Adoption of Artificial Intelligence-Based Employee Experience Chatbots

Abstract

The Adoption of Artificial Intelligence-Based Employee Experience Chatbots model and psychometric measurement instrument, formulated by Pillai, Ghanghorkar, Sivathanu, Algharabat, and Rana (2024), provides an empirical operationalization of the cognitive, motivational, and psychological determinants governing workforce acceptance of conversational artificial intelligence (AI) systems deployed within human resource management (HRM) and employee experience (EEX) architectures. Grounded in Behavioral Reasoning Theory (BRT), the instrument captures how higher-order personal values (specifically, openness to change) interface with contextualized cognitions—divided cleanly into contextualized reasons for adoption and reasons against adoption—to forge overall employee attitudes and subsequent behavioral intentions to accept EEX conversational agents. The multi-dimensional framework integrates key technical and relational affordances: reasons for adoption encompass perceived anthropomorphism, perceived intelligence, interactivity, and personalization, whereas reasons against adoption embody salient cognitive and psychological barriers, notably technological anxiety, perceived risk, and linguistic or language barriers.

Methodologically validated through partial least squares structural equation modeling (PLS-SEM) among operational and managerial personnel across corporate enterprise sectors, including information technology (IT), information technology enabled services (ITeS), and e-commerce companies, the questionnaire utilizes a standardized five-point Likert response format ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). The empirical evaluation demonstrated exceptional psychometric properties: internal consistency was evidenced by composite reliability (CR) and Cronbach’s alpha values uniformly exceeding the recommended 0.80 benchmark across all latent constructs. Convergent validity was authenticated with average variance extracted (AVE) values exceeding 0.50, and discriminant validity was rigorously established according to the Fornell-Larcker criterion and cross-loading assessments. Evaluation of common method variance using Harman’s single-factor test yielded a variance of 26.82%, well below critical thresholds. The instrument serves as a critical diagnostic and evaluative tool for organizational psychologists, human resource leaders, and human-computer interaction (HCI) researchers seeking to diagnose resistance, facilitate user-centric AI adoption, and optimize modern workplace digital transformation.

Keywords

Artificial Intelligence Adoption, Employee Experience Chatbots, Behavioral Reasoning Theory, Perceived Anthropomorphism, Perceived Intelligence, Technological Anxiety, Perceived Risk, Organizational Digital Transformation, Human-Computer Interaction, Conversational Agents, Workplace Technology Acceptance

Authors

The scale and structural adoption model were conceived, developed, and validated by an international team of behavioral science and digital business scholars:

• Rajasshrie Pillai — Department of Management, Symbiosis Centre for Management and Human Resource Development (SCMHRD), Symbiosis International (Deemed University), Pune, India. Specializes in conversational commerce, artificial intelligence in organizational contexts, and human resource analytics.
• Yamini Ghanghorkar — Symbiosis Centre for Information Technology (SCIT), Symbiosis International (Deemed University), Pune, India. Focuses on enterprise systems, workplace digitisation, and employee behavioral dynamics.
• Brijesh Sivathanu — Department of Operations and Information Technology, Symbiosis Institute of Business Management (SIBM), Symbiosis International (Deemed University), Pune, India. Research trajectory concentrates on artificial intelligence, algorithmic workforce management, and advanced consumer behavior modeling.
• Raed Algharabat — Department of Management and Marketing, College of Business and Economics, Qatar University, Doha, Qatar. (ORCID: 0000-0002-3904-4340). Renowned for seminal contributions to digital marketing, telepresence, customer engagement, and algorithmic customer-agent interfaces.
• Nripendra P. Rana — College of Business and Economics, Qatar University, Doha, Qatar; previously School of Management, Swansea University, United Kingdom. Globally recognized scholar in emerging technology adoption, e-governance, and structural equation modeling applications.

Purpose

The rapid integration of conversational artificial intelligence into workplace ecosystems has transformed standard operating workflows within corporate enterprises. Organizations increasingly replace traditional human resource communication helpdesks, onboarding portals, and internal support services with autonomous natural language processing (NLP) dialogue agents, commonly known as Employee Experience (EEX) chatbots. While conversational interfaces promise continuous real-time support, administrative agility, and enhanced transactional efficiency, their implementation frequently encounters psychological friction, operational reluctance, or outright rejection from employees. Existing technology adoption metrics have historically relied on generic models, such as the Technology Acceptance Model (TAM) or the Unified Theory of Acceptance and Use of Technology (UTAUT). While robust, these frameworks often frame adoption decisions along linear lines of utilitarian efficiency (such as perceived usefulness and perceived ease of use) and social influence, largely omitting the complex psychological push-and-pull factors underlying human-AI collaboration.

The primary purpose of the Adoption of Artificial Intelligence-Based Employee Experience Chatbots instrument is to provide an empirically validated, theoretically sophisticated diagnostic inventory designed specifically to measure employees’ psychological attitudes and behavioral intention to adopt and interact with EEX conversational agents. The tool uncovers the cognitive tension that arises when employees contemplate relinquishing human-mediated professional interactions in favor of synthetic, algorithmic systems. By examining both driving rationales (“reasons for”) and inhibitive barriers (“reasons against”) simultaneously, the instrument bridges significant gaps in applied workplace psychometrics and modern management research.

In academic research environments, this scale offers a calibrated operational platform to explore the boundary conditions of algorithmic trust, human-machine dynamics, and relational anthropomorphism within work environments. In corporate applications, the instrument serves organizational development consultants, chief human resource officers (CHROs), and digital change leaders as an actionable diagnostic audit. By administering this survey prior to, during, or following an enterprise-wide conversational AI rollout, leadership can pinpoint the precise psycho-social origins of employee hesitance—whether stemming from technological intimidation, opaque data privacy fears, or perceived conversational inflexibility. Consequently, intervention strategies can be tailored to target specific psychological friction points rather than relying on generic, ineffective enterprise training programs.

Psychological Construct

The construct of artificial intelligence adoption within the employee experience domain is conceptualized as an intricate nomological network of personal values, domain-specific reasons, general attitudes, and ultimate behavioral intentions. The model delineates this system across four core psychometric conceptual layers:

1. Core Human Values: Openness to Change

Rooted in Schwartz’s structural theory of universal human values, Openness to Change reflects an employee’s intrinsic motivational orientation toward novelty, autonomy, innovation, and psychological flexibility. In the context of conversational agent adoption, individuals who exhibit higher openness are intrinsically predisposed toward algorithmic exploration, exhibit cognitive curiosity toward computational tools, and perceive technological modernization not as an existential threat, but as an opportunity for personal and organizational mastery.

2. Domain-Specific “Reasons For” Adoption

Reasons for adoption represent distinct cognitive justifications that directly promote favorable perceptions of conversational systems. These reflect systemic, functional, and relational affordances:

• Perceived Anthropomorphism: The psychological tendency of the employee to attribute human-like attributes, conversational warmth, intentionality, and communicative responsiveness to the algorithmic agent. This subjective humanization diminishes psychological distance, engendering higher relational comfort during organizational inquiries.
• Perceived Intelligence: The degree to which an employee evaluates the conversational agent as capable, logically competent, context-aware, and effective at knowledge retrieval and complex query resolution. High perceived intelligence assures the employee that the chatbot is cognitively equipped to navigate administrative nuances.
• Interactivity: The user’s appraisal of two-way dynamic responsiveness, communicative synchronicity, speed of feedback, and active dialogue management displayed by the chatbot. Interactivity fosters a continuous, fluid conversational loop rather than an iterative, transactional data dump.
• Personalization: The capability of the synthetic agent to deliver customized solutions tailored to the individual employee’s organizational status, history, unique queries, and specific communicative preferences, signaling that the system respects individual identity.

3. Domain-Specific “Reasons Against” Adoption

Conversely, reasons against adoption embody specific cognitive, psychological, and systemic barriers that justify technological avoidance or refusal:

• Technological Anxiety: A discrete state of apprehension, fear, and emotional distress evoked when an employee is confronted with the necessity of interacting with advanced automated computing systems. This includes anxieties regarding technological incompetence, algorithmic errors, or the displacement of intuitive human agency.
• Perceived Risk: An individual’s subjective estimation of adverse consequences, uncertainty, and prospective negative outcomes associated with utilizing an EEX chatbot. These risks frequently encompass organizational exposure, data confidentiality breaches, surveillance anxiety, or potential algorithmic errors that could negatively affect performance evaluations or employment benefits.
• Language Barrier: The employee’s perception of computational inadequacy regarding linguistic nuance, idioms, non-standard dialect interpretation, tone, or colloquial language. When users perceive that an AI agent cannot interpret localized linguistic syntax, communication frustration mounts rapidly.

4. Attitudinal and Behavioral Outcomes

In harmony with cognitive psychology frameworks, these contextual reasons channel directly into the employee’s Attitude toward EEX Chatbots—an overarching, enduring positive or negative emotional and evaluative stance toward working with the tool. In turn, attitude, supplemented by direct cognitive pathways from specific reasons, serves as the immediate precursor driving Adoption Intention, which denotes the employee’s conscious, volitional willingness to actively incorporate the AI chatbot into their ongoing corporate communication practices.

Theoretical Framework

The theoretical framework underpinning this measurement tool is Behavioral Reasoning Theory (BRT), originally formulated by Westaby (2005). Traditional cognitive frameworks, such as the Theory of Reasoned Action (TRA), the Theory of Planned Behavior (TPB), and the Technology Acceptance Model (TAM), posit that beliefs directly determine attitudes, which subsequently predict intentions and observable behaviors. While empirically validated across decades of research, these standard theories operate under several critical limitations when applied to disruptive workplace technologies:

First, traditional acceptance theories frequently model antecedents along a single continuum of utilitarian benefits, presuming that low adoption stems merely from the absence of positive utility (e.g., lack of usefulness). Behavioral Reasoning Theory challenges this paradigm by asserting that decision-making processes involve distinct, non-compensatory cognitive systems: individuals generate distinct reasons for an action and independent reasons against that same action simultaneously. A corporate employee may fully recognize that an AI chatbot is highly intelligent and interactive (compelling reasons for adoption), yet concurrently refuse to utilize the tool due to overwhelming data privacy anxieties or communicative alienation (salient reasons against adoption). Traditional linear models routinely miss this internal cognitive conflict.

Second, BRT incorporates higher-order human values (e.g., openness to change) as macro-level determinants that influence specific contextual reasons. In the Pillai et al. (2024) conceptual model, core personal values do not merely operate as passive distal antecedents; they actively direct how employees process situational cues. Employees who value self-direction and stimulation gravitate toward reasons celebrating the chatbot’s interactive nature and technical prowess, while individuals low in openness may accentuate the risks and communicative disruptions of machine interfaces.

Third, BRT posits dual pathways: reasons directly shape generalized attitudes, but they can also bypass generalized attitudes to directly dictate behavioral intentions. This structural architecture captures intuitive heuristic decision-making, where a critical barrier (such as intense perceived risk) can abruptly derail technology adoption without necessitating an intermediary, fully crystallized negative attitude. By anchoring the measurement instrument in BRT, Pillai and colleagues established a psychometric framework capable of untangling the multi-dimensional cognitive realities characterizing the modern algorithmic workplace.

Validity

The measurement tool was subjected to rigorous psychometric scrutiny to establish its construct, content, convergent, and discriminant validity across target organizational settings.

Face and Content Validity

To confirm that items accurately represented the targeted theoretical dimensions, draft operational indicators were adapted from established literature (e.g., Claudy et al., 2015; Gupta & Arora, 2017; Westaby, 2005) and submitted to an expert panel of organizational psychologists, digital technology academics, and senior HR professionals. Iterative item modifications were conducted based on subject matter expert feedback, eliminating structural ambiguities and ensuring the operational phrasing aligned with modern enterprise conversational AI interactions.

Convergent Validity

Convergent validity evaluates the extent to which distinct items measuring a specific construct intercorrelate and share substantial common variance. Utilizing Partial Least Squares Structural Equation Modeling (PLS-SEM), convergent validity was authenticated based on the guidelines established by Hair, Hult, Ringle, and Sarstedt (2017). Across all latent measurement constructs, the Average Variance Extracted (AVE) values exceeded the requisite 0.50 threshold, demonstrating that each latent variable explains more than half of the variance observed across its respective operational indicators.

Discriminant Validity

Discriminant validity assesses whether constructs within the nomological net are genuinely distinct from one another. This was confirmed using the classic Fornell and Larcker (1981) criterion: the square root of the AVE for every individual construct was strictly greater than the correlation coefficients between that construct and any other latent variable in the model. Additionally, cross-loading analyses confirmed that every measurement item loaded at its highest statistical magnitude on its intended theoretical construct, displaying distinctly lower cross-loadings across alternate constructs, thereby confirming structural discriminant boundaries.

Common Method Bias (CMB)

Because self-report survey methodologies carry the inherent risk of artificial covariation due to measurement format, common method bias was formally assessed using Harman’s single-factor test (Wang et al., 2018). An unrotated exploratory factor analysis revealed that the primary factor accounted for only 26.82% of the total variance. Because this empirical metric falls significantly below the recognized 50% critical threshold, common method variance was ruled out as a threat to construct and structural validity.

Reliability

The reliability of the instrument was examined using both traditional metrics of internal consistency and advanced structural composite indices:

• Cronbach’s Alpha ($lpha$): Widely recognized as a foundational benchmark of internal consistency reliability, Cronbach’s alpha values for all operational subscales comfortably surpassed the canonical threshold of 0.70, with all constructs registering values exceeding 0.80. This demonstrates strong inter-item correlation and conceptual harmony within each latent dimension.
• Composite Reliability (CR): In PLS-SEM applications, composite reliability provides a more nuanced measure of internal consistency because it does not assume equal factor loadings across all indicators. The composite reliability coefficients across the entire structural battery uniformly surpassed the stringent 0.80 threshold. This confirms robust internal consistency, indicating minimal measurement error across the measurement scales.

These statistical metrics demonstrate that the scale items reliably capture employee cognitive states with precision across organizational divisions.

Factor Analysis

The operational framework of the measurement model was validated using advanced latent factor analysis integrated into a two-stage structural equation modeling procedure:

Measurement Model Evaluation

Within the PLS-SEM paradigm, the reflective measurement model underwent empirical inspection to confirm outer item loadings. Standard psychometric convention dictates that standardized outer loadings should exceed 0.708, confirming that the shared variance between the latent construct and its indicator surpasses the variance attributed to measurement error. All retained items exhibited strong factor loadings that loaded cleanly onto their respective conceptual dimensions without problematic cross-loadings.

Structural Fit and Path Relationships

The latent variable structure confirmed the nomological paths dictated by Behavioral Reasoning Theory:

• The values dimension (Openness to Change) showed strong, positive predictive associations with the reasons for adoption (anthropomorphism, intelligence, interactivity, personalization), while exhibiting inverse predictive paths toward reasons against adoption (technological anxiety, perceived risk, language barrier).
• The “reasons for” constructs demonstrated significant positive predictive pathways toward Attitude toward Chatbots, which in turn functioned as a primary conduit to Adoption Intention.
• The “reasons against” dimensions exhibited statistically significant negative associations with employee attitudes, while also exerting direct negative suppression on Adoption Intention, confirming the dual-pathway mechanism central to BRT.
• Model fit indices confirmed the explanatory power and predictive relevance of the endogenous targets ($R^2$ and Stone-Geisser’s $Q^2$ values), proving the framework’s empirical adequacy.

Instrument / Measurement Tool

The Adoption of Artificial Intelligence-Based Employee Experience Chatbots instrument is a self-administered, structured psychometric questionnaire designed for corporate workforce evaluations.

• Test Type: Psychometric inventory / Multi-dimensional survey questionnaire.
• Target Population: Adult workforce members (aged 18 and older) employed within organizational settings utilizing or planning to deploy conversational AI interfaces, including IT, ITeS, e-commerce, and corporate enterprises.
• Administration Format: Computer-assisted web interview (CAWI), internal corporate portal survey, or paper-and-pencil standardized questionnaire.
• Response Scale: Standard 5-point Likert rating scale:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Neutral
  • 4 = Agree
  • 5 = Strongly Agree
• Core Dimensions Assessed:
  • Openness to Change (Personal Values)
  • Perceived Anthropomorphism (Reason For)
  • Perceived Intelligence (Reason For)
  • Interactivity (Reason For)
  • Personalization (Reason For)
  • Technological Anxiety (Reason Against)
  • Perceived Risk (Reason Against)
  • Language Barrier (Reason Against)
  • Attitude toward Chatbots
  • Adoption Intention
• Scoring and Diagnostic Computation: Subscale scores are derived by calculating the mean or sum of the items corresponding to each latent dimension. High scores on “reasons for” combined with high scores on “openness to change” indicate strong psychological readiness for AI adoption. Elevated scores across “technological anxiety”, “perceived risk”, or “language barrier” highlight critical psychological barriers requiring targeted workplace interventions.

Permissions & Fee and Test Year

The scale was developed and published in 2024. The empirical study and underlying measurement structure are published under the intellectual property guidelines of Emerald Publishing Limited in the journal Information Technology & People:

• Test Year: 2024
• Commercial Availability: Non-commercial. The measurement framework is an academic diagnostic instrument.
• Licensing and Fees: Academic and scholarly use of the construct framework is typically permitted without licensing fees under fair use principles, provided formal academic attribution is given to the original authors (Pillai et al., 2024). Commercial implementations, organizational consulting deployments, or proprietary software integrations should seek explicit written permission from the copyright holder and authors.
• Access Location: The original operational validation tables and structural results are detailed in the published source article: Information Technology & People, Vol. 37, No. 1, Table 5, Pages 461–463 (https://doi.org/10.1108/ITP-04-2022-0287).

References

• Claudy, M. C., Garcia, R., & O’Driscoll, A. (2015). Consumer resistance to innovation—a behavioral reasoning perspective. Journal of the Academy of Marketing Science, 43(4), 528–544. https://doi.org/10.1007/s11747-014-0399-0
• Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.1177/002224378101800104
• Guadagnoli, E., & Velicer, W. F. (1988). Relation of sample size to the stability of component patterns. Psychological Bulletin, 103(2), 265–275. https://doi.org/10.1037/0033-2909.103.2.265
• Gupta, A., & Arora, N. (2017). Consumer adoption of m-banking: A behavioral reasoning theory perspective. International Journal of Bank Marketing, 35(4), 733–747. https://doi.org/10.1108/IJBM-11-2016-0162
• Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2017). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (2nd ed.). SAGE Publications.
• Pillai, R., Ghanghorkar, Y., Sivathanu, B., Algharabat, R., & Rana, N. P. (2024). Adoption of artificial intelligence (AI) based employee experience (EEX) chatbots. Information Technology & People, 37(1), 449–478. https://doi.org/10.1108/ITP-04-2022-0287
• Wang, Y., Meister, D. B., & Gray, P. H. (2018). Social presence in enterprise social software: An empirical examination. Information Systems Journal, 28(5), 903–928. https://doi.org/10.1111/isj.12170
• Westaby, J. D. (2005). Behavioral reasoning theory: Identifying new linkages underlying intentions and behavior. Organizational Behavior and Human Decision Processes, 98(2), 97–120. https://doi.org/10.1016/j.obhdp.2005.07.003

Items of the Scale