Understanding factors influencing adoption of digital tools among extension agents of Nepal Open Access

Agricultural extension plays a vital role in improving the livelihood of farming communities through expert support, information dissemination and subsidy disbursement. Farmers are the primary recipients of the extension service and need timely and seasonal information to make informed decisions throughout the year. Given the unique livelihood desires of the farmers and heterogeneity in farming systems, demand-driven extension services are essential (Hainzer et al., 2023). Because of this, the concept of agricultural extension is evolving from linear view of technology and information transfer to more system-oriented perspectives directed towards addressing multiple social and economic problems (Spielman et al., 2021; Takeshita et al., 2023). The success of extension services depends on the effective exchange of information and technology among agricultural researchers, extension agents and farmers (Mustapha et al., 2022).

In Nepal, government-led public extension organizations are responsible for disseminating agricultural knowledge, information and technological practices from researchers to the farmers, while also collecting farmers’ problems and feedback (Sigdel et al., 2022; Thapa et al., 2020; Timsina et al., 2023). After federalization, extension services in Nepal have been decentralized to the provincial and local levels and are delivered through seven provinces and 753 local agricultural extension units. Over the years, Nepal has experienced various extension approaches, including Training and Visit system, Tuki System, Block Production Program, Farmer Group approach, Pocket Package Program, Projectization approach, Farmer Field School and Public–Private Partnership approach (Ghimire et al., 2021; Timsina et al., 2023). Despite these efforts, the effectiveness of the extension service has been poor, achieving little success in the face of evolving government policies, diverse needs of farming communities and advancements in communication technologies.

The agricultural extension service sector in Nepal is characterized by inadequate funding, high transaction costs and limited reach that are not sufficiently tailored to farmers’ needs. This is supported by the fact that the allocated budget in the agricultural sector is only 2–3.5% of the total federal budget and 5–11% of the provincial budget over the past decade (Pandit and Dhakal, 2022). Moreover, 60% of the agricultural budget goes to inputs like chemical fertilizers, and less is allocated for extension programs (Timsina et al., 2023). In addition, only 25% of the farmers benefit from public extension programs (Ghimire et al., 2021). The limited reach is due to a high ratio of extension agents to farm households, which is 1:1328 (Timsina et al., 2023). The ratio is even worse at the provincial level. For instance, out of seven provinces, Bagmati and Gandaki provinces are better staffed for extension services, but the ratios are still high at 1:2200 and 1:1660 respectively (MOEA, 2023; MOEAP, 2023). This gap limits the farmers’ access to extension agents (Amrullah et al., 2023) and is further widened by factors like inadequate expertise and poor communication infrastructure within the extension organizations (Nyarko and Kozári, 2021; Thapa et al., 2020).

Information and Communication Technologies (ICTs) can bridge the gap between extension agents, researchers and farmers (Mustapha et al., 2022), enhance information accessibility and reduce associated costs (Fabregas et al., 2019; Yahaya et al., 2018). Torero and von Braun (2006) define ICT as “the equipment and services that facilitate the electronic capture, processing, display, and transmission of information” (p. 3), and include digital technologies like the Internet, smartphones, data analytics and sensor-based devices. Digital extension tools are defined here as technology-driven and digitally enabled ICTs designed to support agricultural extension services and improve communication with farmers (Rajkhowa and Qaim, 2021). They provide timely information and up-to-date knowledge, enhance feedback mechanisms and help farmers make informed decisions (Singh et al., 2023). Digital extension tools are efficient in addressing challenges that arise from natural disasters like droughts, floods, landslides, pest outbreaks and other emergency needs (Nyarko and Kozári, 2021). Furthermore, situations like the COVID-19 pandemic have also presented the need for digital extension services (Olagunju et al., 2021), as physical distancing measures showed existing practices implausible. Therefore, extension organizations in Nepal can take the opportunity by transitioning to digital systems that lower dissemination costs for service providers and reduce search costs for farmers ( Bhusal et al., 2021; Dissanayake et al., 2022; Kandagor et al., 2018; Olagunju et al., 2021; Timsina et al., 2023).

With the advancements and adoption of digital extension tools, many countries have been successful in improving the effectiveness of agriculture extension services (Nyarko and Kozári, 2021; Spielman et al., 2021). Nepal has also taken some initiatives to digitize the sector through digital weather forecasting, digital seed systems, market information, e-libraries, mobile applications and digital disbursement of subsidies (Gupta, 2022).These efforts are aimed at technology transfer, real-time market information, weather forecasting, advisory services, online marketing, financial services and data management (Paudel et al., 2018; Timsina et al., 2023). However, these initiatives are not integrated across all extension organizations or are not updated regularly, and many service recipients are not aware about them or do not utilize them to their full potential. In addition, their adoption by extension agents is very low (Gupta, 2022; Sigdel et al., 2022), leading to poor performance of extension services. Thus, there is need to identify factors that influence the adoption decision of digital extension tools among extension agents.

Previous studies have primarily focused on farmers’ perspectives and their adoption of digital technologies (Gautam, 2018; Mishra et al., 2023), but little attention has been directed to the perspectives of extension agents. Moreover, past studies have identified factors such as age, education, gender and institutional support as determinants of technology adoption (Dissanayake et al., 2022; Mustapha et al., 2022; Obisesan, 2014; Yakubu et al., 2013) but has often overlooked the influence of participants’ attitudes. Given the prominent role of digital extension tools in service delivery and the array of challenges faced by extension agents, there is need to identify the determinants of their adoption decisions.

In this study, we offer both empirical and methodological contributions to the understanding of adoption of digital extension tools. Empirically, we aim to identify the factors that influence the adoption of digital extension tools among extension agents. In addition to socio-demographic, institutional and professional factors, our study examines the influence of attitudinal factors. Methodologically, we use the case of digital extension tools adoption among extension agents to complement Roger’s Innovation Diffusion theory. This theory provides a foundational framework for understanding the adoption process of new technologies by categorizing adopters into five groups, ranging from innovators to laggards based on time duration required for the adoption (Rogers, 2003; Rogers and Adhikarya, 1979). It measures the rate at which the innovation is diffused and adopted by different adopters (Wani and Ali, 2015). However, it is important to note that adoption rate also varies with increasing awareness of new technologies (Bhusal et al., 2021; Rai and Moktan, 2014), the characteristics of the technology and attitudes of the adopters about the technologies (Dissanayake et al., 2022). In this study we examine the attitudes of the participants to categorize them into distinct segments. This segmentation will enable practitioners to inform targeted interventions and differentiated strategies. These strategies not only address the specific barriers identified within each segment but also mobilize extension agents based on their unique needs and characteristics. This facilitates adoption and effective use of digital extension tools within extension system and help bridging the existing gap with the farmers.

In the next section, we present the data and methods used for our study. After that, we present findings of the study along with their interpretation. Subsequently, we engage in a through discussion to contextualize our findings within the framework of existing literature and practical implications. Finally, we draw conclusions with key implications for policy, practice and future research.

The study employed a survey research method (Fowler, 2014) to collect information from extension agents of Bagmati and Gandaki provinces of Nepal from February to March 2024. These two provinces were purposively selected due to their significant contributions to Nepalese agriculture and relatively advanced digital infrastructure in comparison to other provinces that facilitates the possibility of digital expansion. Bagmati includes the national capital, and Gandaki is an adjoining province. These provinces together have 30% of the total population (NSO, 2023) and contribute 45.8% to national GDP (Ministry of Finance, 2022). The agriculture sector alone contributes 26.60% and 23.5% of the GDP in each province respectively (MOEA, 2023; MOEAP, 2023).

The sample frame consisted of extension agents with hierarchical ranks ranging from sixth to tenth and working in the selected provinces. This selected range of ranks are the core functional group, while ranks lower than these serve in assisting roles and higher ranks perform administrative tasks. A list of 175 such extension agents was compiled from the respective provincial agricultural ministries. The questionnaire was developed and distributed using Qualtrics software, in English and Nepali languages. The study received ethical approval from the Institutional Review Board of The Ohio State University (#2024E0099).

The survey instrument was reviewed by panel of experts. We pre-tested (Dillman et al., 2009) it with experienced extension professionals (n = 10). To enhance response rate, we implemented personalized and repeated communication methods (Monroe and Adams, 2012). We personalized the survey by sending individual emails to each participant, addressing them by their name and communication was repeated through reminder emails with personalized links to partial and non-responders at two weeks intervals. A total of 128 participants provided useable data with response rate of 73.14%.

The survey included inquiries about the usage patterns of digital extension tools, proficiency in using those tools, degree of agreement on 23 statements and other professional, organizational and socio-demographic information. To explore the usage patterns, participants were asked to categorize ten digital tools (phone calls, voice message, office website, email, Viber, short message service-SMS, Facebook, WhatsApp, messenger and mobile applications) into three categories, “not used”, “used sometimes” or “used often”. For proficiency, participants were asked to indicate their frequency of engagement with digital tools, types of tools used and general knowledge in navigating and using them. Specifically, we explored online payment methods, website features navigation skills, use of social media (e.g. Facebook) and familiarity with virtual meeting platforms.

To understand participants’ attitudes, we administered 23 statements across nine topics that are directly or indirectly related to adoption of digital extension tools. These topics include organizational challenges, digital infrastructure, personal preferences, audience characteristics, perceived benefits and barriers, organizational support, workload balance, practical utility and potential of integration of those tools into the existing extension system. Participants expressed their level of agreement on each statement using a 5-point Likert scale. These ratings provided insights into their thoughts, beliefs and preconceptions about the diverse aspects of digital extension systems.

Participants’ professional information like experience, hierarchical rank and organizational support details were collected to understand the background and context details of their workplace. Sociodemographic information collected were gender, ethnicity, education level and annual household income. Table 1 represents the summary characteristics of the participants.

Three empirical methods were used in this study. Firstly, we employed factor analysis to condense 23 attitudinal statements into a smaller set of representative factors. Secondly, a cluster analysis was used to categorize extension agents into distinct segments based on the factors resulting from the analysis of attitudinal variables. Finally, logistic regression modeling was used to identify the determinants of the adoption of digital extension tools among extension agents.

Factor analysis is a statistical tool to identify underlying variables by clustering related variables in the same factor (Shrestha, 2021). Factors are latent constructs formed by combining measured variables and maintaining a similar or minimally reduced amount of the original variance. Before consolidating the 23 attitudinal variables and identifying common factors with high correlations, the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity were executed to assess the appropriateness of conducting the factor analysis. Sequentially, factors were selected to explain the maximum variability within the dataset (Shrestha, 2021). This process retains essential information, allowing to focus on key factors by reducing data complexity and enhancing interpretability (Kim and Mueller, 1978). Factors were retained, and each factor possessed a number of variables loading onto it. Factor loading measures indicate the degree of correlation between the original variables and the identified factor, quantifying how much variability in an observed variable is explained by that factor.

The factor extraction process followed to identify the optimal number of factors using the Kaiser criterion and scree plot method. The Kaiser criterion suggests retaining factors with eigenvalues greater than one, while the scree plot involves plotting eigenvalues in descending order and identifying the point before the last major drop (Costello and Osborne, 2019; Shrestha, 2021). The factor extraction was performed in two stages: first without rotation and subsequently with varimax rotation to achieve a simpler and more interpretable factor structure. Varimax rotation is an orthogonal rotation method that maximizes the variance of loadings within each factor and represents the relationship between variables and factors (Shrestha, 2021). In this method, the spread in loadings is maximized where high loadings become even higher after rotation, while low loadings become lower. The Cronbach’s alpha test was then implemented to measure reliability and internal consistency of the retained factors.

After identifying factors, factor scores were computed as a weighted linear combination of the variables based on the corresponding factor loadings. The scores were then used to partition respondents into distinct groups, a process referred to as segmentation. This process is predominantly used in marketing research where companies identify market segments based on the understanding of consumer behavior, aiming to design efficient segment-specific communication tactics (Kotler and Keller, 2016) and prioritize consumers with the greatest chances of satisfying. We employed a nonhierarchical approach with K-means clustering (Hartigan and Wong, 1979) to divide participants into segments, with each observation assigned to the cluster whose mean is closest (Skevas et al., 2014). The method differentiates segments with greater internal similarity compared to between-segment distinctions with clear boundaries.

The Caliński–Harabasz pseudo-F index was used to determine the optimal number of clusters. Initially, the pseudo-F index showed an upward trend with the addition of more clusters and began to decline after the third. Hence, three clusters were deemed appropriate. Names for each cluster were derived through qualitative analysis of participants’ responses to the attitudinal statements (variables) loaded onto the retained factors, which capture the most prevalent characteristics of each segment. Finally, clusters’ factor score means were submitted to an analysis of variance to examine attitudinal differences among clusters. The segments were labeled “Enthusiasts”, “Conservatives” and “Pragmatists”. The characteristics of each segment are presented in the results section.

A probabilistic model was developed where “adoption” is the dependent variable, treated as binary, indicating whether extension agents choose to adopt digital extension tools or not. The independent variables are characteristics of the participants including socio-demographic, professional, organizational, digital proficiency and the attitudinal segments. The description of variables used in the model with their expected relationships are presented in Table 2.

We estimate the probability of adoption using a logistic regression model. The functional form of the model for “adoption of digital extension tools” is given below:

The explanatory variables were grouped into five vectors. The segment membership (res_seg) consisted of three segments (“Enthusiasts”, “Conservatives” and “Pragmatists”) of participants identified through cluster analysis. Here, the “Enthusiasts” segment was used as the baseline for comparing against “Conservatives” and “Pragmatists”. Professional characteristics (prof_cha) includes the participant’s rank, years of experience and the diversity of services provided. Organizational support (org_sup) includes the type of support received such as training, office space with Internet access and expert support to deliver service through digital extension tools. Participant’s familiarity with mobile applications and the use of analytic platforms are grouped under digital proficiency (dig_prof) vector. Socio-demographic characteristics (socdem_cha) considered are gender and educational level. Our sample does not have individuals identified as third gender or non-binary. The overall rate of correct classification is reported following the cutoff point estimation.

The attitudinal statement variables were submitted to suitability tests including the KMO measure and Bartlett’s test of sphericity. The tests returned a KMO value of 0.56, and Bartlett’s test obtained a p-value smaller than 0.001 (i.e. the variables are not orthogonal and therefore intercorrelated to some degree). Considering the relatively small sample size used in the analysis (n = 128) but the sample representativeness of the extension agents operating in Bagmati and Gandaki provinces of Nepal, the tests indicate that the factor analysis is deemed appropriate for the data.

The factor analysis results suggest that three factors should be retained for rotation based on the criterion of eigenvalues greater than one (Costello and Osborne, 2019) and the total variance explained. In our analysis, the retained factors returned eigenvalues above 1.02 and captured 31.5%, 27% and 14.4%, respectively, of the total variance embedded in the original data. Hence, 73% of the common variance originated from the statement responses were retained in the selected factors (Table 3). These results are superior to standard recommendations for retaining factors that explain at least 50% of the total variance. Furthermore, our results are better than those in other studies that also determined factors based on responses to attitudinal statements and employed data reduction techniques (Skevas et al., 2014; Shrestha, 2021; Birol et al., 2007).

The identified factors were named “Operational challenges”, “Practical utility” and “Support system”, representing five, four and one variable loading onto each, respectively (Table 4). We performed factor rotation to better distribute the variables across the factors. According to Stevens (2002), factor loadings should exceed 0.4 for meaningful interpretation of results. In our application, we set the factor loadings threshold at 0.45, meaning that only variables with measured correlations with the generated factors above 0.45 are representative elements of their respective factors. This criterion ensures that only strong correlations between variables and factors are retained. The loadings threshold adopted in this study is more restrictive than other factor analysis applications found in the literature (Kontoleon and Yabe, 2006; Birol et al., 2007; Shrestha, 2021). Reliability and internal consistency of factors were estimated using Cronbach’s alpha test. The results obtained are reasonable as the correlations between the measured scales and factors “Operational challenges” and “Practical utility” are 0.806 and 0.788, respectively. Hence, internal consistency is confirmed to be high to moderate, with our alpha coefficients similar to results found in other studies (Skevas et al., 2014; Kikulwe et al., 2011). The Cronbach’s alpha test was not conducted for “Support system” as it is a single-variable factor.

In Table 4, we see that the variables loaded under “Operational challenges” reflect associated challenges in the use and operation of digital extension tools as perceived by extension agents. Some of the challenges include organizational resistance, inadequate support, perceived disturbance to work–life balance and farmers’ digital illiteracy to obtain technical advice. The factor “Practical utility” represents the perceived advantages and opportunities offered by applying digital tools for extension service. Some advantages reported by participants include increased recognition of the service, reduced cost and enhanced service coverage. The third factor, “Support system”, represents the importance of prioritizing support mechanisms in terms of training and capacity building for the application of those tools.

Cluster analysis was computed to group extension agents into distinct segments derived from their attitudes regarding ICT use for information dissemination. The K-means method and the Caliński–Harabasz pseudo-F test were employed to decide the optimal number of clusters. Results indicate that the sampled extension agents can be grouped into three distinct segments comprising 30, 32 and 38% of the total 128 study participants (see Table 5).

Based on the qualitative analysis of the participant’s responses to Likert-scale questions, the segments were named “Enthusiasts”, “Conservatives” and “Pragmatists”. These names capture the most prevalent characteristics of the study participants and are summarized in Table 6. Skevas et al. (2014) adopted a similar strategy to name the studied clusters.

Results from an analysis of variance support the qualitative differences in attitudes among segmented extension agents. Pairwise comparisons of factor score means confirm that the three retained factors have means that differ across segments. Two exceptions exist, however. First, the differences between “Practical utility” factor score means in the “Enthusiasts” and “Pragmatists” segments are not statistically significant. Second, Enthusiasts and Conservatives share statistically equivalent score means for the “Support system” factor. This result shows that the factors utilized to aggregate statement responses are relatively important to one another for segmenting the agents’ attitudes toward ICTs to enhance agricultural extension practices. Table 7 summarizes these findings.

A logistic regression model was conducted to assess the likelihood of extension agents’ adoption of digital extension tools. Table 8 presents the result of the logit regression.

According to the results, the “Conservatives” segment is less likely to adopt digital extension tools compared to “Enthusiasts”. Being a member of the “Conservatives” segment decreases the probability of adopting digital extension tools by 27%. The marginal effect for “office rank” is 0.135, indicating that a one-unit increase in rank results in a 13.5% increase in the probability of adopting digital extension tools. “Experience” has a negative coefficient; more experienced extension agents are less likely to adopt digital extension tools. The marginal effect for the variable “experience” is −0.020, suggesting that a one-year increase in “experience” of participants decreases the adoption probability by 2%. Having “office space with Internet access” and “training support” exhibit significant negative effects, decreasing the adoption probability by 32.6% and 28.8%, respectively. However, previous use of agricultural mobile applications significantly increases the adoption probability by 34.8%. Furthermore, results show that being male is associated with a 21.9% higher probability of adoption compared to females. Table 8 reports the log pseudolikelihood value (−71.715) as a measure of model fit.

The capability of the reported model to predict adoption of digital tools by extension agents in Nepal may be of particular interest. We studied our model capability by fitting the Receiver Operating Characteristic (ROC) curve and identifying the optimal cutoff point through the computation of Youden’s index (Kallner, 2018). Results indicate that our model can correctly predict 72.7% of the observations. The area under the ROC curve is estimated at 0.779, demonstrating that the fitted model has a reasonable predictive power considering the relatively small sample size. Table 9 summarizes these findings.

This study sought to explore the factors influencing adoption of digital extension tools among extension agents in the Bagmati and Gandaki provinces of Nepal. Three factors, “Operational challenges”, “Practical utility” and “Support system” derived from the factor analysis reflected that participants’ perceptions are driven by both challenges and opportunities in using digital extension tools. The challenges in integrating digital extension tools expressed by study participants were lack of demand from farmers, absence of standardized digital systems, digital illiteracy among farmers and agents, limited Internet access and concerns about the reliability of information. Previous study also outlined weak network connections, lack of ICT training and inadequate ICT infrastructure as significant barriers for adoption (Nyarko and Kozári, 2021). Participants also acknowledged the benefits of digital extension tools and considered them as plausible options to expand service coverage at a low cost. Perceived benefits were increased recognition, reduced costs, enhanced service coverage and improved competitiveness in the agribusiness sector. They also expressed their willingness to invest time and effort in mastering digital extension tools.

Cluster analysis divided extension agents into three segments: “Enthusiasts”, “Conservatives” and “Pragmatists”. The “Enthusiasts” segment grouped extension agents who are optimistic about digital extension tools. Extension organizations of Nepal can mobilize this segment in digital initiatives to expedite adoption, perhaps by considering them innovators (Rogers, 2003) and using them in peer-to-peer programs. In contrast, the “Conservatives” segment expressed reluctance toward using digital tools. This reservation may stem from concerns about technical reliability, organizational barriers and perceived workload increases. Organizations should encourage this segment by providing digital literacy programs. Educational modules aimed at this group could stress that the integration of digital tools within the existing extension system may ease travel time and enhance coverage. The “Pragmatisms” segment took a balanced approach as they recognize both the benefits and challenges. They showed interest in using digital extension tools along with existing practices. As this segment valued practical outcomes and evidence-based adoption, they may be utilized as the pilot group in testing interventions to promote new digital tools. Their feedback can help develop more effective and user-friendly digital solutions.

The identification of three segments reflects the diverse perspectives of extension agents of Nepal and provides a basis for effective interventions. Results demonstrate that a one-size-fits-all approach is unlikely to succeed in motivating a transition toward digital extension programs. The varying attitudes across segments serve as evidence, and a segmented approach to motivate the adoption of digital tools may be a more effective pathway. It should be stressed that our methodological approach complements Rogers’s diffusion of innovation theory (2003). While Rogers categorized potential innovation adopters based on time duration required for the intended behavior, here participants were segmented based on their attitudes, beliefs and preconceptions. We add to the literature on technology diffusion by demonstrating empirically that segmenting the target population and understanding their unique perceptions are critical starting points to effectively promote technologies in the context of agricultural communication and education.

In the logit model, “hierarchical rank”, previous use of agricultural “mobile applications” and “gender” were found to have a significant positive influence on the adoption decisions. As most of the extension agents with higher ranks are either leading district extension organizations or provincial-level resource centers and technical laboratories, they have greater decision-making authority and access to resources in Nepal. In addition, they are often provided with resources like smartphones, laptops and data packages by their organization. This could be the reason that higher ranked extension agents were more likely to adopt digital extension tools.

Previous use of mobile applications for agricultural communication and news sharing exhibited a significant positive influence on adoption. This may be due to familiarity and comfort that arise from regular use, reducing the perceived complexity and increasing the agents’ confidence in using applications for professional purposes. Due to user-friendliness features and widespread availability of mobile devices, mobile applications can be potential platforms for disseminating agricultural information that originates at provincial or district extension organizations. This finding is also supported by previous studies reporting that extension agents are more likely to utilize advanced technological features once they become comfortable with the basic functionalities of mobile applications (Kandagor et al., 2018).

The dominance of male extension agents (70.30%) shows the presence of gender disparity within the extension system. Mustapha et al. (2022) and Yakubu et al. (2013) also reported similar imbalances in the public extension service in Nigeria. Such disparities not only affect the composition of the workforce but also lead to gender-biased technology development and transfer. Our findings align with Obisesan (2014), who reported gender-based influence on technology adoption and observed a higher adoption rate among male agents. This trend is likely due to their dominant household roles and better access to resources than females (Bhusal et al., 2021; Morris and Doss, 1999). This finding underscores the need for targeted interventions to increase female representation and participation in extension services to ensure equitable access to technologies and bridge the gender digital divide.

In our analysis, the “Conservatives” segment, “year of experience”, “office space with Internet support” and “training” were found to have a significant negative influence on the adoption decision. The “Conservatives” segment is less likely to adopt digital extension tools compared to “Enthusiasts”, plausibly because of their uninformed perceptions. In their opinion, digital tools are costly, increase workload and force them to respond to clients even outside office hours. Fabregas et al. (2019) highlighted that many users believe digital technologies require much more research, development and testing before they can be brought to scale. Conservative participants in our study may have shown reluctance toward adoption due to beliefs further explored by Fabregas et al. (2019).

The study participants’ average years of experience is 11, and their average age is 36 (Table 1). These characteristics suggest a relatively youthful demographic in the study area. Earlier, we presented that hierarchical rank positively influences adoption, and here we see years of experience is negatively associated with adoption. This pattern may be attributed to the fact that most entry-level officer positions begin at level seven, with subsequent promotions to higher ranks. However, it is important to note that level sixth agents are likely to have achieved this rank by entering at level fourth or fifth. In most cases, level sixth extension agents possess more years of experience than those in higher ranks. Hence, technological advancements pose challenges for this rank as they may hesitate to adopt new tools or have less inclination to use them, being accustomed to paper-based work and direct interaction with farmers to provide service. This hesitation contrasts with younger agents, who according to Yakubu et al. (2013), have higher awareness about digital tools and their capabilities.

While prior research has often identified training as a positive influencer of technology adoption (Mustapha et al., 2022; Nyarko and Kozári, 2021; Yakubu et al., 2013), our study presents contrasting findings with training being a negative factor. Study participants expressed concerns about the ineffectiveness of trainings, noting that it often fails to cover ICT-related subjects or if it is offered, the absence of platforms or established working systems within the organization renders it ineffective for adoption. This is further supported by our finding that expert support within the organization was found to have a positive but not significant influence. This suggests that while having access to experts may be beneficial, it is not sufficient on its own to drive adoption. The staggered nature of trainings and investments in ICT support personnel may be leading to unevenly distributed awareness and interest among agents. In other words, not all agents receive the appropriate training and technical support simultaneously, resulting in isolated individuals lacking the necessary resources to fully apply the acquired skills.

Similarly, “office space with Internet support” was found to negatively affect the adoption. This may be due to a disconnect between organizational support and employee performance measures (Mustapha et al., 2022). Mustapha et al. (2022) found that perceived organizational support negatively influences ICT adoption due to lack of motivational factors. Olagunju et al. (2021) also highlighted that lack of motivation among extension agents leads to service inefficiencies. In supplementary open-ended questions, study participants expressed that they use digital tools primarily for personal communication and entertainment rather than official purposes. They also expressed willingness to invest time and effort in mastering digital tools for professional use if sufficiently motivated. Currently, extension organizations lack digitally enabled systems and tend to rely on paper-based and in-person extension practices. As a result, the available Internet facilities are often utilized for personal communication and entertainment. For instance, public service organizations in Nepal have often faced criticism for using social media platforms during office hours, sometimes leading to leakage of official information and dissatisfaction (Rai and Moktan, 2014). In response, the Government of Nepal imposed a ban on social media use during office hours at the central administrative level in 2012 (Rai and Moktan, 2014).

The current extension system in Nepal characterized by limited reach and inadequate budget, realities that are not uncommon among developing countries. While mitigating strategies to address these weaknesses lie in integrating ICTs into existing extension systems, barriers to adoption, including attitudes and behaviors of front-line service providers must be precisely investigated. Our study shows striking results. Of particular interest is that training sessions and the provision of office space with Internet negatively affect the adoption of digital tools among Nepalese extension agents. Moreover, ICT expert support has a non-significant impact on adoption. From these findings, we discuss that the current allocation of resources is not contributing to the adoption of ICTs in Nepal, meaning that resources are not utilized in a productive way. Therefore, policymakers may use our results to revisit their investment plans and resource allocation strategies to initiate a different route to facilitate the integration of digital tools into Nepalese extension services.

The findings also call for the design of training programs that are practical and relevant to the subject matter. Additionally, providing conducive working environments to apply what is learned is paramount. The role of organizational support is crucial alongside the design attributes, hardware, interface and usability characteristics of ICTs that significantly influence outcomes and economic choices for extension agents. System-level factors, such as the organizational structure and governance of extension services, should be favorable to foster adoption.

The attentive reader may notice the usefulness of the present empirical case. While the current literature has stressed the importance of ICTs to help address multi-dimensional food production challenges associated with the growing heterogeneity of farming systems around the world (Nyarko and Kozári, 2021; Spielman et al., 2021), our results invite analysts to further investigate the reasons (i.e. the why and how questions) preventing front-line support providers (i.e. extension agents) from adopting ICTs at faster rates. In the case of Nepal, extension agents’ attitudes matter as much as the quality of training sessions, resource allocation protocols and personal comfort with existing digital applications for obtaining agriculture-related news. The results found for Nepal should not be extrapolated, but the methodology employed in this study could help unveil adoption barriers occurring in other cases.

Moving forward, Nepalese extension organizations should mobilize and promote extension agents based on their unique attitudinal characteristics through targeted interventions and strategies to enhance the widespread adoption of digital tools for effective agro-advisory services. Policymakers should prioritize the development of digital infrastructure within the extension system, and extension organizations should invest in continuous professional development and support systems to encourage innovative practices.

The findings of this research are context-specific to the selected provinces of Nepal and may not be directly applicable to the whole nation or other countries. However, the methodology of segmenting participants based on attitudes, beliefs and preconceptions can be applied in studies of adoption and diffusion of innovations across diverse fields. Future research should replicate similar studies in other contexts with larger sample sizes to capture potential factors affecting adoption decisions while recognizing that segmentation and targeted interventions are likely to increase the chances of successful implementation. Furthermore, studies can be undertaken to investigate extension agents’ motivations that enhance the adoption rate of digital tools over time.

Funding: The authors declare that there is no funding associated with this research.