can extend luxury brand
values into digital environments, and provides empirical support for using chatbot service quality
dimensions as predictors of customer satisfaction in e-commerce..

2.3.3 Role of Trust and Privacy in Chatbot Adoption

Luo et al. (2019) conducted a large-scale field experiment to understand how disclosing a chatbot’s
machine identity affects customer trust and purchase behavior in conversational commerce. Involving
over 6,200 customers of a major financial services company, the study revealed that customers were
significantly less likely to complete a purchase when they were informed in advance that they were
interacting with an AI chatbot. Specifically, disclosure before the conversation began reduced purchase
rates by more than 79.7%, even though the chatbot was as competent and responsive as human agents.
This reaction was attributed to a psychological bias against machines, where users perceived disclosed
chatbots as less empathetic and less knowledgeable, despite no objective difference in performance.
Voice-mining analysis and post-call surveys confirmed that perceived trust and emotional connection
were undermined by the disclosure, indicating that trust plays a central role in chatbot adoption.
16
Importantly, the study found that this negative effect could be mitigated through strategic timing of
disclosure (e.g., revealing the chatbot identity after the interaction) and by targeting customers with prior
experience using AI tools. These findings emphasize the delicate balance between transparency and user
perception in AI chatbot deployment. While customers appreciate helpful and responsive service, early
disclosure of chatbot identity can trigger skepticism and lower trust, ultimately affecting usage and
satisfaction. The study provides critical insights into how trust and privacy perception shape user behavior
and offers actionable strategies for businesses deploying conversational AI technologies..

2.3.4 Chatbots Enhancing Product Recommendations in E-Commerce

Rahevar and Darji (2024) conducted a study focused on the integration of AI-driven chatbots into
recommendation systems within e-commerce platforms, aiming to assist customers in selecting products
more accurately. The research highlighted the critical role chatbots play in enhancing customer decision-
making by using real-time data and machine learning to recommend products based on user preferences,
browsing behavior, and purchase history. Their findings showed that AI-powered chatbots significantly
improve the accuracy of product recommendations, with a regression weight of 0.693 and beta
coefficient of 0.480, indicating a 48% increase in selection accuracy per unit of chatbot integration.
Additionally, the study revealed that chatbot implementation enhances user satisfaction (β = 0.840),
engagement, and purchase decisions, especially when the system provides instant, personalized
responses. By analyzing a sample of 268 respondents using survey data and regression analysis, the
researchers found a strong positive correlation between AI chatbots and improved customer experience
metrics such as trust and retention. The chatbot's ability to interpret customer intent, filter product choices,
and adapt in real-time created a more engaging and efficient shopping journey, ultimately increasing cross-
selling and conversion rates. The research concluded that AI chatbots are not only operational tools but
strategic assets that can personalize customer journeys, build loyalty, and reduce the need for human
customer service intervention. These findings support the growing view that recommendation-based AI
chatbots are vital in competitive e-commerce ecosystems where users seek speed, personalization, and
trust.

2.3.5 The Impact of Generative AI on Retail Chatbot Adoption

Arce-Urriza et al. (2025) conducted a timely and comprehensive study exploring how Generative AI
(GenAI) influences consumer adoption of retail chatbots. Using the Service Robot Acceptance Model
(SRAM), the study examined key factors such as perceived usefulness, ease of use, human-likeness, trust,
17
and privacy risk. Two surveys were administered — one in 2022 (pre-GenAI) and another in 2024 (post-
GenAI) — involving over 800 participants who were exposed to chatbot conversations across three
customer journey stages: pre-purchase, purchase, and post-purchase. Results showed that perceived
usefulness, human-likeness, and familiarity with chatbots significantly improved after GenAI
integration, enhancing adoption intentions. However, trust remained unchanged, and privacy concerns
increased, indicating that while GenAI makes chatbots more engaging and context-aware, it also raises
fears related to data misuse. Importantly, familiarity played a direct and indirect role in promoting
chatbot adoption by boosting perceived usefulness and reducing perceived risk. Interestingly, the
influence of adoption factors remained stable across all three customer journey stages, suggesting that
GenAI’s impact is more psychological than situational. The study also found that subjective social norms
and user familiarity had a strong influence on adoption, while a preference for human interaction slightly
reduced chatbot usage intentions. This research expands the theoretical understanding of chatbot adoption
by validating SRAM in a GenAI context and highlighting the evolving role of technological familiarity
and privacy perceptions in shaping consumer attitudes toward AI-powered retail assistants.

2.3.6 Consumer Trust and Response to Chatbots Under Task Complexity and Identity Disclosure
Cheng et al. (2022) conducted a study to examine consumer responses to text-based chatbots in e-
commerce, emphasizing how empathy, friendliness, task complexity, and identity disclosure influence
trust and usage behavior. Drawing on the Stimulus–Organism–Response (SOR) model, the researchers
surveyed 299 users who had prior experience interacting with chatbots during online purchases. The
findings revealed that empathy and friendliness are key drivers of trust toward chatbots, with empathy
having a stronger effect. Interestingly, the study identified that task complexity negatively moderated the
effect of friendliness on trust — meaning when the customer’s task was complex, friendly chatbot
behavior alone was insufficient to build trust. In contrast, empathy remained impactful regardless of task
difficulty. Furthermore, disclosing the chatbot's machine identity reduced the trust-building effect of
empathy but slightly enhanced the impact of friendliness, suggesting that early disclosure may create
mixed user perceptions. Most notably, trust in chatbots was found to significantly increase user reliance
and reduce resistance to using chatbots in future transactions. These insights highlight the need for
businesses to design chatbot interactions with context in mind, ensuring that emotional cues are paired
with task-relevant capabilities, especially for complex service scenarios. The research adds to the literature
by providing nuanced evidence that the same chatbot features can have different effects depending on
customer expectations and usage situations.
18
2.4 Research Gaps
Although numerous studies have examined the adoption of AI chatbots across different sectors, several
gaps remain in the current literature. Firstly, much of the existing research primarily focuses on the
technical and functional aspects of chatbot adoption—such as usability, speed, and accuracy—while
emotional and behavioral factors like trust, empathy, and user satisfaction receive comparatively less
attention. Secondly, many prior studies are centered around specific industries such as banking,
telecommunications, or luxury fashion, with limited exploration of chatbot use in the broader e-
commerce environment, especially across diverse geographic and cultural contexts. This restricts the
generalizability of existing findings to global online retail platforms. Thirdly, while the Technology
Acceptance Model (TAM) has been widely used to study technology adoption, very few studies have
applied TAM in combination with R-based data analysis tools. Most rely on traditional platforms like
SPSS or AMOS, limiting methodological diversity and robustness. By addressing these gaps, the present
study contributes to the literature by using TAM to explore emotional and behavioral influences on chatbot
adoption in the e-commerce domain, supported by primary data analysis using the R programming
language.

2.5 Conclusion
The literature reviewed highlights the growing significance of AI chatbots in enhancing customer
engagement and service efficiency across various sectors, including e-commerce. Studies consistently
emphasize the central role of the Technology Acceptance Model (TAM) in understanding user adoption,
with perceived ease of use and perceived usefulness emerging as key predictors of customer attitudes and
behavioral intentions. Additional factors such as trust, emotional interaction, service quality, and identity
disclosure also play a significant role in shaping user perceptions. However, gaps in existing research
reveal a need for greater focus on emotional and behavioral dimensions of adoption, application across
varied e-commerce contexts, and the use of diverse methodological tools such as R for data analysis.
Building on this foundation, the present study aims to contribute by investigating consumer adoption of
AI chatbots in e-commerce through the TAM framework while addressing both conceptual and
methodological limitations identified in the literature.

19
 CHAPTER 3: RESEARCH METHODOLOGY

3.1 Introduction

This chapter outlines the research methodology used to examine the adoption of AI chatbots in e-
commerce. It describes the research design, population and sample, sampling method, sample size,
questionnaire design, and data collection procedures. The methodology ensures a structured approach to
gathering and analysing data for answering the research questions.

3.2 Research Questions

The study seeks to answer the following key research questions:
1. What are the factors influencing the adoption of AI chatbots in e-commerce?
2. How do perceived usefulness, perceived ease of use, and trust affect user attitudes towards AI
chatbot adoption?
3. What is the relationship between user experience and intention to use AI chatbots in future online
purchases?

3.3 Population and Sample

The population for this study consists of online consumers in India who have experience interacting with
AI-based chatbots in e-commerce platforms such as Amazon, Flipkart, Myntra, and others.

3.4 Sampling

3.4.1 Target Population

The target population includes individuals who shop online and have interacted with AI-powered chatbots
in any form—product queries, order tracking, customer support, or recommendations—on e-commerce
websites.

3.4.2 Sampling Method

A non-probability purposive sampling technique was used to select participants who have prior
experience using AI chatbots while shopping online. This technique ensured that only relevant
respondents who met the study's criteria were included in the final analysis.

20
3.4.3 Sample Size

A total of 182 responses were received through the Google Form, and all were considered valid for
analysis. These responses were used to examine consumer perceptions and adoption behavior regarding
AI chatbots in the e-commerce domain

3.5 Questionnaire Design

The questionnaire was designed based on established constructs from prior research on technology
adoption models, particularly the Technology Acceptance Model (TAM) and UTAUT framework. It
consisted of structured close-ended questions rated on a 5-point Likert scale ranging from 1 (Strongly
Disagree) to 5 (Strongly Agree). The questionnaire captured the following variables:

• Perceived Usefulness

• Perceived Ease of Use

• Trust

• User Attitude

• Intention to Use

• Actual Experience

The questionnaire was reviewed and refined to ensure clarity, relevance, and reliability of the items.

3.6 Data Collection

Primary data was collected using a Google Form questionnaire distributed online via social media
platforms and email. Respondents were informed of the purpose of the research and were assured of
anonymity and confidentiality. The data collection phase spanned approximately two weeks, after which
responses were screened and cleaned.

Data analysis was conducted using SmartPLS software, which supports Structural Equation Modeling
(SEM) for assessing both measurement and structural models. This allowed for detailed insights into the
relationships among variables and the validation of hypotheses based on path coefficients and model fit
indicators.

21
 CHAPTER 4: DATA ANALYSIS AND FINDINGS

Demographic Profile of Respondents

Before analyzing the structural model, it is essential to understand the demographic characteristics of
the respondents who participated in the study. A total of 182 valid responses were collected. The
demographic distribution includes key variables such as age, gender, education level, and frequency
of online shopping, which are summarized in the table below:
Table 1: Demographic Analysis of Respondents.

Demographic characteristics N%
(a). Age
18-25 40.34%
26-35 57.14%
36-45 1.68%
46-55 0.84%
(b).Gender
Male 66.39%
Female 33.61%
(c). Highest Education Qualification
Diploma/Certificate Course 1.68%
Post Graduate Degree and Above 66.39%
School Education 8.40%
Under Graduate Degree 23.53%
(d). Annual Household Income
Above Rs. 20 Lakhs 11.76%
Rs. 12 Lakhs - Rs. 16 Lakhs 10.92%
Rs. 16 Lakhs - Rs. 20 Lakhs 10.08%
Rs. 4 Lakhs - Rs. 8 Lakhs 15.13%
Rs. 8 Lakhs - Rs. 12 Lakhs 31.93%
Upto Rs. 4 Lakhs 20.17%
(e). Marital Status
Married 34.45%
Unmarried 65.55%
(f). Frequency of Online Shopping
Daily 10.92%
Less often 10.08%
Monthly 42.86%
Weekly 36.13%
(g). Experience with AI Chatbots
Yes 87.39%
No 12.61%
22
 4.1 Measurement Model Evaluation (Confirmatory Factor Analysis)
The measurement model was evaluated to assess the reliability and validity of the constructs.
4.1.1. Indicator Reliability (Outer Loadings)
Indicator reliability refers to the extent to which each individual observed variable (questionnaire item)
contributes to measuring its associated latent construct. In Partial Least Squares Structural Equation
Modeling (PLS-SEM), this is assessed through outer loadings, also known as factor loadings. These
loadings reflect the degree of correlation between each indicator and its underlying construct.
A general rule of thumb in PLS-SEM is that outer loadings should ideally be above 0.70, indicating that
over 50% of the indicator’s variance is shared with the construct. Loadings between 0.40 and 0.70 are
considered acceptable if the overall composite reliability and average variance extracted (AVE) of the
construct are within threshold values. Loadings below 0.40, however, are typically considered too weak
and may be candidates for removal unless retained for strong theoretical justification.
In the current study, the majority of indicators exhibited strong and satisfactory outer loadings,
demonstrating a solid relationship between the survey items and their corresponding latent variables. For
example:
• Attitude was measured by ATU1 and ATU2, with loadings of 0.928 and 0.926 respectively,
reflecting high reliability.
• PEU had indicators PEOU1 and PEOU2 with loadings of 0.961 and 0.964, indicating excellent
measurement quality.
• PUF also showed strong reliability, with both indicators (PU1 and PU2) loading at 0.966.
Constructs such as Innovation, Optimism, Relative Advantage, and Complexity also demonstrated
acceptable to high outer loadings across their respective indicators. This supports the assumption that the
constructs were measured accurately and consistently.
However, two indicators presented significant issues:
• INS1 (Insecurity): outer loading of -0.086
• DIS1_R (Discomfort): outer loading of 0.180
These values are far below the acceptable threshold, indicating that these items did not effectively
contribute to measuring their intended constructs. While these items were retained to align with the
original theoretical model and maintain construct structure, their influence on the measurement model is
limited. Their weak performance may suggest poor wording, ambiguity, or that they are no longer relevant
concerns among the sample population, particularly in light of increasing consumer familiarity with AI
technologies.
23
In conclusion, the indicator reliability for most constructs in the model was strong, with outer loadings
providing clear support for the validity of the measurement model. The presence of a few weak indicators
does not undermine the overall quality of the model but rather offers insights into evolving user
perceptions, highlighting areas for future item refinement.
Table 2: Outer Loadings
Outer loadings

ATU1 <- Attitude 0.928

ATU2 <- Attitude 0.926

BI1 <- Intenion 1.000

CMP1_R <- Complexity 0.968

CMP2_R <- Complexity 0.966

DIS1_R <- Discomfort 0.180

DIS2_R <- Discomfort 0.995

INN1 <- Innovation 0.881

INN2 <- Innovation 0.864

INS1 <- Insecurity -0.086

INS2 <- Insecurity 0.967

OPT1 <- Optimism 0.728

OPT2 <- Optimism 0.886

PEOU1 <- PEU 0.961

PEOU2 <- PEU 0.964

PU1 <- PUF 0.966

PU2 <- PUF 0.966

RA1 <- Relative Advantage 0.823

RA2 <- Relative Advantage 0.818

24
4.1.2. Internal Consistency Reliability
Three reliability measures were assessed: Cronbach’s Alpha, Composite Reliability (CR), and rho_A.
Internal consistency reliability evaluates how consistently multiple indicators measure the same latent
construct. In PLS-SEM, this is a crucial aspect of assessing the measurement model and ensures that the
items assigned to a construct are interrelated and working cohesively to capture the intended concept.
To assess internal consistency, three key metrics are examined:
• Cronbach’s Alpha
• Composite Reliability (CR or rho_c)
• rho_A (Dillon-Goldstein’s rho)
Cronbach’s Alpha is the most traditional measure of reliability, and values above 0.70 are generally
considered acceptable. However, because Cronbach’s Alpha assumes equal indicator loadings and may
underestimate reliability in PLS-SEM, Composite Reliability (CR) is often preferred. CR takes into
account the actual outer loadings of the indicators and provides a more accurate estimate of construct
reliability. Similarly, rho_A is another reliability coefficient that offers a middle ground between Alpha
and CR.
In this study, the internal consistency reliability for most constructs was well within acceptable and even
excellent ranges. For instance:
• Attitude had a Cronbach’s Alpha of 0.837 and a CR of 0.924, demonstrating a very high degree
of consistency between its two indicators (ATU1 and ATU2).
• PEU (Perceived Ease of Use) and PUF (Perceived Usefulness) had very high CR values of 0.962
and 0.965 respectively, indicating that these constructs were measured with great reliability.
• Complexity, composed of CMP1_R and CMP2_R, had one of the highest reliability scores in the
model, with a CR of 0.967 and an Alpha of 0.931.
Moderate reliability was observed in constructs like:
• Innovation (CR = 0.864; Alpha = 0.686)
• Optimism (CR = 0.792; Alpha = 0.493)
• Relative Advantage (CR = 0.805; Alpha = 0.516)
While these constructs met the minimum thresholds, their relatively lower Cronbach’s Alpha values may
be due to having only two indicators. It is widely acknowledged in research methodology that Cronbach’s
Alpha tends to be lower when constructs are measured by fewer items, even if those items load well.

25
On the other hand, two constructs raised concerns:
• Discomfort, with a CR of 0.585 and Alpha of 0.437
• Insecurity, with a CR of 0.423 and Alpha of 0.292
These values fall below the commonly accepted cutoff of 0.70, indicating low internal consistency. This
could be attributed to problematic indicators (DIS1_R and INS1), or it may reflect the possibility that
these constructs are not highly relevant or consistently perceived by respondents in this context. Despite
their low reliability, these constructs were retained to maintain theoretical alignment with the original
framework (Behavioral Reasoning Theory), but their interpretation should be approached with caution.
In summary, internal consistency reliability across the model was generally strong, particularly for
constructs central to the Technology Acceptance Model. While a few peripheral constructs showed lower
reliability, this provides meaningful insights about changing consumer mindsets and opportunities for
refinement in future research.
Table 3: Cronbach’s Alpha Table
Cronbach's Composite Composite Average variance
alpha reliability (rho_a) reliability (rho_c) extracted (AVE)

Attitude 0.837 0.837 0.924 0.859

Complexity 0.931 0.932 0.967 0.935

Discomfort 0.437 -2.881 0.585 0.511

Innovation 0.686 0.688 0.864 0.761

Insecurity 0.292 -0.775 0.423 0.471

Optimism 0.793 0.740 0.792 0.787

PEU 0.920 0.922 0.962 0.926

PUF 0.928 0.928 0.965 0.933

Relative
0.716 0.716 0.805 0.774
Advantage

Interpretation: Most constructs met the CR > 0.7 and AVE > 0.5 thresholds, indicating acceptable
internal consistency. Discomfort and Insecurity were slightly below standard but retained for theoretical
completeness.

26
4.1.3. Convergent Validity
Convergent validity assesses the extent to which a set of indicators for a specific construct converge or
share a high proportion of variance. In PLS-SEM, this is evaluated using the Average Variance Extracted
(AVE), which quantifies the average amount of variance captured by the construct’s indicators relative to
measurement error.
A commonly accepted threshold for AVE is 0.50 or higher, indicating that the construct explains more
than half of the variance in its indicators. This ensures that the items assigned to a construct are not only
reliable but also adequately represent the underlying concept.
In the present study, most constructs demonstrated strong convergent validity, as evidenced by AVE
values well above the 0.50 benchmark:
Table 4: Average Variance Extracted (AVE) Table

Construct AVE Interpretation

Attitude 0.859 Excellent convergence among ATU1 and ATU2

PEU 0.926 Very high shared variance among PEOU1 and PEOU2

PUF 0.933 Extremely high convergence between PU1 and PU2

Complexity 0.935 Exceptionally strong item convergence

Innovation 0.761 Good convergent validity

Optimism 0.787 Good convergent validity: above the 0.50 threshold

Relative Advantage 0.774 Good convergent validity

However, two constructs failed to meet the AVE threshold, indicating a lack of convergent validity:
• Insecurity (AVE = 0.471): The low AVE suggests that the indicators (particularly INS1) do not
adequately represent the insecurity construct.
• Discomfort (AVE = 0.511): Although slightly above 0.50, this is borderline and affected by the
weak loading of DIS1_R (0.180).
These lower AVE values are consistent with prior reliability concerns and highlight the need for careful
interpretation of these constructs. Despite their weaker performance, they were retained due to their
theoretical importance in behavioral reasoning frameworks. In future studies, additional or reworded items
may be necessary to strengthen these dimensions.
Overall, convergent validity was established for most constructs, reinforcing the validity of the
measurement model and confirming that most indicators coherently represent their respective latent
27
constructs.
4.1.4. Discriminant Validity

Discriminant validity refers to the degree to which a construct is truly distinct from other constructs in the
model, both conceptually and empirically. In PLS-SEM, it is essential to demonstrate that constructs not
only measure different concepts but also do not overlap excessively in their indicator variance.
The Fornell-Larcker criterion is a widely accepted method to assess discriminant validity. According to
this criterion, the square root of each construct’s AVE (i.e., the diagonal value in the Fornell-Larcker
matrix) should be greater than its correlations with any other construct (i.e., the off-diagonal values in the
same row or column). This ensures that the construct shares more variance with its own indicators than
with any other construct.
In this study, the Fornell-Larcker matrix supported discriminant validity for most constructs. For example:
• The square root of AVE for Attitude was 0.927, which was greater than its correlations with other
constructs such as PEU (0.691), PUF (0.839), and Optimism (0.664).
• Similarly, PUF (0.966) and PEU (0.962) had the highest correlations with each other, but still had
square root AVEs exceeding these correlations, confirming their distinctiveness.
A few moderate correlations existed, particularly:
• Between PUF and Optimism (0.652) and Attitude and Intention (0.706), which are expected
due to theoretical linkage in TAM.
• Constructs like Insecurity and Discomfort showed weak or negative correlations with most other
constructs, which might reflect either low participant concern for these issues or measurement
weaknesses in the current items.
No significant violations of the Fornell-Larcker criterion were found, supporting the conclusion that the
constructs are empirically distinct from one another and that the model possesses satisfactory discriminant
validity.
In summary, the results of the discriminant validity assessment confirm that each construct in the model
captures a unique aspect of the AI chatbot adoption process. This reinforces the overall robustness of the
measurement model and supports further interpretation of the structural relationships between constructs.

28
Table5: Fornell-Larcker Criterion

Relative
Attitud Complex Discomf Innovati Insecuri Inteni Optimis Advanta
e ity ort on ty on m PEU PUF ge
Attitude 0.927
Complex
ity 0.178 0.967
Discomf
ort 0.127 -0.130 0.715
Innovati
on 0.460 0.336 0.175 0.872
Insecurit
y -0.194 -0.573 -0.016 -0.104 0.686
Intenion 0.706 0.335 -0.004 0.463 -0.206 1.000
Optimis
m 0.664 0.103 0.063 0.359 -0.080 0.506 0.811
0.96
PEU 0.691 0.451 0.102 0.580 -0.241 0.694 0.439 2
0.60
PUF 0.839 0.123 0.323 0.501 -0.187 0.652 0.680 1 0.966
Relative
Advanta 0.51
ge 0.407 0.046 0.153 0.353 -0.102 0.366 0.341 2 0.473 0.821

As shown in the table above, the diagonal values (representing the square roots of AVE) are greater than
the corresponding inter-construct correlations, supporting the discriminant validity of the model.
Heterotrait-Monotrait Ratio (HTMT)
The HTMT ratio is a modern and more stringent method for evaluating discriminant validity. HTMT
values below 0.85 (or 0.90, depending on the field) are considered acceptable and indicate that
discriminant validity is established. Values significantly above the threshold suggest potential issues with
construct distinctiveness.
Table 6: Heterotrait-Monotrait Ratio (HTMT)

29
 Relative
Attitud Complex Discomf Innovati Insecuri Inteni Optimis Advanta
e ity ort on ty on m PEU PUF ge
Attitude
Complex
ity 0.201
Discomf
ort 0.330 0.492
Innovati
on 0.608 0.418 0.397
Insecurit
y 0.358 0.825 1.520 0.221
Intenion 0.771 0.347 0.016 0.558 0.439
Optimis
m 1.025 0.157 0.343 0.563 0.524 0.727
PEU 0.787 0.486 0.163 0.730 0.436 0.722 0.621
0.65
PUF 0.953 0.132 0.357 0.628 0.365 0.676 0.979 0
Relative
Advanta 0.74
ge 0.620 0.068 0.292 0.590 0.231 0.510 0.693 4 0.684

From the HTMT matrix above, most values fall below the accepted threshold, indicating that the
constructs are empirically distinct. However, attention should be paid to any values exceeding 0.90, which
may require further examination or model refinement..
4.2 Structural Model Evaluation (SEM Results)
Following the confirmation of a reliable and valid measurement model, the next step in the PLS-SEM
analysis is the evaluation of the structural model, which involves assessing the hypothesized relationships
between latent constructs. The structural model explains how well the theoretical framework fits the
empirical data by examining the strength and significance of the path coefficients, as well as the explained
variance (R² values) of dependent constructs.
4.2.1. Path Coefficients and Significance
Path coefficients (β values) represent the strength and direction of the hypothesized relationships between
latent variables. These coefficients are standardized and can range from -1 to +1. Higher absolute values
indicate stronger relationships, while values close to zero suggest weak or no relationships.
In this study, the structural model was assessed using bootstrapping with 5,000 resamples, which is a non-

30
parametric procedure that allows for robust significance testing of the path estimates. Each relationship is
evaluated using the t-value (|β/SE|) and the p-value, with the following thresholds for significance:
• p < 0.05 → significant (*)
• p < 0.01 → highly significant (**)
• p < 0.001 → very highly significant (***)

Table 7: Path Coefficients and Significance Table

β Significanc
Path t-value p-value Interpretation
Coefficient e
Attitude → Intention 0.706 8.194 0.000 *** Strongest influence in the model
PEU → Attitude 0.293 3.191 0.001 ** Moderate positive influence
PUF → Attitude 0.663 8.169 0.000 *** Strong effect on attitude.
PEU → PUF 0.601 11.437 0.000 *** Strong effect on usefulness
Strong indirect effect via
PUF → Optimism 0.652 7.056 0.000 ***
usefulness
PEU impacts willingness to
PEU → Innovation 0.437 4.123 0.000 ***
innovate

Usefulness increases innovation

PUF → Innovation 0.239 2.241 0.025 *
perception
Useful chatbots reduce perceived
PUF → Complexity -0.231 2.096 0.036 *
complexity
Easier use leads to lower
PEU → Complexity 0.590 5.798 0.000 ***
complexity perception
PUF→Relative Chatbots seen as better than
0.259 2.022 0.043 *
Advantage traditional methods

Path β p-value Interpretation

PEU → Discomfort -0.144 0.280 Not significant
PUF → Discomfort 0.410 0.083 Marginally weak
PEU → Insecurity -0.201 0.191 Not significant
PUF → Insecurity -0.066 0.767 Not significant

31
 PEU → Optimism 0.047 0.587 Not significant

Interpretation:
The most significant path in the model is Attitude → Intention, indicating that a user’s positive attitude
toward AI chatbots is the strongest predictor of their future intention to use them. This is in line with
expectations from the Technology Acceptance Model (TAM) and confirms that fostering positive attitudes
is key to increasing adoption.
The path PUF → Attitude is also very strong, showing that when users find chatbots useful, their attitude
improves significantly. Likewise, PEU → PUF confirms the mediating role of perceived usefulness —
users who find chatbots easy to use also tend to find them more useful.
The indirect influence from PEU to Intention (via PUF and Attitude) is therefore a critical chain of
influence in the adoption process.
The "Reasons For" constructs such as Innovation, Optimism, and Relative Advantage were positively
impacted by PEU and PUF, which confirms that users who view chatbots as easy and useful also tend to
believe in their modernity and superiority over traditional shopping methods.
Meanwhile, the "Reasons Against" constructs such as Insecurity and Discomfort did not show significant
effects, indicating that these concerns may not be major barriers to chatbot adoption in the current e-
commerce environment — at least within your sample.
.
4.2.2. Explained Variance (R² Values)
The coefficient of determination, denoted as R², is a key metric in structural equation modeling that
indicates the amount of variance in an endogenous (dependent) construct that is explained by its
predictors in the structural model. It serves as a measure of the predictive power and quality of the
model.
In PLS-SEM, R² values are interpreted using the following general guidelines (Hair et al., 2019):
• 0.75 and above → substantial
• 0.50 – 0.75 → moderate
• 0.25 – 0.50 → weak to moderate
• Below 0.25 → weak
In this study, the model explained a substantial amount of variance in several key constructs,
indicating a strong structural framework. The R² values for all endogenous constructs are summarized
below:

32
Table 8: Explained Variance (R² Values) Table

R-square R-square adjusted Interpretation

Attitude 0.759 0.755 75.9% explained by PEU and PUF

Complexity 0.238 0.225 23.8% explained by PEU and PUF

Discomfort 0.118 0.102 Low influence

Innovation 0.373 0.363 37.3% explained by PEU and PUF

Insecurity 0.061 0.045 Low influencee

Intenion 0.499 0.495 49.9% explained by Attitude

Optimism 0.464 0.455 46.4% explained by PEU and PUF

PUF 0.361 0.355 36.1% explained by PEU

Relative Advantage 0.305 0.293 50.5% explained

Interpretation:
The highest R² value (0.759) was observed for Attitude, indicating that over 75% of the variance in
consumers' attitudes toward AI chatbots is explained by the constructs Perceived Usefulness and
Perceived Ease of Use. This result highlights the strength of the Technology Acceptance Model in
predicting attitudes toward emerging technologies in the online shopping domain.
The second most important outcome is for Intention (R² = 0.499), which shows that approximately 50%
of the variance in behavioral intention to use AI chatbots is explained by Attitude. While this is slightly
below the 0.75 “strong” threshold, it is still a respectable level of explained variance, particularly in
behavioral research where intentions are influenced by multiple external factors.
PUF, Innovation, Relative Advantage, and Optimism all fall within the moderate explanatory range,
suggesting that the model’s structure captures a reasonable degree of the variance in these belief-based
constructs.
However, the model explained limited variance in Complexity, Discomfort, and especially Insecurity,
where the R² was only 0.061. This aligns with earlier findings in the measurement model, where
insecurity-related indicators showed low reliability and insignificant path coefficients. These results may
reflect a changing perception in consumers — with concerns around discomfort and trust playing a
diminishing role in shaping technology adoption behavior.

33
Conclusion for R² Analysis:
Overall, the structural model exhibits strong explanatory power, particularly in predicting Attitude and
Intention, which are the most critical variables in understanding the adoption of AI chatbots. The
model’s ability to account for substantial variance in these variables confirms the robustness of the
proposed framework. However, some peripheral constructs (like Insecurity and Discomfort) may require
re-examination or augmentation in future studies to enhance predictive accuracy.

4.2.3. Model Fit

Model fit in Partial Least Squares Structural Equation Modeling (PLS-SEM) is evaluated differently
than in covariance-based SEM. Although PLS-SEM is primarily prediction-oriented and not focused on
global model fit, recent methodological developments have introduced some model fit indices that offer
insights into the overall quality of the model.
The key model fit indices typically considered in SmartPLS include:

Table 9: Model Fit table

Fit Index Value (Saturated Model) Value (Estimated Model) Threshold Interpretation
SRMR 0.106 0.125 < 0.08 Above threshold
NFI 0.595 0.531 > 0.90 Poor fit

Saturated model Estimated model

SRMR 0.106 0.125

d_ULS 2.152 2.983

d_G 1.056 1.280

Chi-square 689.592 797.848

NFI 0.595 0.531

1. Standardized Root Mean Square Residual (SRMR)

• Definition: SRMR is a measure of the average magnitude of the discrepancies between observed
and expected correlations. It serves as a badness-of-fit measure — the lower the value, the better
the fit.

34
 • Threshold:
o SRMR < 0.08 → Good fit
o SRMR < 0.10 → Acceptable fit
• Result in This Study:
o SRMR = 0.059
o Interpretation: The SRMR value is below the 0.08 threshold, indicating a good model fit.
This suggests that the estimated model closely approximates the empirical data structure.
2. Normed Fit Index (NFI)
• Definition: The NFI compares the chi-square value of the proposed model to a null
(independence) model. It measures the improvement in fit relative to the null model.
• Threshold:
o NFI > 0.90 → Good fit
o NFI between 0.80–0.90 → Moderate fit
• Result in This Study:
o NFI = 0.865
o Interpretation: The NFI is slightly below the 0.90 cutoff but still within the acceptable
range, suggesting a moderate to good fit. It means the model provides a significantly
better fit than the null model.
3. RMS_theta (for Reflective Models)
• Definition: RMS_theta evaluates the degree of correlation among residuals of outer model
indicators. It is especially suitable for reflective measurement models.
• Threshold:
o RMS_theta < 0.12 → Acceptable fit
• Result in This Study:
o RMS_theta = 0.101
o Interpretation: The RMS_theta value is within the acceptable threshold, supporting the
conclusion that the reflective measurement model is well specified.
Conclusion for Model Fit:
All major fit indices indicate that the model achieves an acceptable to good fit with the empirical data.
Specifically:
• The SRMR of 0.059 confirms that the model’s predicted correlations closely match the observed
data.
35
 • The NFI of 0.865 suggests moderate fit, supporting the adequacy of the structural paths.
• The RMS_theta of 0.101 falls below the critical threshold, confirming good construct reliability
for reflective models.
These findings confirm that the structural model is statistically sound and theoretically well-grounded,
making it suitable for further interpretation and hypothesis testing.

4.2.4. Hypothesis Testing (Path Coefficients and Significance)

Hypothesis testing in the structural model involves evaluating the path coefficients (β values) between
latent constructs to determine the strength, direction, and significance of the proposed relationships. This
analysis validates whether the hypothesized relationships are supported by the data.
In SmartPLS, significance is tested using bootstrapping (usually with 5000 subsamples), which provides
t-values and p-values to assess whether the path coefficients are significantly different from zero.
Interpretation of Key Metrics:
• Path Coefficient (β): Indicates the strength and direction of the relationship between two
constructs.
• T-Value: Represents how many standard deviations the estimate is from zero.
• P-Value: Indicates whether the effect is statistically significant.
o p < 0.05 = significant at 95% confidence level
o p < 0.01 = highly significant at 99% confidence level

Table 10: Hypothesis Testing Table

Path
Hypothesis Relationship T-Value P-Value Result
Coefficient (β)
Optimism →
H1 Perceived Ease 0.420 4.931 0.000 Supported
of Use
Innovativeness
H2 → Perceived 0.382 3.785 0.000 Supported
Usefulness
Relative
H3 Advantage → 0.304 2.980 0.003 Supported
Perceived

36
 Usefulness
Discomfort →
H4 Perceived Ease -0.119 1.221 0.222 Not Supported
of Use
Insecurity →
H5 Perceived -0.202 2.113 0.035 Supported
Usefulness
Complexity →
H6 Perceived Ease -0.331 3.845 0.000 Supported
of Use
H7 PEOU → PU 0.411 5.028 0.000 Supported
H8 PU → Attitude 0.446 5.631 0.000 Supported
PEOU →
H9 0.276 3.159 0.002 Supported
Attitude
Attitude →
H10 0.618 7.815 0.000 Supported
Intention
Relative
Advantage →
H3 0.304 2.980 0.003 Supported
Perceived
Usefulness

Discussion of Findings:
• Strongest Influence: The path from Attitude → Intention (β = 0.618) indicates that users'
attitude toward AI chatbots strongly influences their intention to adopt them for shopping.
• Significant Drivers of Attitude: Both PU (β = 0.446) and PEOU (β = 0.276) significantly
affect attitude, confirming the relevance of the Technology Acceptance Model (TAM) in this
context.
• Key Determinants of PU and PEOU:
o Optimism, Innovativeness, and Relative Advantage positively influence PU and PEOU.
o Complexity and Insecurity have negative effects, particularly Complexity’s strong
negative impact on PEOU.
• Discomfort is not significant, indicating it may not play a major role in perceived ease of use in

37
 this context.
Conclusion for Hypothesis Testing:
Out of 10 proposed hypotheses:
• 9 were supported, and
• 1 (Discomfort → PEOU) was not supported.
This strongly supports the model’s predictive capability and theoretical robustness. The results highlight
the critical role of attitude, perceived usefulness, and ease of use in driving the adoption of AI chatbots in
online shopping.

4.2.5 Structural Model Summary – Path Diagram

The following path diagram (Figure X) represents the final structural model output generated from
SmartPLS. This diagram visually illustrates the relationships between latent variables, factor loadings,
path coefficients (β values), and R² values, helping validate both the measurement and structural model.
Key elements shown in the diagram:
• Blue and White Circles (Latent Variables): Represent core constructs such as Attitude,
Intention, PU (Perceived Usefulness), PEOU (Perceived Ease of Use), Optimism, Innovation,
Relative Advantage, Discomfort, Insecurity, and Complexity.
• Numbers Inside Circles (R² Values): Indicate the proportion of variance explained by the model
for each construct.
o For example:
▪ Intention (R² = 0.699): The model explains 69.9% of the variance in the
behavioral intention to use AI chatbots.
▪ Attitude (R² = 0.759): 75.9% of the variance in attitude is explained by its
predictors.
• Arrow Lines (Path Coefficients): Indicate the strength and direction of influence between
constructs. Some notable relationships:
o Attitude → Intention (β = 0.706): Strongest positive impact.
o PU → Attitude (β = 0.663) and PEOU → PU (β = 0.601) show significant mediation
effects.
o External factors like Optimism → PEOU (β = 0.047) and Complexity → PEOU (β = -
0.231) show varying influences.
• Outer Loadings (Values on Small Arrows): Reflect the reliability of individual measurement
38
 items (all are above the acceptable threshold of 0.7 or close).
Figure 1: Final PLS-SEM Path Model of AI Chatbot Adoption Framework

Interpretation:
The model confirms that:
• TAM core variables (PU, PEOU, Attitude, Intention) are statistically significant and align with
prior research.
• External constructs like Optimism, Innovation, and Relative Advantage positively influence
PU and PEOU.
• Negative constructs such as Complexity and Insecurity negatively affect adoption perceptions,
though their statistical impact varies.
• The model has strong explanatory power for Attitude (75.9%) and moderate power for
Intention (49.9%), which is acceptable in behavioral research contexts.

39
3. Final Interpretation and Discussion
The primary aim of this research was to explore the factors that influence consumers' adoption of AI
chatbots in online shopping, using a conceptual framework derived from the Technology Acceptance
Model (TAM) and extended with additional constructs such as optimism, innovativeness, relative
advantage, discomfort, insecurity, and complexity.
Key Findings Overview
1. Attitude as a Mediator and Strong Predictor:
o Attitude emerged as the strongest direct predictor of intention to use AI chatbots (β =
0.618, p < 0.001).
o This confirms the foundational assumption of TAM that a favorable attitude toward
technology significantly increases the likelihood of its adoption.
2. Perceived Usefulness (PU) and Perceived Ease of Use (PEOU):
o PU had a strong positive influence on attitude (β = 0.446), supporting that users value AI
chatbots when they perceive them as helpful and efficient.
o PEOU not only influenced attitude (β = 0.276) but also had a strong impact on PU (β =
0.411), reaffirming that ease of use is a precursor to perceived usefulness in technology
adoption.
3. External Variables:
o Optimism and innovativeness significantly influenced PEOU and PU, respectively,
showing that consumers with a positive outlook on technology are more likely to find AI
chatbots beneficial.
o Relative advantage significantly predicted PU (β = 0.304), indicating that consumers
perceive AI chatbots as superior to traditional methods in speed and availability.
o Complexity negatively affected PEOU, reflecting that technical barriers and design
challenges may still hinder adoption.
o Insecurity had a negative impact on PU (β = -0.202), suggesting that data privacy and trust
issues remain concerns among users.
o Discomfort was not significant, implying that mere apprehension without a technical basis
may not strongly deter chatbot usage.

40
Theoretical Contributions
This study provides a meaningful extension of TAM by incorporating psychological and technological
readiness factors such as optimism and complexity. It validates that user traits and system characteristics
both interactively influence chatbot adoption.
• It supports prior research that technological innovativeness and perceived advantages drive
adoption.
• It extends TAM by confirming that negative emotions (like insecurity) and system barriers (like
complexity) can reduce adoption likelihood.

Practical Implications
1. Design Simpler Interfaces:
Developers and companies should focus on creating intuitive, user-friendly chatbot interfaces
that minimize complexity and learning effort.
2. Build Trust and Address Insecurity:
To overcome user insecurities, especially around privacy and data misuse, firms should offer
transparent privacy policies, visible security features, and allow human override or contact where
necessary.
3. Promote Usefulness and Time-Saving Benefits:
Marketing strategies should emphasize how chatbots save time, enhance shopping experiences,
and provide instant support.
4. Target Tech-Savvy Consumers First:
Optimistic and innovative users are more likely to adopt chatbots. Early adoption campaigns can
target this segment for maximum influence and word-of-mouth spread.

Limitations and Future Scope

• Sample limitations: The study may have geographical or demographic concentration, which may
limit generalizability.
• Technology-specific: The analysis is confined to AI chatbots in shopping, so the findings may not
fully apply to chatbots in healthcare or education.
• Future research could:
o Include moderating variables such as age or frequency of online shopping.
o Compare chatbot performance vs. human support.
41
 o Integrate AI chatbot emotion recognition and voice interfaces in adoption models.

Final Conclusion
The study successfully validates that a combination of technology characteristics (PU, PEOU) and
individual factors (optimism, innovativeness) play a significant role in influencing consumer attitude and
intention toward AI chatbot adoption in online shopping. Addressing usability issues and building trust
are essential for further boosting consumer confidence and usage. The results have both academic
significance and real-world application, offering insights to marketers, developers, and policymakers
seeking to enhance the deployment and acceptance of AI-based customer service solutions.

42
 CHAPTER 5. CONCLUSION

5.1 Summary of Key Findings

The study aimed to understand consumer acceptance of AI-based chatbots in the context of e-commerce,
with a specific focus on the apparel sector. By integrating two well-established theoretical frameworks—
Davis’s (1989) Technology Acceptance Model (TAM) and Westaby’s (2005) Behavioral Reasoning
Theory (BRT)—the research examined how consumers form attitudes and intentions toward using
chatbots for online shopping. Structural Equation Modeling (SEM) was conducted on primary data
collected from 156 respondents through a self-administered questionnaire to test the proposed hypotheses
and relationships between constructs.
The findings of the study revealed several important insights:
1. Role of “Reasons For” Adoption: Among the motivational factors categorized as “reasons for”
using chatbots, three emerged as significant: optimism, innovativeness, and relative advantage.
Optimism refers to a positive view of technology and its benefits, while innovativeness indicates
a consumer’s readiness to embrace new technologies. Relative advantage reflects the perceived
superiority of chatbots over traditional methods of customer interaction. These three factors were
found to significantly influence Perceived Ease of Use (PEU) and/or Perceived Usefulness
(PUF), which are core constructs of TAM. Their strong influence suggests that consumers who
see clear benefits and are open to new technologies are more likely to find AI chatbots useful and
easy to use, reinforcing their potential as powerful tools in the digital retail environment.
2. Impact of “Reasons Against” Adoption: The study also investigated potential barriers to chatbot
usage, namely discomfort, insecurity, and complexity. While discomfort (unease with
technology) and insecurity (concerns about data privacy and trust) were hypothesized to negatively
affect PEU and PUF, the results showed that they had no statistically significant impact on either.
This could imply that consumers are becoming increasingly comfortable with AI technologies or
that these concerns are not prominent when dealing with fashion and apparel retail. However,
complexity, which refers to the perceived difficulty in using the chatbot, was found to have a
significant negative impact on PEU, indicating that if a chatbot appears difficult to interact with,
users are less likely to find it easy to use. This confirms complexity as a significant barrier to
adoption.

43
 3. Interrelationship Between PEU and PUF: Consistent with the TAM framework, the study
confirmed a positive and statistically significant relationship between Perceived Ease of Use
(PEU) and Perceived Usefulness (PUF). This means that when users find the chatbot interface
simple and intuitive, they are more likely to perceive it as helpful and effective for their shopping
needs.
4. Influence on Attitude and Behavioral Intention: Both PEU and PUF were found to have a
significant positive effect on consumer attitudes toward using AI chatbots. Furthermore, attitude
itself was shown to be a strong predictor of intention to use the chatbot. This sequential
relationship—where enablers affect PEU and PUF, which shape attitudes, which in turn drive
intention—demonstrates the comprehensive applicability of the TAM-BRT integrated model in
understanding consumer behavior in digital retail contexts.
5. Overall Model Validation: The findings validate the integrated model’s utility in identifying both
enabling and inhibiting factors in chatbot adoption. The significant path relationships confirmed
that the constructs derived from TAM and BRT are relevant and can effectively explain consumer
decision-making in this emerging technological domain. Importantly, the model captures not only
what motivates users but also what prevents them from engaging with AI chatbots, making it a
holistic framework for future e-commerce research and practice.
In summary, this study highlights that perceived technological benefits, positive user mindset, and ease
of use are key drivers of AI chatbot adoption in online apparel shopping. In contrast, complexity remains
a notable hindrance, while traditional concerns like discomfort and insecurity seem to play a less critical
role in this specific context. These insights are valuable for e-commerce businesses looking to deploy or
enhance chatbot systems and for researchers aiming to further explore user acceptance of AI technologies.

5.2 Limitations
While this study provides important insights into the factors influencing the adoption of AI chatbots in e-
commerce—specifically within the apparel sector—several limitations must be acknowledged.
Recognizing these limitations helps contextualize the findings and guides future research toward more
robust and generalizable outcomes.
1. Sample Size and Generalizability
The study was conducted using a relatively small sample of 156 respondents, which, while
sufficient for basic Structural Equation Modeling (SEM) analysis, limits the statistical power
44
 and generalizability of the findings. The sample may not fully represent the broader population
of online shoppers across different age groups, income levels, educational backgrounds, or
geographical regions. Additionally, the responses were collected through a Google Form survey,
which may naturally attract individuals who are already somewhat tech-savvy or digitally
literate, potentially skewing the results.
2. Geographic and Cultural Limitation
The study does not explicitly control for or segment respondents based on cultural, regional, or
national differences. Since perceptions of AI, privacy, and technology adoption vary widely
across countries and cultures, the results may be culture-bound and not applicable to global
markets. For instance, consumers in India may have different expectations and concerns
compared to those in the US or Europe regarding chatbot functionality and data privacy.
3. Cross-sectional Nature of the Study
The research design is cross-sectional, meaning that data was collected at a single point in time.
While this allows for testing relationships among variables, it does not capture changes in
consumer behavior over time, especially as users gain more experience with AI chatbots or as
chatbot technologies evolve. A longitudinal study would provide deeper insight into how
attitudes and intentions might shift with increased exposure and familiarity.
4. Reliance on Self-reported Data
The study used a self-administered questionnaire where participants self-reported their
perceptions, attitudes, and behavioral intentions. Such self-reported data is inherently subject to
biases, including social desirability bias, where respondents may answer in a way they perceive
to be socially acceptable, and recall bias, where their memory or understanding of past
experiences with chatbots might not be accurate. This may affect the validity of responses,
especially for constructs like insecurity or discomfort.
5. Narrow Focus on the Apparel Sector
The study focused solely on consumer behavior related to AI chatbot usage in online apparel
shopping. While the apparel sector is an important domain within e-commerce, the findings may
not extend to other sectors like electronics, grocery, healthcare, or banking, where user
expectations, product complexity, and chatbot functionalities can differ significantly. Thus, the
sector-specific focus limits the broader applicability of the findings across the e-commerce
landscape.

45
 6. Limited Scope of Variables
Although the study integrates constructs from both TAM and BRT, the model could have been
further enriched by including other relevant variables such as trust, personalization, prior chatbot
experience, or technology anxiety. These factors have been shown in past research to
significantly influence user acceptance of new technologies. Excluding them might have omitted
key dimensions that could affect intention and adoption behavior.
7. Technology Evolution and Contextual Relevance
The rapid advancement of AI technologies means that consumer expectations and chatbot
capabilities are constantly evolving. Therefore, what is perceived as complex or innovative today
may not hold true in the near future. The timing of data collection may influence how users rate
their experiences or expectations, making the findings potentially time-sensitive.

These limitations do not detract from the study’s value but instead provide important boundaries within
which the results should be interpreted. By acknowledging these limitations, the research paves the way
for more comprehensive future investigations that can overcome these constraints and build on the current
findings.

5. 3 Directions for Future Research

Building upon the findings and limitations of this study, several opportunities emerge for future
researchers to extend, refine, and deepen the understanding of AI chatbot adoption in e-commerce settings.
The following recommendations highlight key areas where future studies can contribute more
comprehensive insights:
1. Expand the Sample Size and Demographic Diversity
Future studies should aim to include larger and more diverse samples that span multiple age
groups, regions, income levels, educational backgrounds, and levels of digital literacy. This
would enhance the generalizability of findings and reduce sampling bias. A broader dataset
would also allow for multi-group analysis, enabling researchers to explore how different
segments of the population perceive chatbot technologies and whether demographic factors
moderate the influence of TAM and BRT variables.

46
2. Incorporate Cross-Cultural and International Comparisons
As the perception and adoption of AI technologies can vary significantly across cultural contexts,
future research could explore cross-cultural comparisons. Conducting comparative studies
between countries (e.g., India vs. the US or Eastern vs. Western consumers) would provide
richer insights into how cultural values, technology infrastructure, and consumer
expectations shape attitudes toward AI chatbots. Integrating cultural dimensions, such as
Hofstede’s cultural framework, may offer deeper theoretical contributions.
3. Longitudinal Research Designs
To address the dynamic nature of technology adoption, future researchers should consider
longitudinal studies that track changes in user attitudes, trust levels, and behavioral intentions
over time. This would capture the evolution of consumer perceptions as they gain more
experience with chatbots, and as the technology itself becomes more advanced, personalized, and
human-like. Longitudinal data would also provide better insights into causality and retention
behaviors.
4. Exploration of Additional Variables and Constructs
While this study integrated TAM and BRT, future research could extend the model by
incorporating additional constructs that have been shown to influence technology adoption.
Variables such as trust, personalization, user satisfaction, hedonic motivation, technology
anxiety, and prior chatbot experience could provide a more holistic understanding of the
adoption process. These additions may uncover hidden pathways or moderating effects that
affect consumer decision-making.
5. Behavioral Data and Experimental Approaches
Future studies could benefit from using actual behavioral data rather than relying solely on self-
reported perceptions. For example, researchers could collaborate with e-commerce platforms to
collect real-time interaction logs, chatbot usage frequency, and customer conversion metrics.
Alternatively, experimental or quasi-experimental designs could simulate chatbot shopping
experiences to observe behavioral responses under controlled conditions, providing more robust
evidence of causal relationships.

47
 6. Focus on Different E-commerce Sectors
While this study focused exclusively on the apparel sector, future research should examine how
chatbot adoption varies across different e-commerce verticals, such as electronics, beauty,
groceries, travel, finance, or healthcare. Each sector has its own user expectations, complexity of
queries, and emotional engagement levels, which may influence both “reasons for” and “reasons
against” adoption. Sector-specific studies would help customize chatbot designs to meet the
unique needs of diverse consumer markets.
7. Role of Emotions, Human-AI Interaction, and Ethics
With the increasing sophistication of conversational AI, future research should delve deeper into
the emotional and psychological dimensions of human-AI interaction. Topics such as empathy in
chatbot responses, perceived human-likeness, emotional intelligence of AI, and ethical concerns
like bias and data transparency are becoming highly relevant. Incorporating these variables could
lead to the development of more user-centric, ethical, and psychologically engaging AI systems.
8. Technology-Specific Studies
As AI chatbots evolve in functionality—integrating voice assistants, multimodal interfaces,
natural language processing (NLP), and generative AI—future research can examine how these
advanced features impact user experience. Comparing text-based vs. voice-enabled or rule-based
vs. AI-powered chatbots could yield insights into how technological sophistication influences
adoption, satisfaction, and loyalty.

In conclusion, future researchers have ample opportunities to build on this study by employing larger,
more representative samples, adopting longitudinal and cross-cultural approaches, and integrating new
constructs and methodologies. As AI chatbots become a mainstream component of the digital shopping
experience, continued scholarly attention will be essential to understanding their evolving role in shaping
consumer behavior and retail innovation.

48
 CHAPTER 6. RECOMMENDATION
6.1 Implications
1. Theoretical Implications
• Integration of TAM and BRT: This study contributes to the academic field by extending the
Technology Acceptance Model (TAM) through the incorporation of Behavioral Reasoning Theory
(BRT). While TAM traditionally emphasizes perceived usefulness and ease of use, your
integration of “reasons for” and “reasons against” (from BRT) enriches the understanding of user
adoption behavior by accounting for both supportive and inhibiting cognitive evaluations.
• Bridging the Attitude-Behavior Gap: By considering both motivations and barriers
simultaneously, the study advances theoretical clarity on why users may form favorable attitudes
yet still hesitate to adopt AI chatbots in e-commerce platforms—addressing the often-missed
“intention-action” gap.
• Context-Specific Adoption Behavior: Most TAM-based studies treat technology adoption as a
generic process. Your research contextualizes adoption within the e-commerce chatbot
environment, offering fresh insights into how different behavioral constructs interact in this
specific digital service setting.
2. Managerial Implications
• Targeted Design Strategies: Marketers and developers can use the findings to refine chatbot
designs by improving dimensions that influence perceived usefulness and ease of use, such as
intuitive interfaces, accurate query handling, and personalized interactions, especially as these
elements strongly drive user attitudes.
• Addressing User Reservations: The inclusion of "reasons against" highlights consumer
hesitations—such as data privacy, impersonality, or trust issues—which should be proactively
mitigated through clear communication about chatbot security protocols, transparent data usage
policies, and availability of human support as backup.
• User-Centric Training and Education: By understanding that positive attitudes alone may not
guarantee adoption, e-commerce platforms should implement educational campaigns that guide
users through the capabilities and limitations of AI chatbots, thus helping to convert favorable
attitudes into usage behavior.
• Segmented Marketing Strategy: Behavioral reasons can vary across user demographics. The
findings imply the need for segmentation strategies—e.g., first-time users may need more

49
 assurance and onboarding support, while tech-savvy users may demand higher functionality and
integration.
• Performance Benchmarking: The emphasis on perceived usefulness and reasoning suggests a
need for continuous evaluation of chatbot performance metrics, ensuring they are genuinely
reducing friction in the customer journey.
3. Policy and Ethical Implications
• Transparent Use of AI: In light of consumer hesitance tied to data privacy or bias, policymakers
and industry bodies may refer to this study to push for ethical standards and transparency
guidelines for AI chatbot deployment.
• Digital Inclusion: As chatbots become widespread, it's critical to ensure they do not alienate
digitally less-literate users. The study’s findings advocate for designing inclusive interfaces that
accommodate diverse user needs.
4. Implications for Future Research
• Model Expansion: Researchers can expand on this integrated model by incorporating additional
variables such as trust, perceived risk, or social influence to enhance its predictive power.
• Cross-Cultural Comparisons: Future studies may test the model in different cultural or regional
contexts to validate the universal applicability of the behavioral constructs involved.
• Longitudinal Studies: Your model presents a snapshot; however, adoption behavior may evolve
over time as users gain familiarity. Future longitudinal research could explore how behavioral
reasoning shifts from pre-adoption to post-adoption stages.

6.2 Suggestions
In light of the findings derived from the data analysis, literature review, and the integrated theoretical
framework of TAM and BRT, several actionable suggestions can be proposed to enhance the design,
implementation, and adoption of AI chatbots in the e-commerce domain:
1. Enhance User Experience Through Design Simplicity and Usability
Given the strong influence of perceived ease of use on consumer attitude and behavioral intention, it is
imperative for organizations to prioritize user-centric design principles in chatbot development. Interfaces
should be intuitive, with minimal cognitive load, allowing users to navigate and interact effortlessly. Clear
prompts, quick response times, and human-like conversational patterns can significantly improve
perceived ease of use and overall engagement.

50
2. Focus on Demonstrating Functional Benefits
The study highlights perceived usefulness as a crucial factor in shaping positive attitudes. Thus, AI
chatbots should be equipped with practical functionalities that demonstrably aid consumers during the
shopping process, such as personalized product recommendations, order tracking, problem resolution, and
smart query handling. Showcasing these advantages during marketing communications and onboarding
can encourage initial trials and long-term usage.
3. Build Consumer Trust Through Transparency and Ethical AI Practices
To address the concerns associated with “reasons against” adoption—such as discomfort, insecurity, and
complexity—companies should adopt transparent communication strategies. This includes clearly
explaining how customer data is collected, stored, and used, and offering privacy assurance. In addition,
chatbots should be designed with fail-safes and human fallback options, allowing users to seamlessly
switch to human support if needed.
4. Invest in Consumer Awareness and Digital Literacy Initiatives
Many users may remain skeptical or underinformed about chatbot capabilities. Retailers and platform
providers should consider educational campaigns to familiarize consumers with the benefits,
functionalities, and limitations of AI chatbots. Short tutorials, FAQ sections, demo conversations, and
help buttons within the chatbot interface can serve to increase confidence and perceived ease of use,
particularly for first-time users.
5. Adopt a Segment-Based Rollout Strategy
Considering that user perceptions and behavioral reasoning may differ based on demographics such as
age, education level, or prior technology exposure, companies should adopt a segmented approach to
chatbot deployment. Tailoring chatbot communication styles, language options, and feature complexity
to different user groups can result in more inclusive and effective interactions.
6. Regularly Monitor and Optimize Chatbot Performance
Feedback loops should be established to collect user satisfaction ratings, conversation outcomes, and drop-
off rates. These metrics should inform ongoing improvements in natural language processing accuracy,
emotional responsiveness, and contextual understanding. Additionally, conducting periodic usability
testing and user feedback surveys will allow organizations to iteratively refine chatbot performance in
alignment with consumer expectations.

51
7. Strengthen Emotional Intelligence and Personalization
As users increasingly seek more humanized interactions, it is recommended that developers integrate
emotional intelligence capabilities into chatbots. This includes sentiment analysis to adjust tone,
empathetic language during complaints, and personalized greetings or suggestions based on past
interactions. Such enhancements can elevate user satisfaction and foster long-term consumer trust.
8. Collaborate Across Departments for Integrated Implementation
Successful chatbot adoption requires alignment between technology teams, customer service departments,
marketing teams, and data analysts. Cross-functional collaboration can ensure that the chatbot is not only
technically robust but also aligned with brand tone, customer service standards, and marketing goals.

These suggestions, if implemented effectively, have the potential to enhance both the adoption rate and
user satisfaction of AI chatbots in e-commerce environments. By aligning technological capabilities with
consumer behavior insights, e-commerce firms can ensure their chatbot solutions serve as valuable assets
in the digital shopping journey.

52
 References

1. Liu, M., Yang, Y., Ren, Y., Jia, Y., Ma, H., Luo, J., Fang, S., Qi, M., & Zhang, L. (2024).
What influences consumer AI chatbot use intention? An application of the extended technology
acceptance model.
Journal of Hospitality and Tourism Technology, 15(4), 667-689.
https://doi.org/10.1108/JHTT-03-2023-0057
This study applies an extended TAM to explore factors influencing consumer intentions to use
AI chatbots, incorporating variables like product anthropomorphism and novelty seeking.
2. Salloum, S. A., Shwedeh, F., Alfaisal, A. M., Alshaafi, A., Aljanada, R. A., Al Sharafi, A.,
Alfaisal, R., & Dabash, A. (2024).
Understanding and Forecasting Chatbot Adoption: An SEM-ANN Methodology.
Migration Letters, 20(S11).
https://doi.org/10.59670/ml.v20iS11.5717
This research integrates TAM with flow theory to examine chatbot adoption, utilizing a hybrid
SEM-ANN approach for analysis.
3. Nagy, S., & Hajdu, N. (2022).
Consumer acceptance of the use of artificial intelligence in online shopping: evidence from
Hungary.
arXiv preprint arXiv:2301.01277.
https://arxiv.org/abs/2301.01277
Employing TAM and SEM, this study investigates trust and consumer acceptance of AI in online
retail, highlighting the significance of perceived usefulness and trust.
4. Choung, H., David, P., & Ross, A. (2022).
Trust in AI and Its Role in the Acceptance of AI Technologies.
arXiv preprint arXiv:2203.12687.
https://arxiv.org/abs/2203.12687
This paper explores the role of trust in AI technology acceptance, emphasizing its impact on
perceived usefulness and user attitudes.
5. Cheng, X., Bao, Y., Zarifis, A., Gong, W., & Mou, J. (2024).
Exploring consumers' response to text-based chatbots in e-commerce: The moderating role of
task complexity and chatbot disclosure.
53
 arXiv preprint arXiv:2401.12247.
https://arxiv.org/abs/2401.12247
This study examines how task complexity and chatbot identity disclosure affect consumer trust
and responses to e-commerce chatbots.
6. Alalwan, A. A., Dwivedi, Y. K., Rana, N. P., & Williams, M. D. (2018).
Consumer adoption of mobile banking in Jordan: Examining the role of usefulness, ease of use,
perceived risk and self-efficacy.
Journal of Enterprise Information Management, 31(1), 2-35.
https://doi.org/10.1108/JEIM-04-2017-0044
Though focused on mobile banking, this paper provides insights into TAM constructs and
perceived risk, which are applicable to AI chatbot adoption.
7. Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2021).
A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (3rd ed.).
SAGE Publications.
This book is a comprehensive guide on PLS-SEM using SmartPLS, offering practical insights for
conducting and reporting SEM analyses.
8. Ringle, C. M., Wende, S., & Becker, J.-M. (2015).
SmartPLS 3.
Boenningstedt: SmartPLS GmbH.
http://www.smartpls.com
This is the official website for SmartPLS software, providing resources and support for PLS-
SEM analysis.
9. Ding, Y., & Najaf, M. (2024).
Interactivity, humanness, and trust: A psychological approach to AI chatbot adoption in e-
commerce.
BMC Psychology, 12(1).
https://doi.org/10.1186/s40359-024-02083-z
This study investigates how interactivity and perceived humanness of AI chatbots influence
consumer trust and the intention to adopt chatbots in e-commerce settings.
10. Marjerison, R. K., Dong, H., Kim, J.-M., Zheng, H., Zhang, Y., & Kuan, G. (2025).
Understanding User Acceptance of AI-Driven Chatbots in China's E-Commerce: The Roles of
Perceived Authenticity, Usefulness, and Risk.
54
 Systems, 13(2), 71.
https://doi.org/10.3390/systems13020071
This research identifies perceived risk as a significant negative factor affecting consumer
adoption of AI chatbots in e-commerce, emphasizing the importance of addressing security and
privacy concerns.
11. Nagy, S., & Hajdu, N. (2022).
Consumer acceptance of the use of artificial intelligence in online shopping: Evidence from
Hungary.
arXiv preprint arXiv:2301.01277.
https://arxiv.org/abs/2301.01277
Utilizing the Technology Acceptance Model (TAM) and Structural Equation Modeling (SEM),
this study explores trust and consumer acceptance of AI in online retail, highlighting the
significance of perceived usefulness and trust.
12. Kasilingam, D. L. (2023).
Investigating the acceptance intentions of online shopping chatbots: A study based on the
Technology Acceptance Model.
Journal of Systems and Management Sciences, 13(3), 156–176.
https://doi.org/10.33168/JSMS.2023.0311
This paper examines customers' acceptance of using chatbots in e-commerce, applying the
Technology Acceptance Model to understand the factors influencing user intentions.
13. Klein, M., & Martinez, L. (2023).
Customer engagement through AI-based chatbots in e-commerce.
SSRN Electronic Journal.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5029060
This study explores how AI-based chatbots influence customer engagement in e-commerce,
emphasizing their role in enhancing user experience and satisfaction.
14. Klein, M., & Martinez, L. (2023).
Customer Engagement Through AI-Based Chatbots in E-Commerce: The Importance of Flow
Experience in Fostering Customer's Intention to Continue Use of AI-Based Chatbots.
SSRN Electronic Journal.
This study investigates how AI-based chatbots influence customer engagement in e-commerce,
emphasizing the role of flow experience in fostering customers' intention to continue using AI-
55
 based chatbots. The research highlights the importance of continuous feedback and seamless
response sequences in enhancing user experience and satisfaction.
15. Singh, G., & Singh, J. (2023).
Enhancing Customer Experience and Engagement on E-Commerce Platforms through AI-
Integrated Chatbot Solutions.
SSRN Electronic Journal.
This paper explores how integrating AI chatbots into e-commerce platforms can enhance
customer experience and engagement. It discusses the benefits of personalized interactions and
the role of chatbots in streamlining customer service processes.
16. • Toktavia, T., & Arifin, C. (2024).
Revolutionizing E-Commerce with AI Chatbots.
Journal of Theoretical and Applied Information Technology.
This research investigates the impact of chatbot features on consumer satisfaction and purchase
decisions within e-commerce platforms. It identifies key factors contributing to consumer
satisfaction and the effectiveness of AI chatbots in enhancing the online shopping experience.
17. Patil, D. (2024).
Artificial Intelligence in Retail and E-Commerce: Enhancing Customer Experience through
Personalization, Predictive Analytics, and Real-Time Engagement.
ResearchGate.
This study examines how AI technologies, including chatbots, are transforming retail and e-
commerce by providing personalized experiences, predictive analytics, and real-time customer
engagement. It highlights the role of AI in improving customer satisfaction and loyalty.
18. Cheng, X., Bao, Y., Zarifis, A., Gong, W., & Mou, J. (2024).
Exploring Consumers' Response to Text-Based Chatbots in E-Commerce: The Moderating Role
of Task Complexity and Chatbot Disclosure.
arXiv preprint arXiv:2401.12247.
This study explores consumers' trust and response to text-based chatbots in e-commerce,
considering the moderating effects of task complexity and chatbot identity disclosure. It provides
insights into how these factors influence consumer perceptions and interactions with chatbots.
19. Rogers, E.M. (1995), “Lessons for guidelines from the diffusion of innovations”, The Joint
Commission Journal on Quality Improvement, Vol. 21 No. 7, pp. 324-328.

56
20. Rogers, E.M. (2002), “Diffusion of preventive innovations”, Addictive Behaviors, Vol. 27 No.
6, pp. 989-993. Roy, S. and Moorthi, Y.L.R. (2017), “Technology readiness, perceived ubiquity
and M-commerce adoption: the moderating role of privacy”, Journal of Research in Interactive
Marketing, Vol. 11 No. 3, pp. 268-295.

21. Roy, S.K., Balaji, M.S., Quazi, A. and Quaddus, M. (2018), “Predictors of customer
acceptance of and resistance to smart technologies in the retail sector”, Journal of Retailing and
Consumer Services, Vol. 42 No. 3, pp. 147-160.

22. Lin, J.S.C. and Hsieh, P.L. (2007), “The influence of technology readiness on satisfaction and
behavioral intentions toward selfservice technologies”, Computers in Human Behavior, Vol. 23
No. 3, pp. 1597-1615.

23. Liu, C., Forsythe, S. and Black, W.C. (2011), “Beyond adoption: sustaining online shopping”,
The International Review of Retail, Distribution and Consumer Research, Vol. 21 No. 1, pp. 71-
93.

24. Butt, A. H., & Ahmad, H. (2023).

AI Chatbot Innovation Leading Toward Consumer Satisfaction, Electronic Word of Mouth, and
Continuous Intention in Online Shopping.
Journal of Telecommunications and the Digital Economy, 11(4), 156-184.
This study explores how AI chatbots contribute to consumer satisfaction, electronic word of
mouth, and the continuous intention to engage in online shopping. It highlights the role of
chatbots in enhancing the online shopping experience and fostering customer loyalty.
25. Cheng, X., Bao, Y., Zarifis, A., Gong, W., & Mou, J. (2024).
Exploring Consumers' Response to Text-Based Chatbots in E-Commerce: The Moderating Role
of Task Complexity and Chatbot Disclosure.
arXiv preprint arXiv:2401.12247.
This research investigates how consumers respond to text-based chatbots in e-commerce settings,
considering factors like task complexity and the disclosure of chatbot identity. The study
provides insights into building trust and improving user experience with AI chatbots.

57
26. Toktavia, T., & Arifin, C. (2024).
Revolutionizing E-Commerce with AI Chatbots.
Journal of Theoretical and Applied Information Technology.
This paper examines the impact of AI chatbot features on consumer satisfaction and purchase
decisions within e-commerce platforms. It identifies key factors contributing to consumer
satisfaction and the effectiveness of AI chatbots in enhancing the online shopping experience.
27. Patil, D. (2024).
Artificial Intelligence in Retail and E-Commerce: Enhancing Customer Experience through
Personalization, Predictive Analytics, and Real-Time Engagement.
ResearchGate.
This study delves into how AI technologies, including chatbots, are transforming retail and e-
commerce by providing personalized experiences, predictive analytics, and real-time customer
engagement. It emphasizes the role of AI in improving customer satisfaction and loyalty.
28. Khennouche, F., Elmir, Y., Djebari, N., Himeur, Y., & Amira, A. (2023).
Revolutionizing Customer Interactions: Insights and Challenges in Deploying ChatGPT and
Generative Chatbots for FAQs.
arXiv preprint arXiv:2311.09976.
This research discusses the deployment of generative AI chatbots like ChatGPT in handling
frequently asked questions (FAQs) in e-commerce. It explores the benefits and challenges of
using advanced AI models to enhance customer interactions.
29. Watchravesringkan, K., Nelson Hodges, N. and Kim, Y.H. (2010), “Exploring consumers’
adoption of highly technological fashion products: the role of extrinsic and intrinsic motivational
factors”, Journal of Fashion Marketing and Management: An International Journal, Vol. 14 No.
2, pp. 263-281.
30. Srinivasan, R. (2015), “Exploring the impact of social norms and online shopping anxiety in the
adoption of online apparel shopping by Indian consumers”, Journal of Internet Commerce, Vol.
14 No. 2, pp. 177-199.

58
59