Top Management Challenges in Using Artificial Intelligence for Sustainable Development Goals: An Exploratory Case Study of an Australian Agribusiness

Abstract

The integration of artificial intelligence into sustainable agriculture holds significant potential to transform traditional agricultural practices. This transformation of agricultural practices through AI directly intersects with several critical sustainable development goals, such as Climate Action (SDG13), Life Below Water (SDG 14), and Life on Land (SDG 15). However, such implementations are fraught with multifaceted challenges. This study explores the technological, organizational, and environmental challenges confronting top management in the agricultural sector utilizing the technological–organizational–environmental framework. As interest in AI-enabled sustainable initiatives continues to rise globally, this exploration is timely and relevant. The study employs an interpretive case study approach, drawing insights from a carbon sequestration project within the agricultural sector where AI technologies have been integrated to support sustainability goals. The findings reveal six key challenges: sustainable policy inconsistency, AI experts lacking farming knowledge, farmers’ resistance to change, limited knowledge and expertise to deploy AI, missing links in the existing system, and transition costs, which often hinder the achievement of long-term sustainability outcomes. This study emphasizes the importance of field realities and cross-disciplinary collaboration to optimize the role of AI in sustainability efforts.

1. Introduction

Agriculture, as the cornerstone of food security and a vital contributor to the global economy, is under increasing pressure to achieve sustainability while meeting the demands of a growing population [1]. In order to ensure long-term profitability without depleting natural resources or causing irreversible ecological damage, the agriculture industry has to maintain a balance between productivity and environmental responsibility [2]. However, agriculture is one of the most significant contributors to environmental degradation, facing challenges such as greenhouse gas emissions, soil degradation, inefficient resource usage, water scarcity, and biodiversity loss. The United Nations predicted in 2005 that the world’s population would increase by 40% by 2050 to 9.1 billion [3]. Incidentally, Tilman, et al. [4] predicted a 100–110% increase in crop demand to cater to the rising per capita consumption due to the population increase. Therefore, with the growing population and demand for food, the negative demand on the environment further increases [5].

It Is found that sustainable agricultural practices play a vital role in minimizing the negative environmental impacts of agriculture, while also contributing to food security [6]. In addition to the environmental benefits, sustainable agricultural practices lower input costs, increasing the economic viability of farming operations [7]. Therefore, sustainable agricultural practices provide contemporary solutions to pressing environmental challenges.

However, implementing sustainable practices remains a complex process, influenced by a combination of organizational, technological, and environmental factors [6]. Therefore, to face these challenges of environmental sensitivity, with the growth of population and increasing demand for food, academics and practitioners have looked into the role of innovative, data-driven technology solutions [8]. For example, such studies have looked into the role of digital technologies, such as SMAC-IoT (social, mobile, analytics, cloud, and Internet of Things), blockchain, and artificial intelligence (AI), to transform the agricultural sector [9]. For example, wireless sensors enable continuous and precise monitoring of crops, allowing the early detection of issues that lead to sustainable practices such as resource optimization [10]. Moreover, image processing enables plant disease identification and plant species identification, operating on an image to extract useful information from it [11]. Among these technologies, AI provides great potential to agribusinesses, where the adoption of AI purports to provide evidence of its capabilities to derive positive global change [12]. The UN-led initiative AI4Good, which emphasizes how AI can be used to address urgent global concerns, is a noteworthy example of AI being acknowledged as a force for good [13].

While the role of technology in sustainability is clear, there are some interrelated critical issues that agribusinesses face regularly when introducing technologies [14]. They include the status quo bias of the CEOs, uncertainties of the weather and soil conditions that preclude conclusive determination of the return-on-investment of technology investments [15], and challenges related to learning and adaptation of employees [16]. As such, the literature highlights that there is a general reluctance in agribusinesses to introduce new technologies [17]. However, with the advent of AI and the value propositions associated with AI, agribusinesses cannot afford not to invest in AI, especially for sustainable engagements. As such, this paper explores the challenges that the top management of the agricultural sector faces in employing AI to achieve sustainability goals. The sustainable development goals (SDGs) that are directly or indirectly influenced by agriculture and the challenges that top management experiences in introducing AI-related technology projects are explored using the interpretive case study approach. Thus, data are gathered using interviews with the top management of a well-established Australian agricultural firm that engaged in an experimental initiative implementing soil carbon sequestration using AI technologies. Moreover, this study adopts the technological–environmental–organizational framework to capture the factors that hinder the integration of sustainable AI solutions. To the best of our knowledge, at the time of submission, no prior studies have specifically examined top management challenges related to AI integration within the context of sustainable agribusinesses.

2. Literature Review

The literature review explores the growing need for sustainable agriculture and the current practices adopted by farmers to promote environmental sustainability in farming. It also emphasizes how technological advancements, especially AI, support and strengthen these sustainability goals by improving resource efficiency, optimizing decision-making and enabling adaptive agricultural practices.

2.1. Sustainable Development Goals and Agriculture

The concept of sustainability is a multifaceted notion that heavily depends on the context. A widely recognized definition by the United Nations World Commission on Environment and Development describes sustainable development as “satisfying the needs of the present without compromising the ability of future generations to meet their own requirements” [18]. At its core, sustainability is about maintaining a balance between economic growth, social well-being, and environmental protection in order to ensure the long-term well-being of humanity [5].

The United Nations initiated 17 Sustainable Development Goals (SDGs) to provide guidance on sustainability [18]. These goals, which are widely accepted, include several salient facets of socio-economic dimensions, including poverty, education, climate change, and prudent resource utilization [18].

According to the Food and Agriculture Organization [19], agriculture is closely linked to several key Sustainable Development Goals (SDGs), including SDG 1 (No Poverty), SDG 2 (Zero Hunger), SDG 12 (Responsible Consumption and Production), SDG 14 (Life Below Water), SDG 15 (Life on Land), SDG 6 (Clean Water and Sanitation), SDG 7 (Affordable and Clean Energy), and SDG 13 (Climate Action). Among these, SDG 13, SDG 14, and SDG 15 are particularly relevant to environmental sustainability [12].

2.1.1. SDG 13: Climate Action

Global agricultural commodities like corn, soybeans, and cotton must be grown on a large scale, which requires large areas of land. According to Streimikis and Baležentis [20], SDG 13 is a crucial goal linked with the agricultural sector, which focuses on climate-smart agricultural practices. One of the most significant environmental challenges faced by humanity is climate change, which is being caused by an increase in greenhouse gases (GHGs) in the atmosphere. According to Food and Agriculture Organization [21], under the current GHG emissions and climate change, crop yields are projected to decline significantly (maize by 20% to 45%, wheat by 5% to 50%, rice by 20% to 30%, and soybean by 30% to 60%) by the year 2100. Climate conditions expose plants to a range of abiotic stresses such as drought, heat stress, cold stress, and salinity [22]. Additionally, climate change has several adverse effects on agriculture, such as water scarcity, soil fertility depletion, and increased pest infestations in crops [23].

To minimize the risks associated with global warming, it is essential to stabilize atmospheric concentrations of carbon dioxide (CO₂) and other greenhouse gases. Carbon sequestration in agricultural soils, which entails absorbing and storing CO₂ from the atmosphere, is one suggested method for reducing the effects of climate change [24]. This approach is considered a crucial mechanism for reducing the impact of climate change while promoting sustainable soil management.

Farmers have embraced several traditional carbon capture techniques, such as crop rotation, cover crops, and conservation tillage [25]. One significant issue in these traditional methods is the high expense associated with absorbing, moving, and storing CO₂, which prevents many firms, particularly those in developing nations, from conducting them economically [25]. Additionally, traditional capture technologies can lower the overall efficiency of industrial operations and are energy-intensive, such as chemical absorption utilizing solvents based on amines [26]. Another issue with these approaches is their scalability since integrating carbon capture devices into existing infrastructure frequently necessitates significant changes [25]. Additionally, there are safety and environmental concerns associated with the long-term storage of captured CO₂, such as the possibility of leaks and geological instability.

2.1.2. SDG 14: Life Below Water

Agricultural wastewater is one of the main causes of worldwide industrial pollution. The surplus water that runs off the field at the bottom of furrows, boundary strips, basins, and flooded regions during surface irrigation is known as agricultural wastewater (AW) [27]. AW may also comprise wastewater from field crops and processed food production, both of which are run by and for farmers, typically in centralized facilities [28]. These plants generate a considerable amount of AW, which is then collected by pipelines and channels that are appropriately constructed and dumped in a storage tank or pond [29]. AW will cause major environmental issues if released into the environment untreated [30].

However, the efficacy of the various methods that have been investigated for the treatment of AW varies. Physicochemical methods are frequently used to remove nutrients, organic contaminants, and persistent materials from AW [31]. The potential of several techniques to eliminate these pollutants has been studied, including coagulation–flocculation, adsorption, ion exchange, precipitation, photocatalytic degradation, membrane filtration, solvent extraction, catalytic oxidation, and electrochemical oxidation [30]. However, a major challenge with many of these techniques is their limited compatibility with the existing infrastructure of wastewater treatment plants [32]. Current approaches still do not fulfil regulatory discharge criteria for this kind of effluent, even if biological treatments offer a potentially good substitute [33].

2.1.3. SDG 15: Life on Land

Soil plays a crucial role in providing plants with a growing environment. For example, increasing soil organic carbon (SOC) enhances vegetation coverage, tillage control, and the efficiency of production input use [34]. However, the increase in soil nutrition could be counterproductive. For several crops, the optimal soil conditions may range greatly in terms of temperature, humidity, nutrients, structure, etc. [35]. For another, although nitrogen (N) is frequently utilized in fertilizers, nitrous oxide (N₂O) from soil, a well-known greenhouse gas, can also contribute to global warming [36].

Moreover, applying large amounts of inputs, like fertilizers and pesticides, to intensify agriculture can harm the environment and affect the health of local livelihoods [37]. In addition, due to deforestation practices, the growth of agriculture poses the second greatest danger to the protection of biodiversity [38]. Aznar-Sánchez et al. [39] stated that approximately 75% of the world’s forests have disappeared, endangering animal populations’ resilience as a result of agricultural development practices. Therefore, with the anticipated rise in future agricultural demand, a major challenge for sustainable land management lies in balancing increased production with environmental conservation.

One possible solution to counteract the long-term soil degradation associated with traditional tillage practices is through conservation agriculture. It helps in the restoration of soil health by encouraging less soil disturbance, increasing crop diversity, and preserving constant soil cover [40]. Additionally, this approach contributes to environmental sustainability by lowering greenhouse gas emissions, decreasing the need for chemical fertilizers, and enhancing the storage of carbon in the soil [40].

However, the agricultural sector is increasingly being transformed through the integration of advanced information and communication technologies, including blockchain technology and the Internet of Things (IoT) [10]. The swift evolution of these technologies has revolutionized numerous industries, including agriculture, leading to a shift from traditional statistical methods to more data-driven, quantitative approaches.

2.2. Artificial Intelligence in Agriculture

Artificial intelligence broadly refers to the development of intelligent systems capable of performing advanced cognitive functions such as reasoning, perception, learning, problem-solving, and making decisions [41]. Moreover, AI has the potential to positively influence all the SDGs through technology advancements that lead to better results, having a major impact on the environment [12]. For example, AI can improve the effectiveness and efficiency of processes like resource optimization, disease detection, and weather prediction [42].

Moreover, AI solutions offer cutting-edge digital technology for improved and more reliable harvesting in addition to increased production efficiency and the creation of a substitute for cement and steel, which collectively account for 9% of global GHG emissions [43]. The agriculture industry is also leading the way in implementing AI to help farmers overcome conventional obstacles and increase crop productivity while reducing negative environmental effects [44].

AI advancements can be utilized in several farming activities to mitigate carbon footprints, which leads to sustainable agriculture practices.

2.2.1. Field Condition Management

AI becomes a potent tool for evaluating how different seed types react to varied soil conditions when paired with sensors, infrared technology, and image recognition. It provides accurate insights that help adhere to weather patterns. To better comprehend ecosystem behavior and identify problems like soil nutrient deficiencies and other faults, AI-driven systems analyze data on temperature, evaporation, and soil moisture [45]. AI also helps with soil health monitoring [46], estimating the nutritional value of soil [47], and forecasting the probability of disease and insect outbreaks [43]. As a result, less water, fertilizer, and pesticides can be used, which lowers the carbon footprint [43].

2.2.2. Fertilizers and Pest and Weed Control

AI-driven robots and drones play a critical role in maintaining soil fertility and improving crop quality by enabling the natural occurrence of important nutrients, like nitrogen, potassium, and phosphorus. in soil [43]. Additionally, it provides efficient remedies for large crop losses caused by inaccurate weedicides and pesticide use, with weeds accounting for up to 90% of the damage [48] and pests causing around 19% [49]. Robots with AI capabilities can precisely detect and eliminate weeds, differentiate them from crops, and lessen the need for agrochemicals and herbicides. This helps solve environmental issues while also reducing operating expenses [45].

2.2.3. Irrigation

By tracking the specific water requirements of crops, AI-powered thermal imaging cameras provide an innovative way to manage water resources effectively [45]. This approach lowers the risk of agricultural diseases and enhances yield. Additionally, AI technology assists in forecasting weather conditions, analyzing evapotranspiration and evaporation rates, and supporting machine learning-based tools for optimizing irrigation practices [43]. It can also predict the daily dew point temperature. By using AI in irrigation systems, carbon emissions are eventually reduced as less fuel is needed to operate the pumps.

2.2.4. Predictive Agricultural Analytics

Sustainable crop management is enabled by AI-powered drones and satellites that can accurately predict weather changes through image analysis. By examining data like solar radiation, wind speed, temperature, and rainfall, these AI tools enable detailed assessments of possible pest and disease outbreaks as well as plant nutrition levels [50]. This technology aids with the timely delivery of fertilizer, tracks crop water requirements, helps decide the best time to plant seeds, and directs tasks like harvesting, baling, and tilling [51]. Along with predicting supply and demand patterns, it also helps farmers select the best crops or varieties for particular climates and productive areas. By promoting automation, AI encourages the transition to precision farming, resulting in increased yields, improved crop quality, and reduced input usage [52].

2.3. Technological–Organizational–Environmental (TOE) Framework

The TOE framework was introduced by Tornatzky and Fleischer [53] to examine the factors that influence an organization’s adoption and use of new technologies. This framework has been widely used in various fields, including innovation diffusion [54], commercialization [55], digital transformation [56] and rural digital development [57]. The TOE framework categorizes factors that influence technology adoption into three dimensions: technological, organizational, and environmental. Technological characteristics include the technological infrastructure, technological capabilities, and so forth. Organizational characteristics include internal attributes like the availability of resources and the structure of the organization. Environmental characteristics include external influences such as regulatory systems and economic conditions.

The integration of technology within socio-environmental and technological contexts is often effectively explained using the TOE framework [58], which has proven successful in understanding the uptake of technologies such as cloud computing [59] and AI [60]. Additionally, the TOE framework has been widely applied in sustainability studies to explore how organizations adopt environmentally responsible technologies and practices [61,62]. Past studies have evaluated the driving mechanisms of carbon reduction pathways for urban areas using the TOE framework and found that technological and organizational factors play a central role in integrating carbon reduction pathways [63]. Moreover, Monye et al. [64] proposed a conceptual framework for Internet of Things (IoT) technology integration in the energy sector. The authors identified energy efficiency, optimizing energy usage and reducing costs as the drivers of successful IoT integration. In addition, Lin and Chen [65] evaluated risks associated with sustainable water resource management and identified statistical analysis as a technological risk and uncertainty and climate change as environmental risks.

In the context of sustainability, the TOE framework has been widely used to evaluate technology adoption in industries such as manufacturing [62], energy [64], and e-commerce [66]. However, its contribution to the agricultural sector is limited [67]. Therefore, this study contributes to the understanding of how TOE dimensions operate in agricultural settings by identifying context-specific challenges.

3. Methodology

The study utilized a qualitative research methodology, which enabled a contextual understanding of the agricultural industry through the direct experiences of industry experts, as outlined by Walsham [68]. An in-depth study of lived exploration was considered to be more insightful than collecting standardized quantitative data from a broad population; hence, this method was selected. The research followed an interpretive approach, meaning that key concepts and themes emerged naturally from the interview data rather than being predefined [69]. The unit of analysis was at the organizational level, and participants were chosen using a purposive sampling method.

3.1. Case Selection

To explore the challenges that the top management of the agricultural sector faces, we selected a multinational agribusiness with headquarters and operational branches spread across several continents, including Africa, the Middle East, the Americas, Asia, and Oceania. By focusing on the agribusiness based in Oceania, the study captured localized insights while still drawing on the broader context of a globally integrated enterprise. This approach enables us to investigate how regional interpretations and implementations of global sustainability goals are made, particularly within Oceania’s distinct environmental, legal, and cultural context. The organization, headquartered in Australia, has committed to becoming carbon-positive, aiming to reach beyond carbon neutrality by actively removing more carbon dioxide from the atmosphere. Since the research aim of this study is to identify top management challenges in integrating AI, this study employed purposive sampling, requiring the selection of participants from an organization that has already integrated or is in the process of integrating AI technologies for sustainable practices.

3.2. Data Collection

We interviewed experts in the agribusiness to gain an understanding of the operational and strategic challenges associated with integrating AI into sustainability practices, as top management commitment is critical to the AI integration process [70]. Participants for the interviews were specifically chosen based on their direct involvement in the carbon-sequestration project, ensuring their insights were relevant to both AI integration and sustainability-focused agricultural initiatives. Data collection continued until thematic saturation was achieved [71]. As a result, we concluded the data collection process with five participants, as shown in

Table 1. Details of the interview participants.

Code	Designation	Contribution in the Carbon-Sequestration Project
P1	Chief Executive Officer	Strategy oversight
P2	Regional Manager	Regional coordination
P3	Chief Information Officer	Technical contributor
P4	Operations Manager	Workflow management
P5	Project Manager	Project lead

. Although the sample size of five participants may be considered relatively small, each individual was purposefully selected due to their direct involvement in the carbon sequestration program.

The interviews were conducted virtually, with each interview lasting approximately 50 to 60 min. To facilitate thoughtful responses, a semi-structured interview guide was created in advance and shared among participants. All interviews were audio-recorded with the participant’s consent and in compliance with human research ethics guidelines. Following each interview, detailed notes were taken, and the recordings were transcribed for analysis. In order to synthesize emerging themes and patterns, the collected data were systematically arranged, coded, and analyzed using NVivo (version 14) software.

3.3. Interpretive Method of Case Analysis

The purpose of using an interpretive case study approach is to collect in-depth insights on barriers and contextual factors associated with AI deployment and sustainability initiatives [72]. Following Klein and Myers’s [73] interpretive research perspective, this study is grounded in the understanding that knowledge is developed and frequently shaped through social constructs such as shared meaning, language, and collective experiences. This perspective acknowledges that participant interpretations and the researcher’s own position influence how reality is understood within the framework of this study, which explores the integration of AI into sustainability initiatives in agriculture. It also acknowledges the close interaction between the researcher and the subject matter, as well as the situational and organizational contexts that shape the way insights are generated throughout the research process.

In alignment with the thematic analysis framework proposed by Braun and Clarke [74], this study started with an in-depth familiarization phase that involved a careful review of the interview transcripts and supportive documents such as annual reports and presentations delivered by the organization. This phase required multiple readings of the transcripts to uncover key ideas and recurring patterns relevant to the research objective, specifically the integration of AI into sustainability initiatives within the agricultural sector. Moreover, to enhance the credibility of the findings, data triangulation was used by drawing on multiple data sources, including interview transcripts, annual reports, project documentation, and data derived from organization’s official website [75]. In addition, to mitigate potential bias, the analysis involved systematically comparing and contrasting the opinions of all interview participants, ensuring an unbiased interpretation of data.

The initial coding was conducted, allowing codes to emerge naturally from the data, without being influenced by predefined categories. Open coding using the NVivo software produced 115 preliminary codes covering areas like government regulations, cost management, employee behavior, and knowledge and expertise. A sample of these codes is shown in Appendix A (

Table A1. Sample codes of top management challenges.

Open Codes	Axial Codes
We can input data to the system, but the system doesn’t adopt quickly. It doesn’t reflect the changing environment in the field today [Open code: Absence of real-time feedback loops]	Missing links in the existing system
The AI system required digital logs. But some of our farms don’t have internet access. So, we have to manually enter all data to the system. That gap made the whole system hard to use [Open code: Disconnected data systems]
The system needs real-time updates. But some of our farms have poor connectivity. So, it fails when we need it most [Open code: Unreliable internet access]
Our IT staff didn’t understand harvesting timings. It’s not always the same. It created confusions during our pilot project [Open code: Knowledge gap between tech and farmers]	Limited knowledge and expertise to deploy AI
… some managers think AI is just automation. They don’t realize it can also support decision-making [Open code: Low awareness of AI capabilities]
… we couldn’t find any training relevant to our region’s crops or soil types… [Open code: Lack of AI programs]
We tried another AI project once. But it failed and now farmers are wary of trying anything new [Open code: Skepticism from previous failures]	Resistance to change
… farmers trust what they have done for years [Open code: Cultural attachment]
… workers worry these systems will replace them [Open code: Fear of job replacement]
Younger farmers are open to try AI, but the older generation don’t trust technology…… [Open code: Generation gap]
… no roadmap from the government [Open code: Lack of sustainable policy implementation guidelines]	Sustainable policy inconsistency
When we adopt to one policy, it changes… [Open code: Frequent policy shift]
Some regions enforce tech standards strictly. Other ignore them [Open code: Uneven policy enforcement]
… that’s where AI experts often fails in developing models [Open code: Failure to capture tacit knowledge]	AI experts lacking farming knowledge
The AI model kept recommending planting dates based on weather data, but it didn’t factor in our harvest festivals [Open code: Misalignment with seasonal realities]
… for small farms, the cost is not manageable [Open code: Resource strain on small farms]	Transition costs
… setting up these systems costs a fortune. We need hardware, software and training all at once [Open code: High upfront investment]

). Following an iterative process of refinement, these codes were first grouped into 6 axial codes. Next, the 6 axial codes were categorized into three themes using the “technological, organizational, and environmental” framework as the sensitizing device. To make sure it appropriately reflected the data and significantly addressed the main study objectives, each theme was thoroughly examined and precisely described [76,77].

Moreover, to ensure our interpretations were based on both empirical insights and theoretical foundations, we consistently compared emerging themes with raw interview data and the existing literature [78]. The six key propositions that are described in the findings were developed as a result of this iterative procedure. The cycle of coding, refining themes, and interpreting results enabled a deeper understanding of the complexities involved in integrating AI for sustainability within the agricultural sector [79].

4. Findings: Challenges in Integrating AI-Driven Sustainable Solutions

The findings of the study indicate that the lack of a clear regulatory environment, limited technical expertise, missing links in the existing system, farmers’ resistance to change, and transition costs are the challenges faced by top management that hinder the successful integration of sustainable AI solutions.

4.1. Sustainable Policy Inconsistencies Across Geographies

The respondents noted that widespread farming operations across multiple states or territories mean that farming operations are subjected to different environmental protection rules, policies and regulations. As such, one of the key challenges highlighted by participants referred to the inconsistency and ambiguity in policy frameworks related to sustainability standards. Strategic alignment becomes challenging due to the dynamic nature of governmental rules and the inconsistency between national and regional directions. P3 stated that:

“There’s no unified framework we can align with. One state supports AI integration and has developed an assurance framework for AI use, but another is still catching up with basic digitization. So, it slows down everything when we are dealing with farmers in several states. And the way sustainability policies keep changing over time and differ across states really slows us down.”

Therefore, this lack of policy stability not only impedes innovation but also delays the transition from pilot projects to full-scale implementation, especially in multinational contexts where operations span across diverse regulatory landscapes. P2 expressed the idea that:

“Policies keep changing, and sometimes what is encouraged today is restricted tomorrow. So, it’s hard to make long-term investments under such uncertain conditions. Even though we started our pilot project, we are still deciding how to deal with the changing policies. So, our main challenge is to understand how AI integration aligns with environmental and data policies in each of our markets”

These findings align with prior studies emphasizing that policy coherence and institutional support are critical enablers of digital innovation in agriculture [80,81]. Therefore, a stable and supportive policy environment that encourages long-term commitment and cross-border collaboration is required for the successful integration of AI for sustainability practices. Thus, the following proposition was developed.

Proposition 1.

Inconsistent and fragmented policy frameworks across regions negatively influence the integration of AI technologies in sustainability-focused agricultural initiatives.

Our findings are consistent with those of White et al. [71], who highlighted the importance of policy consistency in implementing sustainability practices, stating government policies should be developed with the consideration of public support and preferences. Moreover, Lieu et al. [82] stated that policies are frequently developed independently by government bodies, often without adequately accounting for how the proposed policies might interact with the existing broader policy mix. This existing policy mix is made up of numerous policies developed by different local state governmental bodies, which do not deliberately consider how these policies conflict with one another [83].

4.2. AI Experts Lacking Farming Knowledge

Another significant challenge identified by top management is the lack of contextual farming knowledge among AI experts. The operations manager (P4) highlighted that AI vendors overlook the day-to-day realities of farm life, stating that

“Farmers have their own logic, what works, what doesn’t and farming is complex, one-size-fits-all doesn’t work in agriculture. For example, the quantity we store is not the same as what we dispatch because of shrinkage. So, it needs constant observation.” He further stated that,

“It’s not just about feeding data into a model; it’s about understanding the context behind data, why a farmer planted late, and why a field was left fallow. That’s where AI experts often fail in developing models.”

This lack of contextual knowledge among AI experts not only hampers the development of practical solutions but also undermines the trust among farmers and other agricultural workers, leading to high investments. The chief executive officer (P1) also raises this concern, saying that

“You can’t build tech for farmers without understanding farmers. We’ve seen models that look perfect on paper but completely fail in the field because they don’t consider local practices or weather dependencies. If we have to go back and forth because of these oversights, it costs a lot of money.”

Therefore, the following proposition was developed.

Proposition 2.

The lack of farming knowledge among AI experts negatively influences the integration of AI technologies in sustainability-focused agricultural initiatives.

The lack of farming knowledge issue in building sensor-based technologies in livestock management, as highlighted in our findings, is also supported by Devitt’s [84] study. He stated that these sensors are capable of recognizing the resting of cows through motion tracking, but are unable to capture vocal expressions of distress, which is the method that farmers used to use. This limitation may lead to inaccurate assumptions about the animal’s well-being if such behavioral cues are overlooked by automated systems.

4.3. Farmers’ Resistance to Change

Top management also emphasized the farmers’ resistance to change, often rooted in long-standing practices and mistrust of new technologies. The operations manager (P4) reflected on how farmers rooted in various socio-cultural and economic factors are reluctant to integrate AI for AI-driven sustainability practices.

“Many farmers told us, ‘We’ve been doing this for decades, why change now?’ There’s a strong emotional and cultural connection to their current ways of working. There’s still a belief that AI can’t capture the nuances of the land like they can with their own eyes and hands.”

Moreover, top management frequently expressed concerns about farmers’ and field workers’ fear that AI integration might lead to job replacement. While AI promises efficiency and precision in agricultural operations, many workers interpret these innovations as a threat to their livelihoods. As the project manager (P5) explained,

“We are introducing systems that farmers have never seen before. In some scenarios, they think it’s going to replace them entirely, and that leads to pushback. But at the same time, many of them genuinely care about sustainability. They value the land and the environment, and want to leave something better for the next generation. It’s just that they’re cautious about their job loss.”

Therefore, the following proposition was developed.

Proposition 3.

Resistance to change among farmers negatively influences the effectiveness of AI implementation in achieving sustainable agricultural practices.

Our findings align with those of Rose et al. [85], who emphasize that many farmers are reluctant to adopt digital tools due to barriers such as substantial upfront investment costs, limited digital literacy, and uncertainties regarding the long-term sustainability of these technologies. Moreover, farmers who rely on traditional farming techniques perceived digital innovations as a threat to their familiar practices. The widespread integration of AI tools and the potential benefits of smart agriculture may be hampered by this reluctance [86]. Moreover, Ashraf et al. [87] stated that socio-economic and cultural dynamics such as educational level, cultural norms, and religious beliefs significantly shape the willingness of farmers to adopt modern technologies in agriculture. In addition, the country’s rich cultural and linguistic diversity creates additional barriers to digital literacy, as farmers in different regions struggle to access and understand information on advanced agricultural practices, which leads to resistance [88].

4.4. Limited Knowledge and Expertise to Deploy AI

A significant barrier to the effective deployment of AI in sustainable agricultural practices is the limited theoretical and practical knowledge among stakeholders. Since many operators are unfamiliar with the technical language, operational protocols, and infrastructure requirements related to AI systems, this knowledge gap is noticeable at the grassroots level. During the interviews, the chief executive officer (P1) echoed this concern.

“There’s a steep learning curve. We need technical people on the ground, but we don’t have enough of them. Even among us, there’s confusion on how AI fits into our workflow. So, we struggle to find local trainers who can simplify AI concepts without making it sound like science fiction.”

Stakeholders find it challenging to adapt to and maintain AI systems over time due to the limited availability of qualified technical support and insufficient training. Therefore, for AI-driven sustainable agricultural solutions to be adopted and sustained over the long run, it is imperative to bridge this knowledge gap.

Thus, the following proposition was developed.

Proposition 4.

Insufficient internal expertise and a limited of technical capacity within agricultural organizations hinder the deployment and optimization of AI for sustainability.

Our findings align with those of Medvedev and Molodyakov [89], who emphasize that successful implementation of smart farming technologies necessitates a solid foundation in both theoretical and practical knowledge. This highlights that, beyond understanding the algorithmic logic behind these systems, AI experts must also possess a deep awareness of real-world agricultural practices.

4.5. The Missing Links in the Existing System

Another challenge experienced by the top management is the presence of missing links within the existing agricultural systems, hindering the effective deployment of AI solutions. This concern was highlighted by the chief information officer (P3) during the interviews.

“The analytical team uses one system, the field team another, and every state has a different reporting format and data structures. So, we lose a lot of actionable insights in translation. When we try to make decisions based on AI outputs, we often realize there’s a piece missing, like a technical input that didn’t sync.”

This issue was further emphasized by the project manager (P5), who noted that

“There’s a huge gap between what’s happening in the field and what the dashboard shows. We noticed discrepancies in seed weights. The quantity we store is not the same as what we dispatch because of shrinkage. But our systems don’t account for this, so it flags errors or misreports inventory levels. Our AI models need consistent and clean data, but when the input data itself is inconsistent, like the seed weights or missing storage seeds, it affects the reliability of the whole system”

The inconsistency in data accuracy, lack of real-time monitoring capabilities, and lack of reconciliation mechanisms, which are essential for maintaining the integrity of traceability systems, highlight the importance of aligning digital records with on-ground realities to ensure the successful implementation of AI solutions with less transition costs.

Therefore, the following proposition was developed.

Proposition 5.

The lack of system integration and coordination across AI tools, supply chain actors, and operational data systems weakens the success of AI-driven sustainability initiatives.

Several studies are consistent with our findings, emphasizing that the successful integration of AI in sustainable agriculture practices is severely hindered by the lack of suitable technology infrastructure, particularly as these systems remain in the early stages of development [90,91]. Similarly, McCarthy et al. [92] stated that AI-driven agriculture techniques require robust infrastructure that is capable of bridging the digital and physical realms. Limited integration between sensors and digital networks, absence of reliable cloud-based platforms, insufficient standardization of systems across regions, system compatibility, and inadequate automation are examples of specific technical shortcomings [93].

4.6. Transition Costs

The financial implications of switching to AI-driven agricultural systems present a formidable challenge, particularly for small-scale farmers. These transition costs not only include initial capital outlay for hardware and software, but also for continuing expenditures for staff training, data management, and system maintenance. This concern was raised by the chief executive officer (P1) during the interviews.

“It’s hard to convince farmers to make the leap to AI when the transition requires significant capital. Many small-scale producers are already working on tight margins, and the additional costs for training, equipment, and system integration are beyond their reach.”

Moreover, the compatibility of existing infrastructure and the new AI-driven systems is crucial, as infrastructure incompatibility leads to high transition costs.

“Transition from one technology to another has a huge upfront cost. It’s not just about buying the software. It’s about changing entire workflows, retraining staff, and ensuring that the new system works with our existing infrastructure.” added the regional manager (P2).

Therefore, the following proposition was developed.

Proposition 6.

High transition costs, including financial, technical, and operational burdens, act as significant deterrents to AI adoption in sustainable agriculture projects.

The integration of smart controlled environment agriculture technologies often requires a high initial cost and presents a low return on investment in the short run, as highlighted in our findings and also supported by past studies [15].

Through the analysis, it became evident that technological, organizational, and environmental barriers to integrating sustainable AI solutions are deeply interdependent and collectively hinder the successful integration of sustainable AI solutions, as shown in Figure 1.

Technological barriers, such as missing links in the system and limited knowledge and expertise to deploy AI, impede the integration of sustainable AI solutions and contribute to organizational barriers, such as resistance to change and high transition costs. According to Güner and Sneiders [94], organizations with unclear business requirements and the absence of proper technical infrastructure incur more costs in implementing new technologies.

Moreover, the absence of a clear regulatory framework and financial mechanisms to support AI integration leads to high transition costs and farmers’ reluctance to integrate sustainable AI solutions. Srinivasan and Yadav [95] highlighted that the lack of a legal framework for financial activities often hinders the integration of AI systems in the agricultural sector. Moreover, Alabi [96] emphasized that the lack of adequate knowledge and skills can lead to employee resistance in technology adoption.

These interdependencies among the identified barriers are crucial. For example, the lack of technical knowledge (technological) can intensify resistance to change (organizational), which in turn is exacerbated by an inconsistent policy environment (environmental). Therefore, the TOE framework provides a comprehensive lens to understand how these interdependent barriers impede the integration of sustainable AI solutions in the agricultural sector.

Moreover, these interdependencies underscore the necessity of an interdisciplinary approach in interpreting AI’s role in sustainable agriculture. In addition to the technological advancements, achieving SDGs 13, 14, and 15 requires a convergence of agricultural expertise, policy development, and organizational change. For instance, addressing farmer resistance and knowledge gaps demands inputs from behavioral management and rural development, while designing AI tools suited to environmental goals necessitates collaboration between AI developers and ecologists. Therefore, the intersection of AI and SDGs in this study reveals a critical need for cross-disciplinary engagement to convert AI potential to meaningful sustainable outcomes.

5. Conclusions

This study focused on identifying challenges faced by the top management in integrating AI into sustainable agricultural practices. Overcoming these challenges will enable the successful integration of AI into sustainable agriculture, contributing meaningfully to global sustainability efforts, especially in advancing Climate Action (SDG 13), Life Below Water (SDG 14), and Life on Land (SDG 15), all of which are deeply interconnected with agricultural practices and their environmental impacts.

The study utilized the technological–organizational–environmental framework as the theoretical foundation and explored six key challenges associated with AI integration in sustainable agricultural practices, such as sustainable policy inconsistencies across geographies, AI experts lacking farming knowledge, farmers’ resistance to change, limited knowledge and expertise to deploy AI, missing links in the existing system, and transition costs. These findings, which highlight the misalignment between AI experts and farming realities, illustrate the need for an interdisciplinary approach to work together in designing AI tools that directly support SDGs 13, 14, and 15. Moreover, the findings reinforce that sustainable AI adoption in agriculture cannot be achieved through a single-discipline approach but requires interdisciplinary collaboration that bridges technical, organizational, and environmental domains.

The study findings offer several important implications for AI policy and organizational change. The challenges, such as policy inconsistency and AI experts lacking farming knowledge, highlight the need for coordinated national and regional sustainable policies that support consistency in implementation across states and agricultural zones. Therefore, policymakers should consider developing an inclusive assurance framework that accounts for local farming practices.

Moreover, the findings emphasize the necessity of a structural and cultural transformation that includes fostering interdisciplinary collaboration and promoting change management practices, where organizations actively support learning and adaptability.

Limitations and Future Research

The interdependencies among these identified barriers reveal that a holistic approach is essential to effectively address these complexities. While this study provides valuable insights into the key challenges hindering the integration of sustainable AI solutions in the agricultural sector, the interview data alone are not sufficiently robust to comprehensively capture the interdependencies among these challenges. Therefore, to gain deeper insights into the interdependencies among these barriers, total interpretive structural modelling (TISM) can be employed as a supplementary analytical tool. TISM enables researchers to understand the structural relationships among the identified challenges. To strengthen the generalizability and empirical validity of these interdependencies, a quantitative phase could follow the TISM process.