Abstract
In the digital era, agricultural technology enterprises increasingly rely on external knowledge to cope with interdisciplinary complexity and technological uncertainty. Drawing on the attention-based view, this study examines the associations between boundary-spanning search and disruptive technological innovation, with particular focus on the role of knowledge orchestration capability. Organizational unlearning and digital orientation are further incorporated as contextual conditions shaping these relationships. Using cross-sectional survey data from 412 agricultural technology enterprises in China, this study employs hierarchical regression analysis and bootstrapping procedures to test the proposed associations. The results indicate that both breadth-oriented and depth-oriented boundary-spanning search are positively associated with disruptive technological innovation. Knowledge orchestration capability is positively associated with both forms of boundary-spanning search and with disruptive technological innovation, suggesting that it helps explain how search activities are linked to innovation outcomes. In addition, organizational unlearning strengthens the associations between boundary-spanning search and knowledge orchestration capability, while digital orientation strengthens the relationship between knowledge orchestration capability and disruptive technological innovation. Overall, this study contributes to the literature by distinguishing between two forms of boundary-spanning search and by clarifying how attention allocation and knowledge orchestration are associated with disruptive innovation in agricultural technology enterprises. It also provides context-sensitive insights for managers seeking to improve external knowledge search and integration practices in agricultural technology enterprises. The findings are interpreted as associative rather than causal and are grounded in self-reported, cross-sectional data from the Chinese context.
Keywords
Introduction
Agricultural technology enterprises are increasingly embedded in an environment characterized by intensive interdisciplinary integration and the rapid evolution of technological paradigms (Ribašauskienė et al., 2024). Disruptive technological innovation in this sector relies heavily on knowledge from diverse domains, including the biological sciences, engineering, materials science and digital technologies (Salman & Mohammed, 2024). These knowledge structures are characterized by strong interdisciplinarity, tacitness and ecological coupling, which increase cognitive load and information uncertainty during the innovation process (Salman & Mohammed, 2024). On the one hand, emerging technologies such as advanced breeding platforms, cellular agriculture and controlled-environment agriculture continue to blur traditional industrial boundaries, requiring firms to identify innovation opportunities across a broader range of knowledge domains (N. N. Wang & Cui, 2023). On the other hand, agricultural technological innovation depends heavily on long-term experimentation, contextual validation and experiential accumulation, which encourages firms to maintain sustained and in-depth engagement within selected core knowledge areas (Spanaki et al., 2022). As a result, compared with manufacturing or digitally native industries, agricultural technology enterprises exhibit more heterogeneous boundary-spanning search behaviors in their pursuit of disruptive innovation. These firms must search broadly across diverse external knowledge sources while also maintaining deep engagement in critical domains. As a result, external search becomes central to how they allocate strategic attention (Shao et al., 2024; J. Wang et al., 2020).
Although boundary-spanning search is widely recognized as an important driver of innovation, existing research still pays limited attention to its internal heterogeneity and underlying mechanisms. First, many studies conceptualize boundary-spanning search as a uniform process of external knowledge acquisition (Z. Yang et al., 2024), thereby overlooking differences in search breadth and search depth. From the perspective of the attention-based view (Ocasio, 1997), firms operate under constraints of limited cognitive resources and managerial attention, which lead them to adopt different search strategies that balance broad scanning across multiple domains with focused exploration within specific fields. These strategies are commonly conceptualized as breadth-oriented boundary-spanning search and depth-oriented boundary-spanning search (Katila & Ahuja, 2002; Laursen & Salter, 2006). Prior innovation research has shown that these two forms of search differ fundamentally in knowledge sources, interaction patterns and risk structures (Phelps et al., 2012). However, their distinct roles in the highly interdisciplinary and uncertain context of agricultural technology enterprises have yet to be systematically examined. Second, the acquisition of external knowledge does not automatically translate into innovation outcomes (Laursen & Salter, 2006). This issue is particularly salient in agricultural technology, where knowledge is often fragmented and tightly coupled across technological components (Asiaei et al., 2023). Under such conditions, firms rely heavily on knowledge orchestration capability to integrate heterogeneous knowledge, reconfigure internal processes and enable technological breakthroughs (Rehman et al., 2022). Nevertheless, existing studies provide limited insight into how knowledge orchestration operates as a linking mechanism between boundary-spanning search and disruptive innovation. Third, agricultural technology enterprises differ substantially in organizational inertia, the costs of knowledge renewal and levels of digital infrastructure (Bocean, 2024; Sullivan, 2023). These contextual factors are likely to shape the effectiveness of search strategies, yet they have not been adequately incorporated into prevailing analytical frameworks (Khin & Ho, 2018; Leal-Rodríguez et al., 2019).
To address these theoretical and empirical gaps, this study develops an analytical framework grounded in the attention-based view and conducts an empirical investigation using large-sample survey data from Chinese agricultural technology enterprises. By distinguishing between breadth-oriented boundary-spanning search and depth-oriented boundary-spanning search, the study examines how different forms of boundary-spanning search are translated into disruptive technological innovation through knowledge orchestration capability. In addition, organizational unlearning and digital orientation are incorporated as key contextual conditions to identify how organizational characteristics shape the effectiveness of search behaviors and knowledge integration mechanisms.
This study contributes to the literature in three main ways. First, by distinguishing between breadth-oriented and depth-oriented boundary-spanning search, this study shows how firms allocate attention differently, thereby addressing a notable gap in research on search heterogeneity. Second, by identifying knowledge orchestration capability as a critical mechanism linking boundary-spanning search to disruptive technological innovation, the study deepens understanding of knowledge integration and capability reconfiguration processes in agricultural technology enterprises. Third, by situating the analysis within the technological and institutional context of Chinese agricultural technology firms, the study provides empirical evidence on how boundary-spanning search, organizational capabilities and innovation outcomes interact under conditions of high uncertainty and interdisciplinary complexity, thereby extending the theoretical scope of disruptive innovation research in agricultural systems.
Theoretical Foundation
This study develops its analytical framework primarily on the basis of the attention-based view. According to this perspective, organizational attention is a scarce resource, and the way firms allocate attention shapes the scope and focus of their knowledge search, thereby influencing opportunity recognition and innovation-related outcomes (Ocasio, 1997). In the context of agricultural technological innovation, relevant knowledge sources span multiple domains, including biotechnology, engineering technologies and digital technologies (Sullivan, 2023). Firms are therefore required to allocate attention selectively across heterogeneous knowledge fields rather than attend to all possible technological signals simultaneously (Shao et al., 2024). Different patterns of attention allocation naturally give rise to different search strategies. Some firms emphasize breadth-oriented boundary-spanning search, characterized by broad scanning across multiple knowledge domains and exposure to diverse external sources, whereas others rely more heavily on depth-oriented boundary-spanning search, which involves sustained and focused attention on a limited set of core technological trajectories (Phelps et al., 2012). From this perspective, the attention-based view helps explain why agricultural technology enterprises may adopt different forms of boundary-spanning search under conditions of limited cognitive resources.
However, although the attention-based view explains how firms allocate limited attention across different knowledge domains, it says less about how the knowledge identified through such search is later processed and translated into innovation outcomes (Salman & Mohammed, 2024; Shao et al., 2024). In other words, the attention-based view explains why firms adopt different boundary-spanning search strategies, but it says less about the organizational capability through which attended knowledge is integrated and converted into technological breakthroughs.
To address this issue, this study further draws on the dynamic capability perspective. Dynamic capabilities emphasize firms’ ability to integrate, reconfigure and renew resources in response to environmental change and technological uncertainty (Ribašauskienė et al., 2024). In agricultural technology enterprises, externally acquired knowledge is often heterogeneous, fragmented and strongly context-dependent (Tang et al., 2023). As a result, merely identifying and acquiring such knowledge through boundary-spanning search is insufficient. Firms must also possess the capability to organize, align and recombine heterogeneous knowledge elements so that they can be translated into workable technological configurations and, ultimately, into disruptive technological innovation.
Knowledge orchestration capability links these two theoretical perspectives (Asante et al., 2022). From an attention-based view, firms first allocate attention across diverse external knowledge domains, thereby shaping the breadth and depth of boundary-spanning search. From a dynamic capability perspective, the knowledge obtained through such search must then be structured, integrated and reconfigured before it can contribute to disruptive technological innovation. In this sense, attention allocation helps explain what kinds of external knowledge firms notice and pursue, whereas knowledge orchestration capability explains how that knowledge is transformed into innovation-relevant outcomes. This capability is particularly important in agricultural technology contexts, where innovation frequently depends on combining biological, engineering and digital knowledge under conditions of high uncertainty and long validation cycles.
Building on this logic, this study constructs an analytical framework that treats the attention-based view as the theoretical basis for explaining heterogeneity in boundary-spanning search, while the dynamic capability perspective provides the capability-based explanation for how externally acquired knowledge is mobilized and converted into disruptive technological innovation. Knowledge orchestration capability therefore serves as the critical linking mechanism between attention-driven search behavior and innovation outcomes. This framework helps explain why firms differ in their search behavior and how knowledge integration and recombination connect those differences to disruptive technological innovation.
Literature Review
Boundary-Spanning Search
Boundary-spanning search refers to firms’ efforts to transcend existing knowledge boundaries by acquiring heterogeneous external knowledge in order to compensate for internal limitations and promote innovation (Katila & Ahuja, 2002; Laursen & Salter, 2006). As research has shifted from emphasizing search intensity to examining search structure, scholars have increasingly highlighted the heterogeneity of boundary-spanning search, particularly the ways in which organizations allocate attention across different knowledge domains (Phelps et al., 2012; M. M. Yang & Wang, 2024). From this perspective, boundary-spanning search is commonly conceptualized along two core dimensions. Breadth-oriented boundary-spanning search emphasizes diversified exploration across industries and technological fields, enabling firms to identify emerging opportunities and novel knowledge combinations (Laursen & Salter, 2006; Phelps et al., 2012). In contrast, depth-oriented boundary-spanning search emphasizes sustained engagement and investment within specific knowledge domains, allowing firms to accumulate specialized expertise and reinforce established technological trajectories (Katila & Ahuja, 2002; Sirmon et al., 2011).
Despite these advances, discussion of the internal structure of boundary-spanning search remains incomplete. Existing studies have largely explained search behavior in terms of scope or frequency (Shao et al., 2024), while offering limited mechanistic accounts of how breadth-oriented and depth-oriented boundary-spanning search, as distinct modes of attention allocation, differentially influence technological innovation. Moreover, there is no clear consensus on whether these two forms of search are complementary, substitutive or context dependent under conditions of technological complexity.
Knowledge Orchestration Capability
Knowledge orchestration capability refers to a firm’s ability to integrate, configure and reconfigure knowledge from multiple sources in dynamic environments (Asiaei et al., 2023). It emphasizes the selection, combination and application of heterogeneous external knowledge in order to construct new cognitive structures and technological solutions (Sirmon et al., 2011). Prior studies suggest that this capability can reduce the integration costs associated with knowledge fragmentation and improve the coordination of cross-domain knowledge, thereby supporting complex innovation processes (Salman & Mohammed, 2024).
Although existing research has highlighted the importance of knowledge orchestration (Rehman et al., 2022), its mechanistic role in the innovation process warrants further elaboration. Boundary-spanning search often introduces highly diverse and complex knowledge inputs (Zhu et al., 2024), raising the question of how such knowledge is decoded, integrated and recombined through organizational processes of orchestration. However, systematic theoretical explanations of these mechanisms remain limited. Moreover, whether knowledge orchestration is activated differently under alternative search strategies has received relatively little empirical attention. From a theoretical standpoint, knowledge orchestration capability can be understood as the capability mechanism through which attention-guided boundary-spanning search is translated into innovation-relevant knowledge recombination.
Disruptive Technological Innovation in Agricultural Technology Enterprises
Disruptive technological innovation generally refers to innovations based on new scientific principles or technological trajectories that fundamentally reshape existing production modes, technological systems and industrial structures (Gutmann et al., 2023). In the context of agricultural technology, disruptive innovation often involves the deep integration of interdisciplinary technologies, such as biological breeding, synthetic biology, intelligent agricultural equipment and controlled-environment agriculture (Spanaki et al., 2022). Its defining characteristics include challenging traditional agricultural paradigms, significantly improving resource-use efficiency and reconfiguring agricultural production systems (Spanaki et al., 2022).
Agricultural technology innovation is characterized by high path uncertainty, long experimentation cycles and highly heterogeneous knowledge bases (Yuan et al., 2024). As a result, technological breakthroughs in this sector depend not only on access to external knowledge but also on firms’ capabilities to integrate and coordinate knowledge across disciplinary boundaries (Ribašauskienė et al., 2024). Although prior research has examined agricultural innovation from perspectives such as digitalization (Yuan et al., 2024), organizational learning (Dayan & Benedetto, 2024) and knowledge acquisition (N. N. Wang & Cui, 2023), less attention has been paid to knowledge integration and capability mechanisms. In particular, the process through which technological breakthroughs emerge from the interaction between external knowledge search and internal organizational capabilities has not been fully theorized. Different types of boundary-spanning search may influence disruptive technological innovation through distinct pathways, yet the conditions under which these effects materialize remain insufficiently specified in the existing literature.
Organizational Unlearning
Organizational unlearning refers to the deliberate weakening or abandonment of existing knowledge, routines and cognitive frameworks that are no longer well aligned with environmental conditions, thereby creating space for the absorption and application of new knowledge (Leal-Rodríguez et al., 2019). Prior research suggests that organizational unlearning can reduce knowledge rigidity and path dependence, enhance organizational flexibility in processing heterogeneous knowledge and ultimately strengthen firms’ capacity for technological change (Lyu et al., 2022). In complex innovation contexts, unlearning facilitates the updating of entrenched cognitive structures and improves the efficiency with which firms absorb, match and integrate externally sourced knowledge (Rahardjo et al., 2024).
Despite these insights, the role of organizational unlearning in knowledge integration and innovation processes remains insufficiently explored. In particular, when breadth-oriented and depth-oriented search strategies generate qualitatively different types of external knowledge inputs, how organizational unlearning shapes the activation of knowledge orchestration capability is not yet well understood. Existing studies have paid limited attention to the conditional role of organizational unlearning in shaping disruptive technological innovation, and systematic empirical examinations of its moderating effects remain scarce. Accordingly, examining organizational unlearning from a capability renewal perspective can help clarify how it influences boundary-spanning search and subsequent knowledge integration processes, thereby contributing to a deeper understanding of the internal mechanisms underlying firms’ disruptive technological breakthroughs.
Digital Orientation
Digital orientation refers to a firm’s strategic tendency to adopt and leverage digital technologies to enhance knowledge acquisition, resource allocation and innovation processes (Nasiri et al., 2022). Digital technologies can substantially improve knowledge mobility, information transparency and coordination efficiency, thereby strengthening firms’ knowledge integration capability and responsiveness to innovation opportunities. Prior research suggests that, in dynamic environments, digital orientation enables firms to employ data-driven mechanisms to support knowledge matching, intelligent analysis and innovation-related decision making, ultimately enhancing technological exploration and innovation performance (Arias-Pérez & Vélez-Jaramillo, 2022).
Despite growing evidence of the innovation-enhancing role of digital orientation, its function in complex innovation contexts remains insufficiently specified (J. B. Wang et al., 2024). In particular, when boundary-spanning search generates large volumes of heterogeneous knowledge inputs, how digital orientation shapes firms’ knowledge integration efficiency, technological recombination patterns and innovation trajectories has not been systematically examined. Moreover, although varying levels of digital orientation may condition the extent to which knowledge orchestration capability contributes to disruptive technological innovation, existing studies have paid limited attention to this contingent role.
Hypotheses Development
Boundary-Spanning Search and Disruptive Technological Innovation
Disruptive technological innovation in agricultural technology enterprises often emerges from the recombination of multidisciplinary knowledge and the reconfiguration of established technological paradigms (Spanaki et al., 2022). Such innovation processes depend not only on firms’ access to external knowledge, but also on how they allocate limited organizational attention in complex knowledge environments (Ocasio, 1997). Given the inherent scarcity of attentional resources, firms cannot devote equal attention to all potential technological signals simultaneously. Different patterns of attention allocation are therefore likely to give rise to differentiated boundary-spanning search strategies, which in turn shape innovation-related outcomes (Shao et al., 2024).
From an attention allocation perspective, breadth-oriented boundary-spanning search reflects a strategic choice to distribute organizational attention across multiple technological and knowledge domains (Laursen & Salter, 2006; Shao et al., 2024). Such dispersed attention allocation can expand firms’ cognitive scope and enhance their sensitivity to discontinuous technological signals and emerging knowledge combinations (Li et al., 2024). In the agricultural technology context, technological change often arises at the intersection of the biological sciences, engineering technologies and digital applications, where innovation opportunities are distributed across heterogeneous knowledge systems (Zou et al., 2024). Allocating attention broadly across these domains is likely to be associated with greater recognition of non-obvious innovation opportunities that challenge existing technological paradigms (Tang et al., 2023).
In contrast, depth-oriented boundary-spanning search represents a strategic orientation in which firms concentrate organizational attention on a limited number of key technological trajectories (Katila & Ahuja, 2002; Sirmon et al., 2011). Through sustained and focused attention investment, firms can develop a deeper understanding of complex technological problems and establish more coherent interpretive frameworks for evaluating technological alternatives (Zhu et al., 2024). In agricultural technology innovation, where experimentation cycles are long and technological validation is context dependent, such concentrated attention allocation is likely to be associated with more systematic identification of critical technological bottlenecks and sustained efforts to reconfigure existing technological paths, thereby supporting disruptive innovation outcomes (Sullivan, 2023).
Taken together, breadth-oriented boundary-spanning search and depth-oriented boundary-spanning search reflect distinct attention allocation strategies adopted under conditions of limited cognitive resources. Although they differ in attentional focus and knowledge engagement patterns, both forms of boundary-spanning search are expected to be positively associated with disruptive technological innovation in agricultural technology enterprises through different mechanisms. Accordingly, this study proposes the following hypotheses:
Boundary-Spanning Search and Knowledge Orchestration Capability
For agricultural technology enterprises, knowledge obtained through boundary-spanning search is not only widely dispersed across domains, but also heterogeneous in terms of technological logic, application contexts and validation mechanisms (N. N. Wang & Cui, 2023). As a result, externally sourced knowledge is relevant to later innovation activities only when it receives sustained organizational attention and is effectively processed (Ocasio, 1997; Z. Yang et al., 2025). Accordingly, the implications of boundary-spanning search extend beyond the expansion of knowledge sources to the ways in which firms allocate limited attention to the organization, coordination and internalization of diverse knowledge inputs (Shao et al., 2024).
When agricultural technology enterprises engage in breadth-oriented boundary-spanning search, organizational attention tends to be distributed simultaneously across multiple technological fields and external knowledge sources (Ribašauskienė et al., 2024). Such dispersed attention allocation substantially increases the complexity of information screening and knowledge integration (Asiaei et al., 2023). In agricultural settings, alternative technological solutions often need to be evaluated and matched within specific crop types, ecological conditions and production processes (Sullivan, 2023). This context encourages firms to shift organizational attention away from isolated technological development toward managing interdependencies among heterogeneous knowledge elements (Zou et al., 2024). As a result, breadth-oriented boundary-spanning search is likely to direct greater attention to knowledge orchestration activities and to support firms in coordinating diverse knowledge inputs more systematically.
In contrast, depth-oriented boundary-spanning search reflects a more focused allocation of attention toward a limited set of core technological trajectories (Katila & Ahuja, 2002). This concentrated attention can help firms develop relatively stable cognitive frameworks, allowing newly acquired external knowledge to be continuously monitored, repeatedly validated and embedded within existing technological systems (Gutmann et al., 2023). However, as technological exploration deepens, aligning new and existing knowledge and reconfiguring established technological paths require ongoing organizational attention. Consequently, depth-oriented boundary-spanning search is also likely to be associated with knowledge orchestration capability, as firms rely on more ordered and sustained processes to organize, integrate and recombine knowledge elements.
Taken together, breadth-oriented boundary-spanning search and depth-oriented boundary-spanning search are likely to be positively associated with knowledge orchestration capability, albeit through different attention allocation patterns. Based on the above reasoning, the following hypotheses are proposed:
Knowledge Orchestration Capability and Disruptive Technological Innovation
Disruptive technological innovation generally requires the systematic organization and integration of heterogeneous knowledge elements in order to be translated into viable technological solutions (Spanaki et al., 2022). In agricultural technology enterprises, disruptive innovation often involves the combination of biological processes, engineering principles and digital technologies, resulting in highly dispersed knowledge sources, extended validation cycles and substantial uncertainty in technological trajectories (Spanaki et al., 2022). Under such conditions, the mere availability of external knowledge is insufficient. Firms need the capability to direct limited organizational attention to critical interconnections among diverse knowledge domains and to select, recombine and restructure fragmented knowledge elements. Knowledge orchestration capability underpins this process by enabling firms to organize heterogeneous inputs into more coherent technological configurations, thereby creating conditions under which disruptive innovation may emerge.
Disruptive technological innovation in agricultural contexts is typically accompanied by repeated experimentation and ongoing adjustment (Z. Yang et al., 2025). This process requires firms to sustain attention to evolving knowledge structures over extended periods rather than shift focus prematurely. Knowledge orchestration capability can enhance the coherence of attention allocation by supporting the continuous alignment of emerging knowledge with existing technological frameworks (Rehman et al., 2022). By facilitating the iterative refinement and integration of knowledge elements, this capability allows firms to maintain strategic focus amid technological complexity and uncertainty. As firms allocate attention to optimizing and restructuring core knowledge architectures, they may be better positioned to navigate complex innovation pathways and engage in disruptive technological innovation. Based on the above reasoning, the following hypothesis is proposed:
The Role of Knowledge Orchestration Capability in the Relationship Between Boundary-Spanning Search and Disruptive Technological Innovation
Boundary-spanning search provides firms with opportunities to access novel external knowledge and emerging technologies (Li et al., 2024), yet such knowledge does not automatically translate into disruptive technological innovation. In the agricultural technology context, many externally sourced technologies can demonstrate practical relevance only under specific crop types, ecological conditions and production environments (Zou et al., 2024). When firms acquire external knowledge without sustained attention to its organization, testing and contextual adaptation, that knowledge is more likely to remain conceptual and less likely to be translated into implementable technological solutions or observable innovation outcomes.
From an attention-based perspective, breadth-oriented boundary-spanning search requires agricultural technology enterprises to attend simultaneously to multiple potential technological options. For example, firms may engage in biological breeding research while also exploring digital monitoring systems or intelligent equipment technologies (Sullivan, 2023). This parallel exploration shifts organizational attention away from isolated experimentation within a single technological trajectory toward continuous comparison, coordination and selection across heterogeneous knowledge elements. Only when firms direct sustained attention to the alignment and integration of these diverse technological components can the knowledge acquired through breadth-oriented boundary-spanning search be combined into coherent, system-level agricultural technology solutions.
By contrast, although depth-oriented boundary-spanning search involves a more focused allocation of attention (Phelps et al., 2012), newly introduced external knowledge still needs to be continuously adapted to existing production processes and technological systems. Without such integrative effort, depth-oriented search is more likely to result in incremental refinements rather than substantive restructuring of technological pathways. In this sense, knowledge orchestration is not an auxiliary activity but an essential process through which boundary-spanning search is connected to broader technological reconfiguration.
Accordingly, knowledge orchestration capability reflects agricultural technology enterprises’ capacity to redirect limited organizational attention toward critical knowledge-processing activities following boundary-spanning search. By organizing, coordinating and recombining cross-domain knowledge, firms are more likely to align dispersed technological cues with innovation-related outcomes. This, in turn, is associated with a higher likelihood of disruptive technological innovation. Based on this reasoning, the following hypotheses are proposed:
Organizational Unlearning as an Attentional Condition for Boundary-Spanning Search
In agricultural technology innovation, which relies heavily on long-term experiential accumulation, established technological trajectories, production paradigms and managerial routines provide stability but may also continuously occupy organizational attention (Sullivan, 2023). Such attentional fixation can shape how firms perceive and process newly emerging technological knowledge. When attention remains strongly anchored to prior successful experiences, externally acquired knowledge may receive limited sustained consideration (Salman & Mohammed, 2024), thereby weakening the effectiveness of boundary-spanning search and subsequent knowledge orchestration.
According to the attention-based view, organizational attention is shaped not only by external environmental stimuli but also by entrenched cognitive frames, routines and existing knowledge structures (Ocasio, 1997). In this context, the core function of organizational unlearning is not simply to abandon prior knowledge, but to deliberately weaken reliance on outdated assumptions and rigid cognitive patterns (Sharma & Lenka, 2024). In doing so, organizational unlearning creates room for alternative interpretations and problem-framing approaches. This allows firms to reallocate limited attention more flexibly when processing heterogeneous knowledge generated through boundary-spanning search.
In the context of breadth-oriented boundary-spanning search, agricultural technology enterprises are required to attend simultaneously to technological cues originating from multiple knowledge domains (Shao et al., 2024). Without organizational unlearning, accumulated agricultural experience and mature technological frameworks are likely to serve as default reference points, making new information more likely to be rapidly assimilated into existing schemas or marginalized. Higher levels of organizational unlearning help mitigate path dependence, allowing dispersed attentional allocation to be directed more effectively toward emerging technological signals (Zhang et al., 2024). In turn, this enhances firms’ ability to compare, screen and coordinate multi-source knowledge, thereby strengthening the translation of breadth-oriented boundary-spanning search into knowledge orchestration capability.
In depth-oriented boundary-spanning search contexts, agricultural technology enterprises allocate sustained attention to specific technological trajectories in order to build specialized knowledge bases (Sirmon et al., 2011). Organizational unlearning in this setting does not undermine focused attention itself; rather, it prevents attention from becoming excessively locked into established technological assumptions or predefined solution templates (Dayan & Benedetto, 2024). Higher levels of organizational unlearning help firms maintain reflection and adjustment within core technological pathways and revise critical assumptions when necessary. This, in turn, enhances the flexibility of knowledge orchestration processes and helps prevent depth-oriented search from becoming rigidly path dependent. Based on the above reasoning, the following hypotheses are proposed:
Digital Orientation as an Attentional Condition for Knowledge Orchestration
The attention-based view emphasizes that organizations do not process external information passively, but selectively attend to and interpret it through existing strategic orientations and cognitive frames (Ocasio, 1997). In this sense, strategic orientation not only shapes which issues organizations attend to, but also influences how limited attentional resources are allocated across activities and processes (Khin & Ho, 2018). Digital orientation reflects the extent to which firms strategically emphasize digital technologies and related information, and it shapes organizational patterns of attention allocation (Kindermann et al., 2021). Firms with a stronger digital orientation tend to direct greater attention toward data, algorithms, intelligent systems and information connectivity, thereby enhancing their ability to recognize, compare and process complex technological information. In contrast, firms with a weaker digital orientation are more likely to concentrate attention on isolated technical attributes or established operational routines, which may increase coordination costs during knowledge integration.
Digital orientation does not directly generate new technological knowledge (Nasiri et al., 2022). Rather, by shaping attentional priorities, it influences which aspects of knowledge organization and orchestration receive sustained organizational focus. When firms consistently emphasize digitalization at the strategic level, organizational attention is more likely to be allocated to cross-stage and cross-module information integration, thereby strengthening the role of knowledge orchestration capability in innovation-related processes. For agricultural technology enterprises, such an attentional orientation helps integrate dispersed information from experimental data, production feedback and application contexts, which may support disruptive technological innovation. Based on the above reasoning, the following hypothesis is proposed:
In summary, the conceptual model of this study is shown in Figure 1.
Conceptual model.
Research Design
Sample Selection and Data Sources
This study employs a structured questionnaire survey to examine the associations among boundary-spanning search, knowledge orchestration capability and disruptive technological innovation in agricultural technology enterprises. To ensure the validity and reliability of the survey responses, the target respondents were individuals with managerial or supervisory responsibilities who were directly involved in technological development, R&D coordination or innovation-related decision making within their firms. This sampling strategy was designed to ensure that respondents possessed sufficient familiarity with their firms’ innovation activities.
To enhance sample coverage and response rates, data were collected through multiple channels. First, paper-based questionnaires were distributed to participants enrolled in MBA programs focused on agricultural enterprise management. Importantly, all such respondents were full-time employees of agricultural technology enterprises at the time of the survey, and most held middle- or senior-level managerial positions. The MBA platform served solely as an access channel to practicing managers rather than constituting a student-based sample. Second, electronic questionnaires were distributed via email to managers of agricultural technology enterprises. Third, field surveys were conducted through on-site visits to agricultural technology enterprises located in Shanxi, Shandong, Henan and Jiangsu provinces in China.
The study employed an anonymous and voluntary survey design and did not collect sensitive or personally identifiable information. No intervention, manipulation or potentially harmful procedure was involved, thereby minimizing the risk of harm to participants. All respondents were informed of the academic purpose of the study, assured of anonymity and confidentiality, and provided informed consent before completing the questionnaire. Given the minimal-risk nature of the research design, the absence of identifiable private information and the potential academic and practical value of improving understanding of innovation processes in agricultural technology enterprises, the potential academic and practical benefits of the study were considered to outweigh any foreseeable risks. Formal institutional ethical approval was not required under the applicable institutional and national guidelines.
The sample is geographically concentrated in four provinces that represent major hubs of agricultural technology development in China. These regions provide important contexts for agricultural innovation under conditions of technological transformation and institutional transition. Accordingly, the findings should be interpreted as context-specific empirical evidence rather than as directly generalizable to global or cross-industry settings. In addition, although all respondents possessed practical managerial experience, the use of multiple data collection channels may introduce sample heterogeneity. To address this concern, robustness checks were conducted by controlling for respondents’ background characteristics and data collection channels in the empirical analyses.
Between October 2024 and January 2025, a total of 600 questionnaires were distributed. After excluding incomplete or invalid responses, 412 questionnaires were retained for analysis, yielding a valid response rate of 68.67%. The descriptive statistics reported in Table 1 and all subsequent empirical analyses are based on this final sample.
Descriptive Statistical Results of Basic Enterprise Information.
Source. Questionnaire information.
Variable Measurement
The survey questionnaire comprises two sections: basic information and measurement scales. The basic information section collects data on key enterprise characteristics, including year of establishment, size, ownership type, industry sector and employee position. These variables are incorporated as control variables in the research model. The measurement scale section adopts a seven-point Likert scale, ranging from 1 (“completely disagree”) to 7 (“completely agree”). To ensure the reliability and validity of the measurement scales, the initial questionnaire was developed on the basis of well-established scales from domestic and international studies, with appropriate modifications and refinements to fit the research context. A pre-survey was conducted to further refine the questionnaire items, resulting in the final version.
Disruptive technological innovation was measured using a six-item scale adapted from Spanaki et al. (2022), including statements such as “The enterprise frequently adopts entirely new knowledge and technologies while eliminating outdated knowledge systems.” Boundary-spanning search was measured on the basis of Katila and Ahuja (2002) and Laursen and Salter (2006), who conceptualize it in terms of breadth-oriented boundary-spanning search and depth-oriented boundary-spanning search. Breadth-oriented boundary-spanning search emphasizes scanning across diverse technological domains and external knowledge sources and was assessed with four items, including statements such as “The enterprise can keenly detect market trends and identify emerging customers.” Depth-oriented boundary-spanning search focuses on sustained and intensive engagement within selected core knowledge domains and was also measured with four items, including statements such as “The enterprise maintains close collaboration with core partners to deepen knowledge sharing.” Knowledge orchestration capability was measured using scales from Salman and Mohammed (2024) and Gutmann et al. (2023), covering three dimensions: knowledge acquisition, integration and utilization. This construct was assessed with six items, such as “The enterprise improves its business-related knowledge composition by acquiring or eliminating certain knowledge resources.” Organizational unlearning was measured following Sharma and Lenka (2024) and Rahardjo et al. (2024), using five items, including “The enterprise continuously adjusts outdated routines and processes to adapt to changes.” Digital orientation was assessed on the basis of Nasiri et al. (2022) and Arias-Pérez and Vélez-Jaramillo (2022), using five items, including “The enterprise has a clear understanding of how emerging digital technologies enhance business value.” This measurement approach ensures conceptual clarity and consistency and provides a basis for examining the proposed relationships among the focal constructs.
Measurement Model and Validity
To assess the reliability and validity of the measurement scales, the measurement model was systematically evaluated using SPSS 25.0 and AMOS 24.0. First, internal consistency reliability was examined using Cronbach’s alpha and composite reliability (CR). As reported in Table 2, Cronbach’s alpha values for all latent constructs range from .816 to .902, and all CR values exceed the recommended threshold of 0.80, indicating satisfactory internal consistency and measurement reliability. Second, convergent validity was assessed through standardized factor loadings and average variance extracted (AVE). The results show that most standardized factor loadings exceed 0.70, and all items are above the minimum acceptable level of 0.50. Although a small number of items exhibit factor loadings slightly below the ideal threshold, these items were retained because they are theoretically meaningful and contribute to content validity. Moreover, removing these items did not lead to substantive improvements in scale reliability or overall model performance.
Results of Reliability and Validity Test.
Note. The bold diagonal values in Table 2 represent the square root of the AVE (Average Variance Extracted) for the respective variables.
Confirmatory factor analysis further indicates that the measurement model demonstrates acceptable fit indices (CMIN/DF = 2.356, RMSEA = 0.057, IFI = 0.920, CFI = 0.919), supporting the adequacy of the model’s underlying factor structure. In addition, all AVE values exceed 0.50, providing further evidence of convergent validity. Discriminant validity was first assessed using the Fornell-Larcker criterion, according to which the square root of each construct’s AVE exceeds its correlations with other constructs. To strengthen the assessment of discriminant validity and align with contemporary methodological recommendations, the heterotrait-monotrait ratio (HTMT) was further examined following Henseler et al. (2015). As shown in Table 3, all HTMT values are well below the conservative threshold of 0.85, providing additional support for satisfactory discriminant validity among the constructs.
Heterotrait-Monotrait (HTMT) Ratios.
Note. HTMT ratios below 0.85 indicate satisfactory discriminant validity among constructs. All values reported are below the conservative threshold, suggesting adequate discriminant validity.
Common Method Bias
Given that all variables in this study were collected from the same respondents using a single questionnaire at a single point in time, the potential for common method bias (CMB) cannot be fully ruled out. To mitigate this concern, several procedural remedies were implemented during the questionnaire design and data collection stages. Specifically, respondent anonymity was assured, participants were informed that there were no right or wrong answers, and measurement items for different constructs were interspersed rather than grouped, thereby reducing respondents’ ability to infer the hypothesized relationships.
In addition to these procedural controls, statistical tests were conducted to assess the severity of common method bias. First, Harman’s single-factor test was performed, and the results indicated that no single factor accounted for the majority of the variance. Furthermore, confirmatory factor analysis (CFA) was used to estimate a single-factor model in which all measurement items were constrained to load onto a single latent construct. The fit of this model was substantially worse than that of the proposed multi-factor measurement model, suggesting that common method variance is unlikely to pose a serious threat to the validity of the empirical findings.
Multicollinearity Assessment
Prior to hypothesis testing, potential multicollinearity among the independent variables was examined. The results indicate that variance inflation factor (VIF) values for all explanatory variables across the regression models are well below the commonly accepted threshold of 5, suggesting that multicollinearity is not a serious concern. In addition, pairwise correlation coefficients among the main independent variables were all below 0.70, providing further evidence that multicollinearity is unlikely to bias the regression estimates.
After completing the measurement validation and diagnostic tests, hierarchical regression analysis was employed to test the research hypotheses. Although structural equation modeling is widely used in related studies, regression-based approaches are more transparent for estimating and interpreting specific hypothesized relationships, particularly when multiple interaction effects are involved. Moreover, regression analysis combined with bootstrapping procedures provides robust estimates of indirect and conditional effects, which align closely with the analytical focus of this study.
Results
Descriptive Statistics and Correlation Analysis
This study uses SPSS 25.0 to compute descriptive statistics, including means, standard deviations and Pearson correlation coefficients for all key variables. As shown in Table 4, breadth-oriented boundary-spanning search is positively correlated with disruptive technological innovation (r = 0.606, p < .01), and depth-oriented boundary-spanning search is also positively correlated with disruptive technological innovation (r = 0.273, p < .01). In addition, both breadth-oriented boundary-spanning search (r = 0.386, p < .01) and depth-oriented boundary-spanning search (r = 0.239, p < .01) are positively correlated with knowledge orchestration capability. Knowledge orchestration capability is also positively correlated with disruptive technological innovation (r = 0.423, p < .01). Given the descriptive and cross-sectional nature of the data, these correlations should be interpreted as statistical associations rather than evidence of causal relationships.
Results of Means, Standard Deviations and Correlations.
Note. *represents p < .05. **represents p < .01.
Hypothesis Testing
Main Effects and Explanatory Role Analysis
Table 5 reports the regression results with disruptive technological innovation as the dependent variable. After controlling for respondents’ position, firm age, firm size, ownership type and industry sector, Model 2 shows that breadth-oriented boundary-spanning search is positively associated with disruptive technological innovation (β = .604, p < .001). Model 3 further indicates a significant positive association between depth-oriented boundary-spanning search and disruptive technological innovation (β = .273, p < .001). In addition, Model 4 shows that knowledge orchestration capability is positively associated with disruptive technological innovation (β = .436, p < .001). These results are consistent with Hypotheses H1a, H1b and H3. Across these models, adjusted R2 increases with the inclusion of the focal variables, and variance inflation factor values remain well below conservative thresholds, suggesting that multicollinearity is not a concern.
Regression Results of Boundary-spanning Search, Knowledge Orchestration Capability and Disruptive Technological Innovation.
Note. ***represents p < .001.
Table 6 presents the regression results with knowledge orchestration capability as the dependent variable. Model 8 indicates that breadth-oriented boundary-spanning search is positively associated with knowledge orchestration capability (β = .395, p < .001), while Model 9 shows that depth-oriented boundary-spanning search is also positively associated with knowledge orchestration capability (β = .254, p < .001). These findings support Hypotheses H2a and H2b.
Regression Results of Boundary-spanning Search on Knowledge Orchestration Capability.
Note. ***represents p < .001.
To examine the explanatory role of knowledge orchestration capability, Models 5 and 6 in Table 5 further include knowledge orchestration capability alongside the boundary-spanning search variables. When knowledge orchestration capability is introduced, Model 5 shows that breadth-oriented boundary-spanning search remains positively associated with disruptive technological innovation, although the magnitude of the coefficient is reduced (β = .519, p < .001). Similarly, Model 6 indicates that the association between depth-oriented boundary-spanning search and disruptive technological innovation remains significant but is attenuated after the inclusion of knowledge orchestration capability (β = .174, p < .001).
To avoid relying solely on coefficient comparison logic, a bias-corrected bootstrap procedure with 5,000 resamples was conducted. The results indicate that the indirect effects of both breadth-oriented and depth-oriented boundary-spanning search through knowledge orchestration capability are statistically significant, with 95% confidence intervals that do not include zero. These findings suggest that knowledge orchestration capability constitutes an important explanatory mechanism linking boundary-spanning search to disruptive technological innovation, consistent with Hypotheses H4a and H4b.
Moderation Role Analysis
Table 7 reports the regression results examining the moderating role of organizational unlearning in the relationship between boundary-spanning search and knowledge orchestration capability. After accounting for the main effects, the interaction term between breadth-oriented boundary-spanning search and organizational unlearning is introduced in Model 11 and shows a positive and statistically significant coefficient (β = .104, p < .05). This result indicates that the positive association between breadth-oriented boundary-spanning search and knowledge orchestration capability becomes stronger as the level of organizational unlearning increases, thereby supporting Hypothesis H5a.
Moderating Role of Organizational Unlearning in the Relationship between Boundary-spanning Search and Knowledge Orchestration Capability.
Note. *represents p < .05. ***represents p < .001.
Turning to depth-oriented boundary-spanning search, Model 13 introduces the interaction term between depth-oriented boundary-spanning search and organizational unlearning. The interaction effect is positive and statistically significant (β = .246, p < .001), suggesting that organizational unlearning also strengthens the association between sustained, focused cross-domain search and firms’ ability to organize and integrate heterogeneous knowledge. Accordingly, Hypothesis H5b is supported.
Table 8 presents the regression results examining the moderating role of digital orientation in the relationship between knowledge orchestration capability and disruptive technological innovation. In Model 15, the interaction term between knowledge orchestration capability and digital orientation is positive and statistically significant (β = .095, p < .05). This finding indicates that the association between knowledge orchestration capability and disruptive technological innovation is stronger among firms with higher levels of digital orientation, lending support to Hypothesis H6.
Moderating Role of Digital Orientation in the Relationship between Knowledge Orchestration Capability and Disruptive Technological Innovation.
Note. *represents p < .05. **represents p < .01. ***represents p < .001.
Across the moderation analyses, the inclusion of interaction terms is associated with incremental increases in explained variance, while variance inflation factor values remain well below conservative thresholds, suggesting that multicollinearity is unlikely to bias the estimates. These results indicate that organizational unlearning and digital orientation function as relevant contextual conditions under which the examined associations vary in strength.
To facilitate the interpretation of the interaction effects, moderation plots were generated using simple slopes at one standard deviation above and below the mean of each moderator. As shown in Figures 2 and 3, the positive associations between boundary-spanning search and knowledge orchestration capability are more pronounced at higher levels of organizational unlearning. Specifically, for both breadth-oriented and depth-oriented boundary-spanning search, the slopes are steeper under high organizational unlearning than under low organizational unlearning, indicating a stronger association at higher levels of unlearning. Figure 4 shows a similar pattern for digital orientation. The positive association between knowledge orchestration capability and disruptive technological innovation is steeper when digital orientation is high than when it is low, suggesting that the relationship varies systematically across different levels of digital orientation.
The moderating role of organizational unlearning on the relationship between breadth-oriented boundary-spanning search and enterprise knowledge orchestration capability.
The moderating role of organizational unlearning on the relationship between depth-oriented boundary-spanning search and enterprise knowledge orchestration capability.
The moderating role of digital orientation on the relationship between knowledge orchestration capability and enterprise disruptive technological innovation.
Robustness Test
To further strengthen the credibility of the empirical findings and address concerns regarding contemporary practices in mediation and moderation analysis, a series of bootstrap-based robustness tests were conducted. Following current recommendations in mediation analysis, the robustness of the indirect effects was assessed using bias-corrected bootstrap procedures with 5,000 resamples. As reported in Table 9, the total effect of breadth-oriented boundary-spanning search on disruptive technological innovation is 0.4034, with a 95% confidence interval of [0.3520, 0.4548], indicating a statistically significant association. More importantly, the indirect effect through knowledge orchestration capability is 0.0571, with a 95% confidence interval of [0.0316, 0.0894], which does not include zero. This provides direct statistical evidence of an indirect pathway, independent of changes in the magnitude of the direct effect.
Bootstrap Estimates for Total, Direct and Indirect Associations.
Note. Effect Size (b) represents unstandardized regression coefficients. Bootstrap confidence intervals are bias-corrected and based on 5,000 resamples. An indirect association is considered statistically significant when the 95% bootstrap confidence interval does not include zero.
A similar pattern is observed for depth-oriented boundary-spanning search. The indirect effect through knowledge orchestration capability is 0.0892, with a 95% confidence interval of [0.0447, 0.1490], again excluding zero. Consistent with contemporary mediation practice, these results confirm the existence of statistically meaningful indirect effects without relying solely on coefficient reduction logic, thereby supporting Hypotheses H4a and H4b.
The robustness of the moderation results was further examined using bootstrap confidence intervals for the interaction terms. As shown in Table 10, the interaction between breadth-oriented boundary-spanning search and organizational unlearning is positive and statistically significant (β = .0618, p < .05), with a confidence interval excluding zero. Likewise, the interaction between depth-oriented boundary-spanning search and organizational unlearning is statistically significant (β = .2584, p < .001), providing robust support for Hypotheses H5a and H5b. In addition, the interaction between knowledge orchestration capability and digital orientation is positively associated with disruptive technological innovation (β = .1349, p < .05), thereby supporting Hypothesis H6. The statistical significance and stability of these interaction effects under bootstrap estimation further reinforce the robustness of the moderation findings.
Bootstrap Estimates of Interaction Terms in Moderation Analyses.
Note. Effect Size (b) represents unstandardized regression coefficients. Bootstrap confidence intervals are bias-corrected and based on 5,000 resamples. An interaction effect is considered statistically significant when the 95% confidence interval does not include zero.
To assess the stability of the results across alternative model specifications, additional analyses were conducted by varying the order of variable entry and estimating simplified model structures. Across these specifications, the direction and statistical significance of the key coefficients remained substantively unchanged. Although the cross-sectional nature of the data precludes causal inference, the consistency of the results across multiple analytical approaches helps alleviate concerns about model dependency and potential reverse causality.
Discussion and Findings
This study shows that both forms of boundary-spanning search are positively associated with disruptive technological innovation in agricultural technology enterprises. It also clarifies the roles of knowledge orchestration capability, organizational unlearning and digital orientation in shaping these relationships.
Importantly, these findings should be interpreted in light of the Chinese agricultural innovation context. In recent years, China has developed a policy-supported agricultural innovation system characterized by sustained government intervention, regionally differentiated pilot programs and the accelerated rollout of rural digital infrastructure. Compared with sectors in which innovation is driven primarily by market-based technological competition, agricultural technology innovation in China is more deeply embedded in public support arrangements, including agricultural modernization policies, digital agriculture pilot programs, smart farming demonstration projects and collaborative innovation platforms linking enterprises, universities and research institutes. These institutional arrangements increase firms’ access to heterogeneous external knowledge, lower the barriers to cross-domain collaboration, and create conditions under which external search is more likely to translate into innovation outcomes. In this sense, the Chinese agricultural innovation context helps explain why boundary-spanning search, knowledge orchestration capability and disruptive technological innovation are positively associated in this study.
Prior research has consistently shown that boundary-spanning search supports innovation by enabling firms to access diverse external knowledge (Khanagha et al., 2017; J. Wang et al., 2020; M. M. Yang & Wang, 2024). However, most of this evidence comes from manufacturing or information-intensive industries, where knowledge is relatively modular and technological validation cycles are comparatively short (Li et al., 2024). In contrast, agricultural technology enterprises operate in environments characterized by biological uncertainty, long experimentation periods and fragmented application contexts (Z. Yang et al., 2025). Against this backdrop, the present findings indicate that boundary-spanning search remains positively associated with disruptive technological innovation, suggesting that engagement with external knowledge is also relevant in biologically complex and resource-constrained sectors.
The distinction between breadth-oriented and depth-oriented boundary-spanning search provides a clearer understanding of how firms allocate attention across knowledge domains. Breadth-oriented boundary-spanning search, which reflects broader and more exploratory attention allocation, is positively associated with disruptive technological innovation (Shao et al., 2024). This pattern is consistent with arguments that exposure to diverse knowledge sources enhances firms’ ability to recognize non-obvious technological combinations (Zhu et al., 2024). At the same time, depth-oriented boundary-spanning search, characterized by sustained attention to specific external technological trajectories, is also positively associated with disruptive technological innovation. This pattern suggests that, unlike in many technology-intensive sectors where exploratory breadth tends to dominate, agricultural technology enterprises rely more heavily on sustained, domain-focused external search to accommodate biological variability, long validation cycles and localized production constraints.
The results further suggest that knowledge orchestration capability helps explain how boundary-spanning search is associated with disruptive technological innovation. While existing studies recognize knowledge orchestration as a key organizational capability (Lyu et al., 2022; Rehman et al., 2022), empirical evidence on how it operates in agriculture-related innovation remains limited. The findings suggest that external knowledge acquired through boundary-spanning search does not translate automatically into innovation outcomes. Instead, its relevance appears to depend on firms’ ability to organize, align and recombine heterogeneous knowledge into workable technological configurations. This pattern is consistent with the view that agricultural technologies often require substantial contextual adaptation before firms can turn them into innovation outcomes. The interpretation is particularly relevant in the Chinese context, where policy-supported collaboration and expanding digital infrastructure may further facilitate the integration and recombination of heterogeneous knowledge.
The moderating role of organizational unlearning provides additional insight into the conditions under which boundary-spanning search contributes to knowledge orchestration capability. In contrast to studies that mainly examine organizational unlearning as an internal adjustment process (Dayan & Benedetto, 2024), the findings suggest that unlearning also functions as a contextual condition shaping the effectiveness of boundary-spanning search. Specifically, organizational unlearning strengthens the associations between both forms of boundary-spanning search and knowledge orchestration capability. This suggests that when firms are less constrained by established routines and entrenched cognitive frames, attention allocated to external knowledge is more likely to be redirected toward integrative and coordinative activities. In agricultural technology enterprises, where accumulated experiential knowledge can be deeply embedded (Yuan et al., 2024), the ability to weaken reliance on outdated assumptions may facilitate more effective integration of externally sourced knowledge.
The findings related to digital orientation further indicate that it shapes how knowledge orchestration capability is associated with disruptive technological innovation. While prior studies often conceptualize digital orientation as a strategic driver that directly enhances innovation outcomes (Bocean, 2024), the present results suggest a more nuanced role. Rather than acting as a direct driver, digital orientation appears to function as an attentional condition that influences how firms process and recombine knowledge. In agricultural technology enterprises, where levels of digital maturity may vary considerably, a stronger digital orientation may help firms focus attention on data connectivity, information transparency and cross-stage coordination. In turn, a stronger digital orientation is associated with a stronger linkage between knowledge orchestration activities and innovation-related outcomes. This relationship may be especially salient in China, where uneven rural digital infrastructure means that the effectiveness of knowledge orchestration is likely to vary across regions and firms.
Conclusion
Theoretical Implications
This study contributes to the literature on knowledge management and innovation by clarifying how boundary-spanning search is associated with disruptive technological innovation in agricultural technology enterprises under conditions of biological uncertainty and uneven digital development.
First, by situating disruptive innovation research in the agricultural technology sector, this study extends insights that have largely been derived from manufacturing and information-intensive industries. Rather than assuming high digital maturity or modular knowledge structures, the findings show how innovation-related mechanisms operate in contexts characterized by long experimentation cycles, fragmented application environments and resource constraints. In this sense, the study refines rather than generalizes prevailing innovation theories and highlights the importance of sector-specific boundary conditions.
Second, drawing on the attention-based view, this study differentiates between breadth-oriented and depth-oriented boundary-spanning search. Prior research often treats boundary-spanning search as a unidimensional construct and implicitly emphasizes exploratory breadth. The present findings suggest that, in agricultural technology enterprises, both broad exploratory attention and sustained domain-focused attention are relevant. This nuance extends attention-based arguments by showing how different modes of attention allocation coexist and jointly shape innovation-related outcomes in biologically complex sectors.
Third, the study advances understanding of knowledge orchestration as an explanatory mechanism linking external search to innovation outcomes. Consistent with dynamic capability perspectives, the findings suggest that access to external knowledge alone is insufficient. Innovation-related outcomes depend on firms’ ability to organize, align and recombine heterogeneous knowledge elements in context-specific ways. This insight contributes to the integration of attention-based and dynamic capability perspectives by highlighting knowledge orchestration as a key process through which attention allocation is translated into innovation-related outcomes.
Finally, by examining organizational unlearning and digital orientation as moderating conditions, the study clarifies how internal learning dynamics and strategic orientation shape the effectiveness of knowledge orchestration. Importantly, these effects should be interpreted as contingent rather than universal. The findings suggest that organizational unlearning and digital orientation are associated with stronger manifestations of existing capabilities, but do not substitute for basic organizational resources or institutional support.
Practical Implications
The findings offer several practical insights for agricultural technology enterprises while recognizing heterogeneity in firm size, resource endowment and digital maturity. For early-stage or resource-constrained agri-tech firms, breadth-oriented boundary-spanning search may be particularly useful for identifying emerging technological opportunities and alternative problem-solving approaches. However, given their limited absorptive capacity, such firms should prioritize selective engagement with external knowledge sources and gradually build basic knowledge orchestration routines rather than pursue excessive search breadth. For more established agricultural technology enterprises, depth-oriented boundary-spanning search can support the deep adaptation of external technologies to specific production contexts. These firms may benefit from investing in internal coordination mechanisms, such as cross-functional R&D teams or platform-based collaboration, to strengthen knowledge orchestration capability. Organizational unlearning should be approached as a selective and deliberate process. Rather than abandoning accumulated experience, firms can periodically reassess entrenched routines and assumptions that hinder the integration of new knowledge, especially in areas affected by technological change. With respect to digital orientation, the findings suggest that digital tools are most effective when they support transparency, connectivity and coordination in knowledge-related activities. For firms with lower levels of digital maturity, incremental investments in data integration and information-sharing systems may yield greater returns than large-scale digital transformation initiatives.
Limitations and Future Research Directions
This study has several limitations that should be acknowledged. First, the cross-sectional design limits the ability to establish causal relationships or temporal sequences among boundary-spanning search, knowledge orchestration capability and disruptive technological innovation. Although robustness checks and bootstrap procedures were employed to strengthen the credibility of the findings, future research could adopt longitudinal designs, panel data or event-based approaches to better capture how these processes evolve over time and to address potential reverse causality.
Second, the study relies primarily on self-reported survey data, which may raise concerns about common method bias and perceptual measurement. Although both procedural and statistical remedies were implemented to reduce this risk, future studies could combine survey data with archival indicators, patent data or other objective measures of innovation outcomes to strengthen empirical inference.
Third, the sample was drawn from agricultural technology enterprises operating within the Chinese institutional context. As a result, the observed relationships may be shaped by country-specific cultural, regulatory and policy conditions. Caution is therefore warranted when generalizing the findings to other national or institutional settings. Comparative studies across regions, industries or countries would help clarify the boundary conditions of the proposed relationships. In particular, while this study incorporates the Chinese agricultural innovation context into the interpretation of the findings, the relevant institutional factors were not directly operationalized in the empirical analysis. Future research could examine how policy support systems, regional pilot programs and digital infrastructure development condition the effectiveness of boundary-spanning search and knowledge orchestration across institutional settings.
Finally, while the present study provides quantitative evidence on the associations among external search, knowledge orchestration and disruptive innovation, it offers less insight into the micro-level processes through which these relationships unfold. Future research could therefore complement quantitative analysis with qualitative case studies or mixed-method approaches to examine how firms allocate attention, orchestrate heterogeneous knowledge and engage in organizational unlearning in practice. Experimental or quasi-experimental designs may also help clarify causal mechanisms and improve understanding of how these capabilities interact under different technological and institutional conditions.
Footnotes
Ethical Considerations
This study involved human participants and was conducted in accordance with established ethical standards for social science research. The research employed an anonymous and non-invasive questionnaire survey. No sensitive or personally identifiable information was collected, and no intervention or manipulation was involved. Given the minimal-risk nature of the study design and the absence of identifiable private information, formal ethical approval from a research ethics committee was not required under the applicable institutional and national guidelines.
Consent to Participate
Participation in the study was entirely voluntary. Before completing the questionnaire, all participants were informed of the academic purpose of the research, the anonymous and confidential nature of their responses, and their right to decline or discontinue participation at any time without penalty. Proceeding with the questionnaire was taken as an indication of informed consent.
Author Contributions
All authors contributed to the conceptualization and design of the study. The first draft of the manuscript was written by Yunqi Chen and Liqing Zhou. Material preparation, data collection and analysis were performed by Yichu Wang. Liqing Zhou was responsible for language editing and proofreading.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Humanities and Social Science Fund of the Ministry of Education of China (Project Number: 23YJC630018).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The data presented in this study are available on request from the corresponding author.*