Digital upskilling in the gig economy: a study on Gen Z’s adoption of e-learning using UTUAT model

The global labor market is undergoing a structural transformation driven by the rapid diffusion of digital technologies such as the Internet of Things (IoT), artificial intelligence (AI), big data analytics and the industrial IoT (IIoT) (Kar et al., 2021). These technologies blur the boundaries between physical and digital workspaces, giving rise to new forms of digital tasks, platform-mediated work and gig-based employment (Abdelli and Abid, 2025). Although these developments have led to more economic opportunities, they have also increased the volatility of skills, the professional obsolescence of specific jobs and disparities in access to learning opportunities, which are raising critical issues of learning equity and digital inclusion (Allioui and Mourdi, 2023). The gig economy has not only become a form of labor market but also an informal ecosystem of learning, where workers must revitalize, renew and reuse digital skills based on how algorithmic management, platform needs and in comparison with traditional labor, gig employment transfers skill development to individuals, with access to digital upskilling becoming one of the most important factors of inclusion in digital markets (Korobkova et al., 2025).

The problem is especially relevant to Generation Z (born between 1997 and 2012), which is the first generation to have completely entered the labor market with thorough immersion in digital technologies. Often described as digital natives, Gen Z proves to be highly familiar with digital environments and highly desiring of autonomy, flexibility and self-directed careers. Consequently, a large proportion of Gen Z employees are becoming more attracted to freelance and gig-based jobs as their main or additional revenue model (Octobre and Raillard, 2020). Nevertheless, the gig economy requires constant digital updating, with workers being compelled to leverage the skills they possess and discover new technologies at the same time to keep up with competition, or in this case, the concept is referred to as individual ambidexterity (Seemiller and Grace, 2017).

Although digital familiarity exists, there is new evidence to indicate that digital familiarity does not necessarily result in fair distribution of task-specific and advanced digital literacy. There is still a substantial gap in Gen Z’s capacity to use more complicated technologies like AI-based tools, data analysis platforms and work systems controlled by algorithms (Salam et al., 2024). Digital upskilling, therefore, has turned into an inclusionary as well as exclusionary process in the gig labor market. The considerations that influence the use of digital upskilling by Gen Z are thus critical toward the development of inclusive digital systems and equitable learning experiences (Adetunla and Chowdhury, 2025).

With the growing role of digital technologies in mediating work and learning, it is now possible to consider how people accept, adopt and maintain engagement with digital technologies as a core focus of education and inclusion research. There have been various theoretical frameworks formulated to explain technology acceptance behavior, such as the “Technology Acceptance Model” (TAM) (Davis, 1989), the “Theory of Planned Behavior” (TPB) (Ajzen, 1991) and the “Theory of Diffusion” (commonly termed as the innovation diffusion theory) (Rogers et al., 2019).

The focus of these frameworks is on the cognitive evaluations, social norms and innovation characteristics as the driving force of adoption. Based on these views, Venkatesh et al. (2003) suggested the Unified Theory of Acceptance and Use of Technology (UTAUT), which combines constructs of various previous models into an exhaustive model (Persada et al., 2019). UTAUT models the influence of technology adoption by four broad determinants, namely, performance expectancy (PE), effort expectancy (EE), social influence (SI) and facilitating conditions (FC). UTAUT has gained tremendous popularity in studying technology adoption within organizations, educational and societal settings due to its capability to explain in a broad way (Nnaji et al., 2023). UTAUT has also been a widely applied research instrument in the field of higher education and digital learning research to examine the adoption of the learning management system, online platforms and AI-assisted educational resources (Granić, 2022).

Nevertheless, the current uses are more so in institutional structures, formal education settings and enrolled learners. As a contrast, digital platforms and gig-based learning spaces are informal and decentralized learning spaces, in which upskilling is self-directed, market-based and unequally supported (Hail et al., 2024). This shows that there is a significant research gap: the scarcity of research on UTAUT within gig economy learning, digital inclusion and learning equity, especially among members of Gen Z (Akter and Ahmed, 2025). As more and more gig workers depend on digital devices to make their earnings, it is necessary to learn how the acceptance and continued usage of these devices affect access to education and economic activity. The connection of the need to fill this gap with the special issue theme of quality education to all, informal learning ecosystems and inclusive digital markets is straightforward.

Since Covid-19, there has been a significant rise in digital platforms facilitating the evolution of the “Gig Economy.” During the pandemic, employees faced numerous challenges, like salary cuts, layoffs and forced leaves, to name a few. The era demanded upskilling to meet these challenges and to stand out from the crowd. Many industries shifted to a new model of online work, leading to demand for more tech-savvy employees. Another trend that cropped up simultaneously was “Gig,” wherein an employee has more flexibility to work (Singh et al., 2025). Considering these two major transitions, the educational institutions found it utmost crucial to prepare students to compete in such a challenging scenario and convert them from job seekers to “Gig Workers.”

The term “gig” offers workers the benefits of short-term, project-based employment opportunities (Turekulova et al., 2024). The expansion of information and communication technologies (ICTs) has transformed learning into a sociotechnical process that extends beyond formal institutions into digital platforms, online communities and work-based environments (Alharbi, 2023; Khan et al., 2017). For Gen Z, digital upskilling often occurs within gig platforms, AI tools, MOOCs and peer-driven ecosystems, positioning the gig economy as a site of informal and continuous learning.

Extending UTAUT, UTAUT2 incorporates intrinsic drivers such as hedonic motivation and habit, which are particularly relevant for Gen Z. Enjoyment, personalization and gamification significantly influence sustained engagement with digital learning tools, while habitual use enables long-term skill development and mastery (Elasaria and Nurabiah, 2024; Gupta et al., 2025).

In the context of professional development, Gen Z and young adults represent a digitally literate cohort that is highly receptive (96.2%) to the adoption of technology for upskilling. This demographic values flexibility, favoring systems where learning can take place “anytime and anywhere” to better balance personal schedules with demanding work environments. Their acceptance of new digital tools is strongly driven by PE, meaning they are most willing to engage when they believe the effort will directly increase their productivity or improve their chances of receiving a pay raise. Conversely, they are deterred by high costs and incompatibility between new digital platforms and the systems they already use at work (Venkatesh et al., 2003).

E-learning has become very common and comfortable for several reasons. Majorly, flexibility and global access to courses lead the list (Turekulova et al., 2024). Learning without worrying about traveling and relocating makes a student gain global exposure without leaving their motherland. The continuous upskilling provides the most relevant skills faster than conventional methods (Livingstone and Bulger, 2014). Such faster acquaintances with the updates enable learners with digital competency and the ability to perform with precision and speed. Razak and Zahidi (2024) highlight job security as the major benefit of e-learning acquired skills.

PE is a measure of whether Gen Z employees think that learning digital skills will make them more employable, more productive and earn higher salaries or not within the gig platforms. The competencies that have direct economic benefits and payoffs establish skills that encourage learning behavior in the digital market, which is highly competitive and evaluated by algorithms (Alblooshi and Hamid, 2022; Camilleri, 2024):

EE refers to the question of the perceived comfort of learning new digital capabilities and using them. In general, Gen Z is confident in using digital tools, but high cognitive or technical complexity may serve as an obstacle to inclusion, especially in the case of people who are not provided with the structure of assistance. Therefore, EE has an important role in making or breaking digital upskilling as an opportunity or a filter (Faida, 2017; Yusoof et al., 2025):

FC refer to access to infrastructure, devices, reliable internet and learning resources (Ho et al., 2025). Unequal access to these conditions remains one of the most significant barriers to equitable participation in digital markets, particularly in emerging economies and marginalized communities (Yu et al., 2024):

SI captures the role of peers, mentors, online communities and professional networks in shaping perceptions of digital skills. In gig work, peer-based knowledge sharing and community-driven learning often substitute for formal instruction, although reliance on informal networks can also reinforce existing inequalities (Lorenzetti et al., 2020; Van Zoonen et al., 2023; Waldkirch et al., 2021):

Flexibility has been identified as a major contextual factor influencing the BI to use digital technologies, especially in digitally enabled learning and working environments. Although flexibility is not a major component of the original UTAUT model (Venkatesh et al., 2003), it has been identified as a factor that positively affects users’ BI to use digital learning platforms by overcoming structural limitations and improving perceived usefulness and ease of use (Park and Youl, 2009; Sun et al., 2008). In the context of Gen Z learners participating in digital upskilling, especially in the gig economy, flexibility enables self-directed learning at one’s own pace, location and modality, thus improving BI to use such platforms (Duggan et al., 2019; Wood et al., 2019). From the perspective of digital inclusion, flexible learning platforms ease access for diverse and disadvantaged users, thus improving their willingness to engage with technology-enabled skill development (Helsper, 2021; van Dijk, 2020). Therefore, based on previous research studies extending the UTAUT model, flexibility has been identified as a major positive predictor of BI to use digital upskilling platforms, especially among younger generations seeking flexible and career-oriented learning pathways (Venkatesh et al., 2012; Tamilmani et al., 2021):

By far, the most valid and the most reliable predictor of the adoption of digital learning systems is the BI of UTAUT and its expansions (Venkatesh et al., 2003; Venkatesh et al., 2012). Past studies have shown that people with strong intentions are better inclined to accept and continue using e-learning systems in case they fulfill their skill development and career goals (Al-Fraihat et al., 2019; Park and Youl, 2009). Digitally, BI supports the bridge between access and successful use of the digital platform, particularly in the case of resource-limited users (Helsper, 2021; van Dijk, 2020). In the gig economy, intention plays a major role, with digital learning tools providing the cultivation of employability and gatekeeping to flexible employment opportunities instantly (Wood et al., 2019):

Despite UTAUT’s popularity in explaining technology and e-learning adoption, previous literature suggests that the explanatory potential of UTAUT can be further enhanced by adding context-specific variables, especially in an online learning context where flexibility is also a crucial factor (Alhashimalsayed, 2025; Venkatesh et al., 2003). Still, a significant part of the literature addresses general student groups or pandemic-related online learning settings, where flexibility is commonly viewed as a platform characteristic, as opposed to a behavioral requirement due to work/study imbalance (Chatti and Hadoussa, 2021; Songkram et al., 2023). This limitation is further amplified in the context of Gen Z learners in the gig economy, where learning has to be merged with irregular time schedules, mobile work arrangements and unceasing upskilling requirements (Rahi et al., 2019; Wu et al., 2025). Consequently, there is an apparent gap in the research on whether flexibility is an independent antecedent of BI when adopting e-learning among Gen Z gig workers (Abdaljaleel et al., 2024; Nain, 2023).

Flexibility is the most underexplored but pivotal construct when it comes to online/digital upskilling learning, as it provides a learner with locational independence to balance multiple tasks without any physical mobility. Moreover, the remote nature of e-learning platforms offers benefits like cost reduction of commuting and operating from the original place of residence (Arbaugh, 2000). Rare studies incorporated flexibility along with UTAUT as a direct driver of BI in recent times. The literature presents mixed results stating that flexibility contributes (Alkhuwaylidee, 2019; Alrawashdeh et al., 2012), and those exploring the UTAUT using system flexibility, where the flexibility did not contribute to the BIs. Although flexibility showed a strong correlation with system quality and system enjoyment (Batucan et al., 2022).

Studies that included flexibility as a variable in UTAUT were more from healthcare (Thanthrige et al., 2025), fintech and banking (Tariq et al., 2024). Thus, considering the previous studies, where flexibility is not consistently incorporated into UTAUT across the e-learning industry, this indicates the potential as a variable demonstrating or leading to BIs. As such, the research not only extends the traditional acceptance model of e-learning but also integrates adoption within the socio-behavioral context of the gig economy. The importance-performance map analysis (IPMA) findings add to the strength of the study by demonstrating that flexibility is important and operational.

According to the UTAUT and its variations in digital learning studies, the current study hypothesizes a conceptual framework that explores the factors of digital learning system adoption. As it is shown in the model, PE, EE, SI and FC are assumed to be independent variables directly affecting the BI to use digital learning systems. BI is a mediating variable, which converts positive perceptions of the users about the learning system into real “Adoption of the Digital Learning System” (ADLS). Moderating (FLEX) as an independent variable of BI and adoption (ADP) is added. The model presupposes that by thinking of the digital learning platforms as helpful, convenient, social, effectively facilitated and flexible, the learners will most likely develop a strong desire to use them, which will result in an increased adoption rate and long-term use (Figure 1).

This given model is highly applicable within the framework of digital upskilling, in which learners are becoming increasingly dependent on technology-based learning systems as a means of improving competencies, productivity and employability (Table 1).

In the existing research, the hypothesis was tested, and the mentioned relationships were tested using an empirical approach through a proposed model (Figure 1). An online survey was floated via WhatsApp and email to the targeted respondents. The total number of respondents who returned the questionnaire was 242. Keeping in mind time and cost limitations, purposive sampling was used to ensure that participants had the right experience with AI-enabled digital upskilling, thereby confirming the quality of data (Hair et al., 2018; Menon, 2025; Sekaran and Bougie, 2016).

The respondents were mostly Gen Z men and women aged 18–22 and engaged in the survey, in particular in Tier 2 city of India. The respondents were included on the basis of their willingness to participate in the survey, and particularly those who have certain experience of e-learning. As Gen Z has a high exposure to technology as compared to other groups, they were considered apt for the survey, thus comprising the targeted group.

Confirmatory factor analysis was applied to test the proposed model, and path analysis was done with structural equation modeling.

Data collection instruments were developed after a thorough review of the literature and theoretical models presented by renowned authors. The questionnaire incorporated variables from established models, sourced from various studies. PE, EE, FC and SI scale adopted from Venkatesh et al. (2003) and Kokoç (2020). The flexibility scale from Kokoç (2020) was incorporated. Further, BI was adopted from Chao (2019) and Mehra et al. (2022). The adoption scale was adopted from Hsiao (2013) and Mehra et al. (2022). The scales were modified as per the context, thus were re-standardized. Although the questionnaire was distributed to over 500 respondents, only 242 filled-in questionnaires were usable for further analysis making a response rate of 48% were included in the study. Male respondents were 128, and the rest were female respondents. A five-point Likert scale was used for data collection, ranging from 1 (strongly disagree) to 5 (strongly agree).

Statistical tools, including reliability and validity assessments, as well as common method bias evaluation, were used to re-standardize and validate the questionnaire. Confirmatory factor analysis (CFA) was used to confirm the factors in the study, followed by structural equation modeling (SEM) to test the hypothesized relationships. The IPMA was conducted to further understand the variable which needs more attention in e-learning.

Harman’s single-factor method was used to identify potential bias in the data, as per Podsakoff et al. (2003). All constructs were included in the analysis, and the results indicated that the data were free from biases. Common method bias (CMB) was evaluated using SPSS, and the results were within the acceptable threshold limit of 50% (Harman, 1976).

The measurement model performs very well and gives confidence in the quality of the constructs used in our study (Figure 2). All item loadings are strong and fall within acceptable ranges. PE items load between 0.65 and 0.78, EE between 0.61 and 0.78, FC between 0.64 and 0.71, SI between 0.66 and 0.738, BI between 0.67 and 0.74 and ADP between 0.64 and 0.79. These values mean that the items in the surveys are meaningful and understand what they were supposed to measure. FLEX, which had a factor loading between 0.87 and 0.89. The model further accounts for a significant amount of variance for BI and ADP as per the threshold value of R < 1. This, in real life, it would imply that the model elicits a big portion of what motivates intention and real adoption of the digital learning system (Table A1). The reliability analysis for all constructs shows high levels of internal consistency in this model, and that all Cronbach values are above the 0.70 threshold (ADP: 0.772; BI: 0.800; EE: 0.773; FC: 0.729; FLEX: 0.88; PE: 0.777; SI: 0.751), indicating that all scales are reliable to use in UTAUT. This is again confirmed by composite reliability scores (all above 0.87), which exceed the threshold limit of 0.70. These measures meet the Fornell and Larcker (1981) criteria and justify the further AVE measurement (with a value of > 0.50) and the discriminant validity (Hair et al., 2021). AVEs were majorly found to be greater than SIC’s thus reflecting discriminant validity as per threshold limit (Baldus et al., 2015) (Table 2).

Analyzing the direct effects from SEM (Table 3) revealed that the influence of FLEX on BI was positive and significantly high (β = 0.181, t = 6.804, p < 0.001), which means that FLEX is the strongest predictor of BI. There was also an important indirect effect of FC on BI (β = 0.181, t = 2.088, p = 0.03). The direct effect of SI was also found to be significant (β = 0.174, t = 2.062, p = 0.04), indicating SI as one of the major predictors after FLEX and FC. However, the effects of PE (β = 0.126, t = 1.793, p = 0.07) and EE (β = 0.011, t = 0.155, p = 0.87) were not significant. Based on the results, it can be confirmed that FLEX and SI are strong predictors of BI, but EE and PE do not contribute significantly toward BI. BI as moderator, when studied as an independent variable, significantly affected ADP (β = 0.793, t = 21.815, p  < 0.001), thus conforming BI as a strong predictor for ADP (Figure 3).

Analyzing the indirect effect, the results of the structural equation modeling show that the relationship between the selected antecedents and ADP is mediated by BI. The indirect effect of FLEX on ADP via BI was positive and significantly high (β = 0.372, t = 6.563, p < 0.001), which means that FLEX is the strongest predictor of ADP through BI. There was also an important indirect effect of FC on ADP via BI (β = 0.144, t = 2.032, p = 0.043). The indirect effect of SI was also found to be significant (β = 0.138, t = 1.969, p = 0.050), indicating support for the mediation. However, the indirect effects of PE (β = 0.100, t = 1.644, p = 0.102) and EE (β = 0.009, t = 0.156, p = 0.876) were not significant. Based on the results, it can be confirmed that BI is a meaningful mediator for FLEX, FC and SI but not for EE and PE (Table 4).

IPMA shows that FLEX is the most influential construct in the model, with performance score of 73.705. The index value of 3.948, indicates a room for improvement in the score. Similarly, BI is also found to have a high relevance in the model, with a performance value of 72.294 and an index value of 3.892, which reaffirms BI as a key mediating construct although there is room for improvement in scores. SI’s performance score is moderate at 68.912, and its index value is 3.756, which means that it is moderate in terms of relative standing. Similarly, FC is recorded with the lowest performance score (67.295), indicating that it is moderate in terms of relative standing. FC and SI show moderate importance values of 0.144 and 0.138, respectively; thus, they matter but not very urgently (Figure 4).

PE had performance values of 70.644, despite a reasonable effect (0.100), confirming that performance (3.826) is already adequate considering its importance. The effect of EE is a near-zero effect on ADP (0.009), so improving it will not meaningfully enhance ADP. Summarizing the IPMA results, it is concluded that FLEX and BI are the most strategically relevant constructs, while FC and SI are less strong, and therefore merit focus on improvement efforts (Ringle and Sarstedt, 2016; Hair et al., 2017) (Table 5).

For digital upskilling, flexibility has appeared as the most significant and positive predictor of BI. Users are significantly more likely to intend to use the system if they perceive it as flexible. It matches the user’s needs; it also supports various user preferences. The overall flexibility reduces disruption, making the user more certain of adoption from anywhere and anytime without relocating at times (Abdaljaleel et al., 2024; Adapa et al., 2019; Jo and Bang, 2023) (H5). FC is completely unavoidable as access to necessary resources, the support system and compatibility with the technology reduces barriers to intentions, and thus, users gain confidence in using the same (Alhashimalsayed, 2025; Yang and Qian, 2025) (H3), although not supported by Utomo et al. (2021).

H4,SI has a significant positive impact on BI. The findings indicate that in the current era of social media and the internet, the opinions of others have a meaningful contribution; others play a meaningful role in shaping a user’s intent to adopt. Be it peer opinions, reviews or any influencer endorsement, all in an integrated manner, can impact the BIs. Moreover, a sense of belongingness is generated via many social media groups, which can affect the BI supported by Taamneh et al. (2022), Liang et al. (2024) and partially supported by Nugraha et al. (2025). PE did not affect the BIs. In the gig economy, factors like flexibility, FC and SI overshadow the PE. Especially in a gig economy, a gig worker is more flexible and income-driven. Moreover, a gig worker has a quality of being tech-savvy, which diminishes the motivational importance of perceived performance gains. Therefore, the routinized use of tools generated a familiarity bias, which leads to overlooking PE supported by Jabagi et al. (2019) and Kurian and Madhavi (2024)(H1).

Specifically, for the digital learning methods, for H2,EE showed an insignificant impact on BI. At times, the users have the intention to use, irrespective of the amount of effort required to put in. Considering the fact that nowadays, the organizations have well-established helpdesk and customer relationship managers, due to which the task needs less of efforts supported by Wu et al. (2025) and contradicted by Rahi et al. (2019).

H6. Finally, behavioral intention (BI) strongly affects the adoption (ADP), considering that the behavioral intention is associated with individuals’ upskilling goals. In case the learner considers that high value, the intention toward adoption increases. Here, the intention included self-motivation, which is the most crucial parameter for adoption (Songkram et al., 2023).

The SEM results confirm that FLEX, FC and SI strongly predict BI and PE and EE do not. From a practical point of view, the IMPA results further support this interpretation, as the FLEX result in terms of performance and index value is the highest, while BI is second, suggesting that these are the most leveraged operational and strategic levers to boost substantive performance in terms of ADP. Therefore, interventions to increase flexibility-related mechanisms and BI are more likely to be statistically influential and more actionable than interventions to increase other, less statistically influential (e.g. FC and SI) and insignificant ones (EE and PE). From a managerial aspect, the e-learning platforms must not promote the design and ease aspect of the platform, as PE and EE do not contribute to BI and indirectly to adoption. Rather, an intense investment is to be made in making a supportive learning ecosystem.

The platforms must be designed asynchronously. Starting with the content, it has to be user timing and pace-driven, i.e. the user is not bound to access at a decided time or duration. Moreover, the platform can be accessed from any range of device, be it mobile, laptop, desktop, tablet, to name a few, and further data can be saved in multiple modes. Another important outcome the research has elaborated is the FC. Troubleshooting is a must when it comes to an IT support system, which is the bare minimum expected. Being the e-learning, the “internet” becomes the backbone of the entire system; thus, a reasonable speed internet with uninterrupted connectivity makes any learner focus more on the learning part. In many instances, the platform needs to provide customized FC matching the needs of individuals.

Another important finding revealed that SI affects the BI. Thus, more opinion leaders, peers, top management and supervisors share the content and its relevance. It is the human tendency to listen to the actual associated user rather than any paid promotion. Public recognition and acknowledgments are quite fruitful. For instance, announcing “achiever of the month,” “star performers,” “celebrity testimonial,” “user feedback,” “thank-you note from top management,” “award ceremonies,” “best employee of the month” can act as a game changer, to which a user connects instantly.

The present paper discusses the topic of digital upskilling platforms, their adoption and BI among e-learners. Being integrated with other models of technology adoption and incorporating variables of other models may present a better picture of the digital upskilling platform adoption. It is possible to add a qualitative research approach. Qualitative research may provide more information about the emotions and feelings related to digital learning and the real causes of e-learning implementation. Although the research was done with the help of SEM using Smart PLS, investigating alternative software would give more insights, taking into account the specific limitations of the tool expansion.

The study was conducted with respect to Indian students in a Tier 2 city without sufficient or efficient physical facilities. But the focused geographical location could affect the generalizability of the results. This is recognized, and future research should consider using more diverse samples from multiple locations to improve generalizability.

The study could also be improved by expanding the scope, which would allow increasing the geographical areas of the investigation. Since the tendency toward e-learning is growing, the inclusion of participants with different age groups might provide an interesting piece of information.

The research on online upskilling sites identifies the most important aspects causing the BI of e-learners. A product in the form of constructs such as flexibility, FC and SI, increases the intention, desire and at the same time, stimulates the e-learners.

The two most important constructs, i.e. flexibility and FC, have come out as the most significant factor that the BI and the process of commitment by working on the intent using peer perception and flexible systems. SI, remains a game changer if used in an effective manner, as human connects to human faster than anything. By contrast, PE and EE are least affected by tech-savvy characteristics of the user and the introduction of robust and well-established customer support and helpdesk. Finally, BI is the strongest predictor in the adoption process. These results can be used by organizations that encourage digital upskilling efforts.

This research received no specific grant from any funding.

The authors sincerely thank all the respondent for sharing their valuable responses via shared data collection tool.