Artificial Intelligence and New Quality Productive Forces: Evidence from Vietnam’s Banking Sector

Abstract

This study examines how artificial intelligence (AI) contributes to the formation of new quality productive forces (NQPF) at the employee level. While prior research has largely treated AI as an external technological driver, this study investigates whether AI becomes embedded within employees’ capabilities through confidence and skill transformation. Using survey data from 303 employees in Vietnamese commercial banks, the study applies exploratory factor analysis and regression models to analyze the relationships among AI confidence, skill transformation, work experience, and NQPF. The results show that AI confidence has a significant positive effect on NQPF, and this relationship is strengthened by skill transformation. However, work experience weakens this effect, suggesting uneven adaptation across employee groups. These findings indicate that the impact of AI on productive transformation depends not only on technological deployment but also on workforce capability development. The study contributes to the literature by providing micro-level evidence on how AI may be internalized within labor processes in emerging economies.

1. Introduction

Artificial intelligence (AI) has become a central site of contradiction in contemporary capitalism. On the one hand, AI enhances productivity and generates new forms of value creation. This transformation is consistent with broader arguments that digital technologies reshape value creation logics and enable innovation-driven development (Andriyansah & Saputra, 2025). On the other, it may displace workers, deepen stratification, and produce new forms of precarity. Classical labor process theory has long argued that technological change can intensify control and reshape skill hierarchies within production systems (Braverman, 1974). Recent research on human–AI interaction further documents how AI adoption alters workplace dynamics and performance outcomes (Glikson & Woolley, 2020). These contradictions resonate with broader debates on technological change and the restructuring of productive forces, in which innovations simultaneously expand the material basis of accumulation and destabilize established forms of labor control (Z. Li & Qi, 2014; Petach & Wilson, 2024).

The emergence of the concept of “new quality productive forces” (NQPF) provides a new vocabulary for capturing these transformations. Prominent in Chinese political discourse, NQPF emphasizes innovation-driven development that moves beyond reliance on cheap labor and resource-based advantages (Xie et al., 2025). From a Marxist perspective, productive forces are defined as the unity of instruments of labor and labor power (Marx, 1844; Marx & Engels, 1975, 1988). The notion of NQPF therefore signals a qualitative transformation of this unity, raising the theoretical question of whether AI should be regarded as a constitutive component of the productive forces rather than a peripheral adjunct.

Theoretical Tension: AI as an External Driver or a Constitutive Element of NQPF?

Although recent empirical studies have established a positive relationship between AI and NQPF (Tan et al., 2025; J. Li et al., 2025; Wu & Zhan, 2026), the theoretical interpretation of AI’s role within this relationship remains unsettled. Specifically, an unresolved tension persists between two fundamentally different understandings of how AI relates to productive forces.

In the dominant empirical approach, AI is operationalized as an external explanatory variable influencing NQPF. Tan et al. (2025) construct an enterprise-level NQPF index using text analysis and machine learning techniques and demonstrate that AI significantly enhances NQPF through technological innovation efficiency and resource allocation optimization. Similarly, J. Li et al. (2025) measure AI across strategic, application, and innovation dimensions and show that AI development promotes NQPF while generating spatial spillover effects. Wu and Zhan (2026) further identify NQPF as a driver of advanced manufacturing cluster upgrading, mediated by technological innovation and talent agglomeration. Across these studies, AI functions as a catalyst or enabling condition, whereas NQPF is treated as an outcome variable reflecting high-quality development. In this framework, AI improves productive performance but does not alter the internal structure of productive forces.

However, this interpretation becomes theoretically incomplete when examined through the structural lens proposed by Chin et al. (2025). By conceptualizing NQPF as comprising three interrelated components—laborers, labor resources, and labor objects—Chin et al. (2025) re-anchor NQPF in the classical definition of productive forces as the unity of labor power and means of production (Marx, 1844; Marx & Engels, 1988). If labor quality constitutes an intrinsic component of NQPF, then AI’s integration into workers’ skills, decision-making processes, and cognitive frameworks cannot be reduced to a mere external driver. Instead, AI may reshape the very structure of labor power itself. In such a case, AI would not simply promote NQPF; it would become constitutive of it.

Despite this growing body of literature, three important gaps remain. First, existing studies predominantly adopt a firm-level perspective, treating AI as an external technological input while overlooking how it is internalized within employees’ capabilities. Second, current empirical approaches focus on macro or organizational indicators of NQPF, providing limited insight into the micro-level mechanisms through which productive forces are transformed. Third, the role of employee-level factors—such as AI confidence and skill adaptation—in shaping this transformation remains underexplored.

As a result, the literature has yet to explain how AI becomes embedded within labor processes and whether such internalization alters the structure of productive forces rather than merely enhancing them.

Accordingly, this study addresses the following research question:

To what extent does AI, when internalized through employees’ confidence and skill transformation, function as an external driver versus a constitutive element of new quality productive forces?

This tension reflects two divergent ontological understandings of technology. In management and strategic studies, technology is often framed as a higher-order dynamic capability that enables firms to reconfigure resources under environmental turbulence (Teece, 2007; Tan et al., 2025). Within this view, AI enhances adaptability and efficiency but remains instrumental. In contrast, the classical theory of productive forces suggests that technological transformations can reorganize the material foundations of production, redefining the relationship between labor and means of production (Marx, 1844; Marx & Engels, 1988). Under this perspective, AI has the potential not merely to optimize processes but to reconstitute the structure of productive forces.

The emerging literature on AI confidence further intensifies this theoretical tension. Foundational trust theory defines trust as a willingness to accept vulnerability based on positive expectations (Rousseau et al., 1998), and subsequent work extends this logic to technological systems (Twyman et al., 2008). More recent studies on human–AI interaction demonstrate that AI-specific trust significantly shapes collaboration, reliance, and performance outcomes (Glikson & Woolley, 2020; Kong et al., 2023; Chong et al., 2022; Gkinko & Elbanna, 2023). These findings imply that AI becomes internalized through workers’ perceptions and behavioral engagement. Yet, this literature stops short of connecting such internalization to the structural transformation of productive forces. AI confidence is examined as a determinant of technology adoption rather than as a mechanism through which labor power itself is reconfigured.

Therefore, the core theoretical issue is no longer whether AI enhances NQPF—this relationship has been empirically documented—but rather how AI should be situated within the structure of NQPF. If AI is treated merely as an exogenous technological input, it belongs to the domain of instrumental facilitation. However, if AI is deeply embedded in workers’ capabilities, cognition, and skill formation, it may constitute a structural component of new quality productive forces.

This unresolved tension is particularly salient in service sectors such as banking, where AI systems directly participate in credit evaluation, fraud detection, and decision-making processes. In such contexts, AI does not operate solely as backend infrastructure but becomes intertwined with frontline labor. Consequently, the central research question of this study emerges: when AI is internalized into employees’ skills and confidence structures, does it remain an external driver of NQPF, or does it become a constitutive element of new quality productive forces? Furthermore, through what mechanisms does this transformation occur?

Yet scholarly perspectives remain divided. Political economy approaches have long argued that technological change can reshape both the technical organization of production and the social relations that structure it (Bowles, 1985; Shaikh & Tonak, 1994). Although these works do not address AI directly, they provide a theoretical lens for understanding how contemporary digital technologies may transform the material foundations of productive forces. In contrast, management studies tend to conceptualize AI as a dynamic capability that enables firms to reconfigure resources without itself constituting productive forces (Chin et al., 2025). This unresolved tension crystallizes the central puzzle of whether AI can become a constitutive element of productive force or remain merely as an external driver.

Clarifying this question is critical because it engages broader debates on the relationship between productive forces and relations of production, which have long been associated with issues of inequality and labor stratification (Roncolato & Willoughby, 2017; Perry, 2025). In contemporary contexts, AI intensifies these longstanding tensions, particularly in emerging economies undergoing rapid digital transformation. Vietnam provides a strategic vantage point for addressing this debate. As a rapidly developing economy committed to digital transformation, Vietnam illustrates how emerging economies confront technological change under rigid institutional constraints. The banking sector, which is both labor-intensive and tightly regulated, makes these contradictions especially visible: AI integration enhances efficiency while simultaneously producing new hierarchies of skill and trust, privileging digitally confident employees and marginalizing others.

This study makes a primary contribution by shifting the analysis of new quality productive forces from the firm level to the employee level. While existing studies predominantly conceptualize AI as an external driver of productivity at the organizational level, this study provides a micro-level empirical examination of how AI is internalized within labor processes. In addition, the study contributes by identifying AI confidence as a key mechanism through which this internalization occurs, linking individual-level cognition to the formation of productive forces. Finally, by focusing on Vietnam’s banking sector, the study offers context-specific insights into how AI-driven transformation unfolds in an emerging economy setting.

2. Theoretical Framework and Hypotheses Development

2.1. New Quality Productive Forces

2.1.1. Philosophical Roots of NQPF: Marx’s Theory of Productive Force

Karl Marx’s theory of productive forces provides the foundational lens for examining how technological change interacts with labor processes and institutional arrangements. In Marxist political economy, productive forces are defined as the unity of labor power and the means of production (Marx, 1844; Marx & Engels, 1988). Labor power refers to the physical and intellectual capacities of workers. The means of production comprise both the means of labor (tools, machinery, digital systems) and the objects of labor (the materials or entities upon which labor acts and which are transformed in the production process).

This distinction is crucial. Objects of labor do not refer to organizational objectives or strategic goals. Rather, they denote the substantive materials transformed through labor. In industrial production, these include land and raw materials. In contemporary service sectors such as banking, objects of labor are primarily intangible and include financial data, credit records, customer information, and risk profiles. These informational materials are processed, evaluated, and transformed into financial decisions and services.

Marx further emphasizes the contradiction between developing productive forces and existing relations of production. When innovative technologies expand productive capacity but institutional structures fail to adapt, structural tensions emerge (Marx & Engels, 1988). Later Marxist scholars elaborate this dynamic. Braverman (1974) shows how technological innovation can restructure skill hierarchies and labor control. Bowles (1985) and Shaikh and Tonak (1994) analyze how technological change reshapes production processes and surplus extraction. These works do not address AI directly, but they provide a theoretical foundation for situating contemporary digital transformation within the long-standing debates of political economy.

2.1.2. The Emergence of NQPF Concept

Building on this tradition, the concept of NQPF has gained prominence in recent policy and academic discourse (Xie et al., 2025; Chin et al., 2025). NQPF refers to a qualitative upgrading of productive forces driven by technological innovation, digital transformation, and more efficient utilization of resources. In both Chinese and Vietnamese contexts, the concept is associated with innovation-led development and structural transformation beyond low-cost, labor-intensive growth models.

Chin et al. (2025) propose a multidimensional framework for operationalizing NQPF in the AI development context, comprising three components: Laborers (quality of labor power), Labor resources (technological and organizational means of production), and Objects of labor (the materials being transformed through production).

While this framework provides an important structural foundation, it is primarily developed at the enterprise level, relying on firm-level indicators such as AI investment, innovation intensity, and strategic orientation. The micro-level mechanisms through which labor power itself is transformed remain less clearly specified.

In this study, we retain the tripartite structure proposed by Chin et al. (2025) but adapt it to the labor-process level within AI-enabled banking. Specifically, we clarify that the laborer dimension refers to capability enhancement, adaptability, and skill upgrading; labor resources refer to AI infrastructure, training systems, and organizational support that constitute the means of production; and objects of labor refer to financial and informational materials transformed in banking operations.

Table 1. NQPF conceptualization.

NQPF components	Adapted definition to the AI-enabled transformation context
Laborer	Enhancement of labor power through skill upgrading, adaptability, and AI-related competencies
Resources of Labor	AI infrastructure, digital systems, training programs, and organizational support that constitute the means of production
Objects of Labor	Financial data, customer information, credit profiles, and risk assessments that are transformed through AI-mediated labor processes

presents the descriptive statistics and correlation matrix of the key variables used in the analysis.

2.2. AI Confidence

To understand how AI fosters the emergence of new quality productive forces, it is crucial to examine the role of employee AI confidence. Confidence, or trust in competence, reflects a psychological state in which individuals are willing to rely on a system based on positive expectations about its competence (Rousseau et al., 1998; Twyman et al., 2008). While traditionally grounded in interpersonal relationships (Yuan et al., 2022), confidence has increasingly been explored in the context of human–AI interactions, where individuals must engage with non-human agents whose decision-making logic may be opaque, adaptive, and dynamic (Chowdhury et al., 2023; Kong et al., 2023; Chong et al., 2022). Beyond individual cognition, organizational communication, trust, and supportive culture play a critical role in facilitating effective human–AI interaction (Saputra et al., 2026).

From the perspective of Self-Determination Theory (Deci & Ryan, 1985; Ryan & Deci, 2000, 2023), AI confidence facilitates the satisfaction of employees’ psychological needs for autonomy, competence, and relatedness. Given the complex nature of AI behavior and potential risks associated with erroneous or biased outputs, confidence enables employees to interpret AI interventions as beneficial rather than threatening (Glikson & Woolley, 2020). When employees perceive AI as reliable, transparent, and capable of enhancing their performance, their cognitive AI confidence increases, leading to greater engagement with AI-enabled systems (Gkinko & Elbanna, 2023). This behavioral and cognitive shift not only enhances individual capability but also contributes to the organizational capacity to adopt and sustain technologically mediated innovation—central to the notion of NQPF (Chin et al., 2025). In the context of productive forces, this trust empowers employees to leverage AI tools in ways that amplify efficiency, knowledge integration, and problem-solving—hallmarks of NQPF, which emphasize innovation-driven, sustainable, and intelligent modes of production (Xie et al., 2025). Moreover, when employees trust AI to provide accurate insights or automate routine decisions, they redirect cognitive resources toward higher-value tasks such as innovation, strategic thinking, and collaboration.

In summary, AI confidence enables a psychological and functional alignment between human labor and machine intelligence, which is essential for cultivating the innovation-centric, sustainable, and adaptive characteristics of new quality productive forces. Employees’ confidence in AI strengthens their willingness to use it as a transformative tool, supporting the transition from traditional productivity logic to digital, high-quality development. From the perspective of productive forces, AI confidence does not merely influence technology usage but shapes how labor power incorporates technological capabilities. When employees trust and feel confident in AI systems, they are more likely to integrate these systems into their decision-making processes, effectively extending their cognitive and operational capacities.

In this sense, AI confidence functions as a mechanism through which technological resources are internalized within labor power. Rather than remaining external tools, AI systems become embedded in everyday work practices, contributing to the structural formation of productive forces. Consequently, AI confidence plays a critical role in transforming the relationship between labor and the means of production.

Thus, we posit the following hypothesis:

Hypothesis 1. 

AI confidence is positively associated with NQPF.

2.3. Experience

Work experience refers to the accumulated task, role, and industry tenure that shapes both tacit and explicit knowledge, organizational routines, and interpretive schemas (Cohen & Levinthal, 1990; Argote & Miron-Spektor, 2011). While experience enhances efficiency, accuracy, and pattern recognition, it can also lead to cognitive entrenchment, stable mental models that reduce openness to novel approaches (Dane, 2017). Work experience can play a dual role in AI-enabled environments. On the one hand, experience enhances absorptive capacity, domain knowledge, and contextual judgment, which may facilitate the effective use of AI systems. Experienced employees are often better able to interpret AI outputs and integrate them into complex decision-making processes.

On the other hand, as highlighted in the literature on cognitive entrenchment and path dependence, accumulated experience may also reinforce established routines and reduce openness to new technological approaches. This creates a tension in which experience can both enable and constrain the integration of AI into productive processes. Building on this dual perspective, this study focuses on the constraining effect of experience in the context of rapid AI-driven transformation.

There are two established arguments from organizational management scholars supporting that even when employees possess high AI confidence, extensive experience may weaken its translation into NQPF behaviors. Firstly, over time, experience consolidates routines and forms a dominant logic that guides decision-making but may bias individuals against adopting unfamiliar tools such as AI. Such path dependence can limit adaptability in fast-changing technological environments (Nelson & Winter, 1985). Entrenched schemas reduce the willingness to reconfigure work processes, hindering the integration of AI into value-creating activities (Prahalad & Bettis, 1986; Dane, 2017). Extended tenure often locks individuals into historical trajectories, raising switching costs and reinforcing exploitation of familiar solutions over exploration of new technological possibilities (Sydow et al., 2009; March, 1991).

Secondly, work experience can heighten resistance to change when core professional practices are threatened. Strongly ingrained professional identities may amplify threat–rigidity responses, curbing experimentation with AI-enabled methods (Petriglieri, 2011). Experienced employees often protect incumbent skills and role authority, perceiving AI-enabled process changes as challenges to their status. This defensive stance diminishes their willingness to embed AI in core workflows, even when they are confident in its capabilities.

In summary, our argument points to a paradox of expertise in the age of AI: while work experience provides a valuable foundation of firm-specific knowledge, it can also foster entrenched cognition and identity protection that impede the very adaptability required to convert AI confidence into innovation-driven productive forces. Based on this, we propose Hypothesis 2:

Hypothesis 2. 

Work experience negatively moderates the AI confidence–NQPF relationship.

2.4. Skill Change

Skill change refers to the process by which individuals and organizations acquire and adapt competencies to meet emerging demands. It encompasses the concept of change competence, defined as the dual capacity to leverage existing skills while consistently building new ones to address evolving challenges (Reineholm et al., 2024). Skill change varies depending on operational or strategic contexts, underscoring its nuanced and multi-dimensional character.

The positive relationship between skill changes and NQPF is supported by several arguments. Firstly, NQPF is characterized by high performance and continuous innovation, necessitating a workforce that is both adaptable and committed to ongoing skills upgrading (Chin et al., 2025; Xie et al., 2025). Facilitating skill change empowers organizations to effectively implement technological advances and transition to modern production models, which are core requirements of NQPF (McGuinness et al., 2023). Empirical research further dispels concerns of widespread deskilling, instead finding that technology frequently prompts the rebuilding and enhancement of workforce capabilities (Burchell et al., 1994). Additionally, studies on skill-biased technological change highlight a positive complementarity between new technologies and the demand for advanced skills, with clear evidence that organizations investing in skill development realize gains in both productivity and innovation (McGuinness et al., 2023). Systematic investment in skill change thereby strengthens knowledge integration and operational efficiency, driving the sustainable, innovation-focused growth central to NQPF (Kolade & Owoseni, 2022).

Hence, we posit the following hypothesis:

Hypothesis 3. 

Skill change is positively related to NQPF.

Taken together, the proposed hypotheses form a unified framework explaining how AI becomes embedded within the structure of productive forces at the employee level. AI confidence represents the cognitive mechanism through which technological resources are internalized into labor power, while skill transformation captures the capability dimension of this integration. Work experience, in turn, conditions this process by either facilitating or constraining the extent to which AI can be effectively incorporated into productive activities.

Rather than examining isolated relationships, the model tests an integrated process in which technological, psychological, and experiential factors jointly shape the formation of new quality productive forces.

3. Research Methodology

3.1. Research Design

This study employs a quantitative survey analysis to examine how AI is internalized as a constitutive element of NQPF. The quantitative component provides the structural foundation of the analysis, cross-sectional survey design to examine whether AI can be conceptualized as a constitutive element of NQPF in Vietnam’s banking sector. Rather than treating AI as an external influence, the research investigates its role as part of the latent construct of productive forces. The analytical framework combines reliability testing, factor-based validation, and regression analysis. The analytical process consisted of three steps: (i) assessing the reliability of the measurement scales; (ii) conducting Exploratory Factor Analysis (EFA) and Principal Component Analysis (PCA) to identify and measure NQPF; and (iii) performing OLS regression analysis to test the three stated hypotheses.

3.2. Sample Size and Recruitment

This study employed a purposive sampling strategy to collect survey data from employees working in commercial banks in Vietnam. The banking sector was selected due to its rapid integration of AI technologies in credit assessment, fraud detection, customer service automation, and internal operational systems. Given the increasing digitalization of banking services, employees in this sector provide an appropriate context for examining the relationship between AI integration and the formation of new quality productive forces.

3.2.1. Selection Criteria

Respondents were required to meet the following criteria:

(1)

they must be full-time employees working in a commercial bank in Vietnam;

(2)

they must have direct or indirect exposure to AI-enabled systems or digital automation tools in their daily work;

(3)

they must have at least six months of work experience to ensure sufficient familiarity with organizational processes and AI applications.

These criteria ensured that participants had meaningful exposure to AI technologies and were able to evaluate their confidence, skill transformation, and perceptions of organizational change.

3.2.2. Survey Procedure

Data were collected through a structured questionnaire administered primarily via an online survey platform. The online format was chosen to facilitate broad geographic coverage and ensure respondent anonymity. Banking employees are accustomed to digital systems, making online distribution appropriate and efficient.

The questionnaire included sections measuring AI access and confidence, technological infrastructure, labor skill transformation, training and adaptation, organizational culture and support, job security and satisfaction, and future perceptions regarding AI. Most variables were measured using a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). In addition, several open-ended questions were included to capture qualitative reflections. An introductory statement explained the academic purpose of the study, guaranteed confidentiality, and emphasized voluntary participation.

3.2.3. Data Collection Period and Sample Size

Data was collected between June 2025 and September 2025. The survey covered employees from multiple commercial banks located in major economic regions of Vietnam, including Hanoi and Ho Chi Minh City. A total of 303 responses were obtained and included in the final dataset. The characteristics of the sample are summarized in

Table 2. Demographic and Occupational Composition.

Demographic Construct	Categories	Frequency	Percentage (%)
Age group	25–34	118	38.9
	35–44	90	29.7
	45–54	35	11.6
	55+	2	0.7
Work experience	<1 year	65	21.5
	1–3 years	50	16.5
	4–6 years	32	10.6
	>6 years	156	51.5
Bank Type	State-owned bank	85	28.1
	Joint-stock commercial bank	205	67.7
	Foreign bank	1	0.3
	Others	12	4.0
Job position	Customer relationship officer	137	45.2
	Bank teller	28	9.2
	Branch manager	20	6.6
	Technology specialist	7	2.3
	Others	111	36.6
Total		303	100

. Table 2 presents the demographic and occupational composition of the respondents, including gender, age, experience, and managerial roles.

A pilot test was conducted with a small group of banking employees (n = 10) to evaluate the clarity, wording, and contextual relevance of the questionnaire. Participants were asked to provide feedback on item comprehension and applicability to their work context. Based on this feedback, minor revisions were made to improve wording clarity and ensure that the items were appropriately adapted to the Vietnamese banking context.

The questionnaire was originally developed in English and translated into Vietnamese using a translation and back-translation procedure to ensure semantic equivalence.

3.2.4. Data Cleaning and Screening Procedures

All respondents confirm direct experience with AI-related tools or systems in their work. Participation was voluntary, anonymous, and no personal identifiers are collected. The structured questionnaire is administered online in Vietnamese, using closed-ended Likert-scale items to ensure comparability and quantitative rigor. Although the questionnaire included several open-ended questions, these responses were used primarily for contextual understanding and were not systematically analyzed in this study.

Data cleaning was initially conducted using Microsoft Excel. Data cleaning procedures were actively applied prior to analysis. Responses were screened for completeness, duplicate submissions (based on timestamps), and inconsistent answering patterns. All 303 responses were retained in the final dataset because no invalid cases were identified after the screening process.

Statistical analysis was conducted using Stata 17. Exploratory Factor Analysis (EFA) was first applied to examine the underlying structure of the measurement items. Principal Component Analysis (PCA) was then used to construct composite indices for key constructs. Finally, Ordinary Least Squares (OLS) regression models were estimated to test the proposed hypotheses. Robust standard errors were applied where appropriate to address potential heteroskedasticity.

3.3. Variable Measurement

All constructs use multi-item Likert scales (1 = strongly disagree, 5 = strongly agree). Reliability tests confirm internal consistency, with Cronbach’s alpha values exceeding 0.83. The dependent variable, NQPF, is operationalized in line with Marxist political economy as the unity of labor power and instruments of labor, reflected in three dimensions: (i) human capability enhancement, (ii) supportive organizational culture, and (iii) future-oriented cognition. Independent variables include AI confidence, work experience, and skill change, while demographic and infrastructural factors serve as controls. The variables used in this study are defined and operationalized as shown in

Table 3. List of variables and definitions.

Variable Name	Definition
ai_use	Frequency of using AI tools
ai_conf	Self-confidence in using AI
ai_support	AI support in decision/tasks
inf_net	Network quality
inf_system	System stability
inf_error	Error frequency
inf_support	Technical support availability
inf_modern	Modernity of infrastructure
sk_change	Job transformation due to AI
sk_adapt	Adaptability to AI
sk_data	Data/digital skills
sk_auto	Ability to handle automation
sk_creat	Creativity in digital tasks
at_pos	Feeling threatened by AI
at_fear	Fear of AI displacement
at_mot	Motivation to learn
at_trust	AI Confidence and recognition
trn_form	Formal training programs
trn_self	Self-learning
trn_onjob	On-the-job learning
trn_eff	Training effectiveness
oc_collab	Collaboration culture
oc_innov	Innovation encouragement
oc_support	Internal communication about AI
fut_trust	Optimism about AI’s future role
fut_conf	Confidence in AI creating more opportunities
fut_opp	Willingness to recommend others to work in sector

.

3.4. Measurement Design

To operationalize the dependent variable, this study adopts a two-stage approach combining Exploratory Factor Analysis (EFA) and Principal Component Analysis (PCA). First, EFA is employed to examine the underlying structure of the observed variables and to identify how the dimensions of NQPF manifest at the employee level in AI-enabled work contexts.

Based on the EFA results, only items with factor loadings greater than or equal to 0.50 on the dominant factor are retained for subsequent analysis. These indicators capture closely related empirical dimensions of NQPF, including adaptability, AI engagement, training effectiveness, and future orientation. Together, they reflect how employees adopt, interact with, and integrate AI into their work processes, as well as how they develop capabilities and behavioral orientations in the context of digital transformation.

In the second stage, PCA is applied to construct a composite index of NQPF. Specifically, the NQPF score (nqpf) is computed as the first principal component (Comp1), representing a weighted linear combination of the retained indicators, where the weights are derived from the corresponding eigenvectors. This approach ensures that each indicator contributes to the composite index according to its relative importance in explaining shared variance, rather than being equally weighted. The first principal component accounts for 67.35% of the total variance, indicating a high degree of convergence among the measured dimensions.

From a theoretical perspective, these empirical dimensions can be mapped onto the tripartite structure of NQPF. Specifically, adaptability and AI engagement reflect the enhancement of labor power (laborers), while training-related and organizational support elements represent the enabling conditions of production (labor resources). Regarding the object-of-labor dimension, in the context of AI-driven digital transformation, data—together with AI-enabled banking products and services—constitutes the central content of the labor process. In this study, this dimension is captured through future-oriented variables, which indicate the extent to which employees adapt to interact with data and engage in the processing of AI-supported banking products and services, thereby reflecting how these elements become central to the object of labor.

Accordingly, the nqpf index should be interpreted as an empirically integrated representation of interrelated components of NQPF at the micro level. This approach captures the convergence of key elements of new quality productive forces in AI-enabled contexts, while maintaining methodological rigor and empirical tractability.

For the regression analyses, the focal independent variable, ai_conf, is conceptualized as a perceptual and psychological construct reflecting employees’ confidence in their ability to work with AI systems. It represents a unidimensional latent construct capturing perceived competence in engaging with AI in the workplace. The observed items are treated as reflective indicators of this underlying construct, representing different manifestations of employees’ confidence rather than distinct conceptual components. Accordingly, ai_conf reflects perceived confidence in AI use rather than actual productivity outcomes or structural capability development. The high internal consistency of the scale (Cronbach’s α = 0.88) further supports its treatment as a coherent and unidimensional construct. Tenure in banking (exp_group) is self-reported and categorized into standard analytical groupings. To test the moderation hypothesis (H2), an interaction term (ai_conf × exp_group) is computed by multiplying the mean-centered variables. Age group is included as a control variable.

While the theoretical framework conceptualizes NQPF as a tripartite structure comprising laborers, labor resources, and objects of labor, the empirically derived factor primarily captures the labor and organizational dimensions at the employee level. As such, the resulting composite should be interpreted as a partial, micro-level operationalization of NQPF rather than a complete representation of its full theoretical structure.

3.5. Ethics Considerations

This study was conducted in accordance with standard ethical principles for social science research involving human participants. Participation in the survey was entirely voluntary. Prior to completing the questionnaire, respondents were informed about the academic purpose of the study, the anonymity of their responses, and their right to withdraw at any time without any consequences.

No personally identifiable information was collected or used in the analysis. Although the online survey platform recorded email addresses automatically, these identifiers were not included in the analytical dataset and were removed during the data cleaning process. All responses were treated confidentially and were used solely for academic research purposes.

The questionnaire did not involve sensitive personal questions, medical information, or experimental manipulation. The research focused on employees’ perceptions of AI integration, skill transformation, and organizational adaptation in the workplace. As such, the study posed minimal risk to participants.

Data were securely stored and accessed only by the research team. The study complied with institutional research integrity guidelines and general ethical standards for survey-based research.

This study employs a cross-sectional design, which limits the ability to establish causal relationships. While the proposed framework assumes that AI confidence and skill transformation influence NQPF, reverse causality cannot be fully ruled out, as employees with higher perceived productive capability may also report greater confidence in AI.

To assess the robustness of the results, alternative model specifications were estimated. Model 1 presents the baseline regression, while Model 2 introduces the interaction term between AI confidence and work experience. The results remain consistent in terms of coefficient signs and statistical significance across specifications, supporting the stability of the findings.

In addition, multicollinearity diagnostics were conducted using variance inflation factors (VIF). All VIF values are within acceptable thresholds, indicating that multicollinearity is not a concern in the estimated models.

4. Findings

4.1. Measurement Scale Reliability

The internal consistency of the measurement scales is assessed using Cronbach’s alpha. All scales achieve α values ranging from 0.839 to 0.948, exceeding the acceptable threshold of 0.70 (Ahmed et al., 2025), which indicates high internal consistency and construct validity. The reliability of the measurement scales was assessed using Cronbach’s alpha, as reported in

Table 4. Reliability of the group of related variables.

Group of Related Variables	Items	Cronbach’s α
AI-related variables (ai_use, ai_conf, ai_support)	3	0.880
Technological Infrastructure (inf_net, inf_system, inf_error, inf_support, inf_modern)	5	0.839
Innovation-Supportive Culture (oc_collab, oc_innov, oc_support)	3	0.948
Future Orientation (fut_trust, fut_conf, fut_opp)	3	0.931
Skill Adaptation (sk_change, sk_adapt, sk_data, sk_auto, sk_creat)	5	0.885
AI Training (trn_form, trn_self, trn_onjob, trn_eff)	4	0.938

.

4.2. EFA and PCA

4.2.1. Exploratory Factor Analysis (EFA)

EFA was conducted using the Principal Factors extraction method with Promax rotation. The Kaiser–Meyer–Olkin (KMO) measure was 0.935 and Bartlett’s Test of Sphericity was significant (p < 0.001), confirming that the data were suitable for factor analysis.

First, the laborer dimension is reflected in human capability enhancement, including continuous skill adaptation and targeted training. Variables related to skill development and learning (e.g., sk_data, sk_auto, trn_form, trn_self) capture the upgrading of labor power as employees acquire the competencies required to work with AI systems.

Second, the labor resources dimension is represented by organizational support structures, including AI-related training systems, collaborative practices, and innovation-supportive culture. Indicators such as oc_collab, oc_innov, oc_support, and training effectiveness reflect the availability and usability of technological and organizational means of production.

Third, the object-of-labor dimension, while less directly observable in survey-based data, is partially captured through employees’ future-oriented cognition. Variables such as fut_trust, fut_conf, and fut_opp reflect how employees perceive and engage with data, information, and AI-mediated decision contexts, which constitute the transformed objects of labor in service-based environments such as banking.

Taken together, these dimensions represent the micro-level manifestation of the unity between labor power, technological resources, and transformed work objects, consistent with the theoretical structure of NQPF.

Exploratory factor analysis was conducted to examine the underlying structure of the constructs. The results, presented in

Table 5. EFA Results (Principal Factors, Promax Rotation).

Observed Variable	Factor 1	Factor 2	Uniqueness
inf_net	0.558	0.208	0.498
inf_system	0.544	0.245	0.475
inf_support	0.540	0.234	0.493
sk_data	0.575	0.334	0.314
sk_auto	0.536	0.342	0.363
at_fear	0.754	0.127	0.294
at_mot	0.696	0.148	0.364
at_trust	0.613	0.286	0.321
trn_form	0.735	0.119	0.334
trn_self	0.724	0.177	0.281
trn_onjob	0.847	0.048	0.228
trn_eff	0.878	0.019	0.208
oc_collab	0.888	−0.053	0.268
oc_innov	0.930	−0.094	0.236
oc_support	0.953	−0.100	0.203
fut_trust	0.900	−0.088	0.282
fut_conf	0.816	−0.036	0.370
fut_opp	0.850	−0.044	0.323

, show that the observed variables load strongly on a dominant factor. As presented in Table 5, two factors with eigenvalues greater than one were extracted from the EFA. Factor 1 demonstrates high loadings on a coherent set of variables encompassing AI-related training, skill adaptation, organizational culture, and future orientation. The training dimension, reflected in trn_form, trn_self, trn_onjob, and trn_eff, captures both formal and informal learning processes—whether structured, self-directed, or on-the-job—that enhance the adaptability of the workforce. Skill adaptation is indicated by variables such as sk_data, sk_auto, at_mot, at_trust, and the reverse-coded at_fear, which together represent the readiness of individuals to reconfigure their skills and embrace new work processes in the context of AI integration. The organizational culture component, as measured by oc_collab, oc_innov, and oc_support, reflects the presence of collective norms, collaborative practices, and supportive environments that foster innovation and experimentation. Finally, the future orientation dimension, captured through fut_trust, fut_conf, and fut_opp, encapsulates employees’ cognitive and affective stance toward AI-driven transformation, including their trust in technological outcomes, confidence in their own capacity to adapt, and perception of emerging opportunities. The co-loading of these conceptually distinct but mutually reinforcing elements onto a single factor indicates that they share substantial common variance. This suggests that, at the micro level, the dimensions of labor capability, organizational support, and future-oriented cognition operate as an integrated system of productive capacity. Rather than representing a reduction of the theoretical structure, Factor 1 reflects the empirical convergence of these dimensions in the context of AI-enabled work.

Consequently, the identification of a dominant factor through EFA supports the treatment of NQPF as a unidimensional latent construct at the empirical level, while retaining its multidimensional conceptual foundation.

Factor 1 can be interpreted as an empirical representation of key dimensions associated with NQPF, particularly those related to labor capability and organizational support. However, it does not fully capture the object-of-labor dimension specified in the theoretical framework.

4.2.2. Principal Component Analysis (PCA)

Following the EFA results, principal component analysis was conducted to construct a composite index. As shown in

Table 6. PCA Results.

Component	Eigenvalue	% Variance	Cumulative %
Comp1	12.1235	67.35	67.35
Comp2	1.0480	5.82	73.18
Comp3	0.7002	3.89	77.07

, the first principal component explains a substantial proportion of the variance and is used as the NQPF measure in subsequent analyses. Only those observed variables that exhibited rotated loadings of at least 0.50 on Factor 1 were retained for the construction of the composite NQPF measure through PCA. This threshold is consistent with commonly accepted standards in factor analysis, where loadings of 0.50 or higher are considered practically significant in representing a latent construct. Variables with lower loadings contribute less to the shared variance and may introduce measurement noise, thereby weakening construct validity.

By retaining only high-loading indicators, the PCA operates on a set of variables that are both empirically and theoretically aligned with the definition of NQPF as an integrated manifestation of skill adaptability, organizational support, and forward-oriented cognition in the context of AI adoption.

The PCA results further indicate that the retained indicators converge into a dominant component capturing the majority of shared variance among these dimensions (eigenvalue = 12.1235; explained variance = 67.35%). This dominant component reflects an empirically integrated latent dimension in which human capability enhancement, organizational support, and future-oriented cognition are closely intertwined.

Rather than implying that these elements are theoretically indistinguishable, the first principal component represents their empirical convergence within the specific context of AI-enabled work. This convergence is consistent with the Marxian notion of the unity of productive forces, in which labor power and the means of production are internally related and co-evolve in practice.

Accordingly, the composite score derived from this component (nqpf) is used as a parsimonious representation of this integrated capability structure in the subsequent analysis.

4.3. Hypothesis Testing

To test the direct effects (Research Question 2, Hypothesis H1 and H3, Ordinary Least Squares (OLS) regression was employed using the specification:

$$nqp{f}_i= {\beta }_0+ {\beta }_1\ast ai\text{\_}con{f}_i+{\beta }_2\ast \exp \text{\_}grou{p}_i+{\beta }_3\ast sk\text{\_}chang{e}_i + X_i+ {\epsilon }_{it}$$

where $nqp{f}_i$ represents the composite NQPF score for respondent i, $ai\text{\_}conf$ denotes ai confidence, $\exp \text{\_}grou{p}_i$ refers to work experience group, $sk\text{\_}chang{e}_i$ refers to job transformation due to AI and $X_i$ refers to vector of control variables. All independent variables were mean centered to facilitate subsequent moderation analysis.

To evaluate Hypothesis H2, an interaction term was computed between AI confidence and work experience, yielding the model:

$$nqp{f}_i= {\beta }_0+ {\beta }_1\ast ai\text{\_}con{f}_i+{\beta }_2\ast \exp \text{\_}grou{p}_i+ {\beta }_3\ast ai\text{\_}con{f}_i\ast \exp \text{\_}grou{p}_i+{\beta }_4\ast sk\text{\_}chang{e}_i +X_i+ {\epsilon }_{it}$$

The conceptual research model is presented in Figure 1, illustrating the proposed relationships among AI confidence, skill change, work experience, and NQPF, including the moderating role of experience.

The regression results are presented in

Table 7. Regression Results.

	NQPF (Model 1)	NQPF (Model 2)
ai_conf	1.6887 *** (0.210)	3.070 *** (0.439)
exp_group	0.353 * (0.165)	1.556 ** (0.507)
sk_change	0.893 *** (0.199)	1.171 *** (0.190)
inf_error	−0.212 (0.153)	−0.059 (0.151)
age_group	−0.362 (0.218)	−0.323 (0.220)
ai_conf × exp_group		−0.333 (0.129) **
Constant	−15.517 *** (1.875)	−15.434 *** (1.625)

Note: N = 303; * p < 0.05; ** p < 0.01; *** p < 0.001.

. AI confidence is positively associated with NQPF, supporting Hypothesis 1. The interaction term between AI confidence and work experience is negative and significant, indicating a moderating effect and supporting Hypothesis 2. Regression analysis builds on the factor results by showing how the latent construct of NQPF is stratified across the workforce. AI self-confidence has a large positive effect on NQPF (β = 3.070, p < 0.001), and skill change also contributes significantly (β = 1.171, p < 0.001). These findings indicate that while infrastructure and training constitute the instruments of labor, the extent to which workers can mobilize them depends on their confidence and adaptability.

Work experience alone has a positive association with NQPF (β = 1.556, p = 0.002), but the interaction between AI self-confidence and experience is negative and significant (β = −0.333, p = 0.010). This result indicates that the positive association between AI confidence and NQPF becomes weaker as work experience increases. In other words, AI confidence has a stronger effect on NQPF among less experienced employees, while this effect diminishes for those with higher levels of experience. This result highlights a structural inertia: employees with long tenure benefit from accumulated routines, but these very routines reduce the marginal returns of AI proficiency. In contrast, younger, or less experienced, employees translate AI confidence into larger productivity gains, positioning themselves as the vanguard of new productive forces.

To further illustrate the moderating effect of work experience, Figure 2 presents the interaction between AI confidence and NQPF across different experience groups.

As shown in Figure 2, the slope of AI confidence is steeper for employees with lower levels of experience, indicating a stronger positive effect on NQPF. In contrast, for more experienced employees, the slope becomes flatter, suggesting that accumulated experience reduces the marginal impact of AI confidence. This pattern is consistent with the negative interaction effect observed in the regression results.

Both models were estimated using OLS with a sample of 303 respondents.

The model explains a substantial proportion of variance in NQPF (R² = 0.554), indicating strong explanatory power. Comparing Model 1 and Model 2, the inclusion of the interaction term improves model fit, suggesting that the moderating role of work experience contributes additional explanatory value.

In terms of substantive effects, AI confidence and skill change exhibit the largest coefficients, indicating that these variables play a more prominent role in explaining variations in NQPF compared to other predictors. Variance Inflation Factor (VIF) values were below 2.5 across all predictors, indicating the absence of problematic multicollinearity.

The regression results indicate that AI confidence has a positive and statistically significant effect on NQPF. Work experience is also positively associated with NQPF. However, the interaction term between AI confidence and work experience is negative and significant, suggesting that the positive effect of AI confidence on NQPF decreases as work experience increases. This pattern indicates that AI confidence has a stronger association with NQPF among less experienced employees, while the effect becomes weaker for those with higher levels of experience.

5. Discussion

The empirical results provide clear support for the proposed hypotheses. Specifically, H1 is supported, as AI confidence shows a positive and statistically significant association with NQPF. H2 is also supported, with skill change exhibiting a strong positive effect on NQPF. Finally, H3 is supported, as the interaction term between AI confidence and work experience is negative and significant, indicating that work experience weakens the positive relationship between AI confidence and NQPF.

5.1. Theoretical Implications

The findings of this study provide micro-level evidence that both complements and refines the existing literature on AI and NQPF.

First, the positive and significant effect of AI confidence on NQPF is broadly consistent with prior enterprise-level studies demonstrating that AI promotes high-quality development (Tan et al., 2025; J. Li et al., 2025). However, while earlier research predominantly operationalizes AI as an exogenous technological driver influencing organizational-level NQPF indices, the present study shifts the analytical focus to the labor level. The results suggest that the transformation associated with AI does not originate solely from technological deployment, but from the extent to which AI is internalized through employees’ confidence and perceived competence. In this sense, AI does not automatically function as a productive force; it becomes one only when embedded in workers’ cognitive and behavioral capabilities. This micro-foundational perspective deepens the existing AI–NQPF framework by highlighting the subjective dimension of productive transformation.

Second, the role of skill transformation directly engages with the structural framework proposed by Chin et al. (2025), who conceptualize NQPF as comprising laborers, labor resources, and labor objects. The empirical evidence presented here substantiates the centrality of the labor component by demonstrating that AI-driven productive enhancement is contingent upon skill upgrading. AI infrastructure alone does not generate qualitative transformation; rather, productive forces are reshaped through the co-evolution of technological systems and human capabilities. In this regard, the study not only supports the tripartite structure of NQPF but also clarifies the mechanism through which the “laborer” component evolves under AI integration. This finding aligns with broader perspectives that digital transformation reshapes value creation mechanisms and supports innovation-led upgrading (Andriyansah & Saputra, 2025).

Third, the negative moderating effect of accumulated work experience introduces an important distributional dimension that has received limited attention in prior AI–NQPF research. Conventional human capital theory often treats experience as a cumulative source of productive advantage. Yet the findings suggest that in AI-intensive environments, experience does not uniformly enhance the productivity gains associated with AI confidence. Instead, adaptability and digital competence appear to become relatively more salient than tenure-based expertise. This implies that AI integration may reconfigure internal hierarchies of productive capacity within organizations. Rather than uniformly elevating all employees, AI reshapes the distribution of advantage across groups, favoring those better positioned to adapt technologically.

The empirical results of the EFA are broadly consistent with the theoretical conceptualization of NQPF, while also revealing how its dimensions manifest in an integrated manner at the employee level. Drawing on the conceptual framework presented earlier, NQPF is understood as the transformation of productive forces through the integration of three interrelated components: laborers, labor resources, and objects of labor. At the micro level, these components are operationalized through empirically observable dimensions that capture how AI is embedded in everyday work processes.

Compared to the literature on AI confidence and human–AI interaction (Glikson & Woolley, 2020; Kong et al., 2023; Chong et al., 2022), which primarily examines performance outcomes and technology acceptance, this study extends the analytical scope by linking AI confidence to the structural formation of productive forces. Confidence is not merely a predictor of usage behavior; it becomes a mechanism through which labor power is reconstituted in AI-enabled contexts.

Taken together, these findings contribute to resolving the theoretical tension outlined in the Introduction: whether AI functions merely as an external technological driver of NQPF or as a constitutive component of productive forces. The evidence suggests that AI can transition from an instrumental input to a structural element when it is deeply integrated into employees’ skills and confidence structures. However, this transformation is uneven and shaped by the internal distribution of adaptability and experience within the workforce.

5.2. Practical Implications

Building on the empirical findings, several practical implications can be derived. Organizations should pay particular attention to developing AI-related confidence among employees, especially those at earlier stages of their careers. Training programs and organizational support mechanisms should be designed not only to enhance technical skills but also to build confidence in interacting with AI systems.

Moreover, the moderating role of work experience suggests that experienced employees may require different forms of intervention, such as targeted reskilling or mindset-oriented training, to overcome potential resistance or inertia. This highlights the importance of differentiated workforce strategies rather than a one-size-fits-all approach to AI adoption.

At the policy level, workforce development initiatives should consider heterogeneity across experience groups, ensuring that AI-related transformation does not lead to uneven capability development within organizations.

5.3. Limitations and Future Research Directions

This study has limitations that should be considered when interpreting the findings. First, the focus on a single sector (banking) may limit the generalizability of the results, as the nature of AI adoption and productive processes may differ across industries. Therefore, the observed relationships may be context-specific rather than universally applicable.

Second, the findings are grounded in employees’ perceptions of AI-related capabilities and work processes. While this perspective is particularly valuable for capturing how AI is experienced and internalized at the micro level, it may not fully reflect objective productivity outcomes. Future research could extend this approach by incorporating additional sources of evidence, including performance-based indicators and multi-level data, to provide a more comprehensive understanding of AI-enabled productivity.

Third, this study reflects a snapshot of AI-enabled transformation at a particular point in time. As the relationships among AI adoption, skill transformation, and productive outcomes are likely to evolve, future research could further explore these dynamics through longitudinal or comparative designs, thereby offering deeper insights into how AI contributes to the transformation of productive forces over time.

These findings should be interpreted with caution, as the empirical measure of NQPF reflects a perception-based, micro-level proxy and does not fully encompass the structural dimensions of productive forces, particularly the object-of-labor component.

6. Conclusions

The findings suggest that AI may function as a constitutive element in the development of productive forces, as reflected in its strong association with NQPF at the employee level. The empirical findings provide several practical implications. First, the positive association between AI confidence and NQPF suggests that organizations should prioritize initiatives that enhance employees’ confidence in interacting with AI systems, including targeted training and supportive learning environments. Second, the strong effect of skill change highlights the importance of continuous reskilling and upskilling programs to ensure that employees can effectively adapt to AI-enabled work processes. Third, the negative moderating effect of work experience indicates that a uniform approach to workforce development may be insufficient. Instead, organizations should adopt differentiated strategies, with greater emphasis on supporting less experienced employees in leveraging AI, while addressing potential inertia or resistance among more experienced staff. At a broader level, these findings also imply that investments in AI infrastructure and governance mechanisms should be complemented by human-centered capability development to ensure balanced and inclusive productivity transformation. These findings highlight the importance of trust, internal communication, and organizational alignment in enabling AI-driven transformation (Saputra et al., 2026). Practically, banks and policymakers in emerging economies should institutionalize continuous reskilling, invest in reliable digital infrastructure, build transparent AI governance to foster trust, and tailor support for experienced employees to enable inclusive, sustainable digital transformation.