Agricultural farms are extremely pressed to get a rewarding return on their investments which leads to, at times of the year with high workload, farmers not getting much sleep at all.This is confirmed by a farmer who says that since he works so much, some hours are nearly unpaid.Implementing AI in agriculture could potentially mitigate these intense periods of large workloads somewhat, which would give social values back to the farmers.Another dimension of investments and implementation of new technology in agriculture, is that investments in smart farming are not always viewed as necessary by farmers but rather something neat and trendy.Thus, such investments are described to be paid by the “amusement account”.This is confirmed by a farmer that states that most of the technological investments made on his farm are motivated by his interest and fascination with technology.Respondent C4 says that a lot of farmers gladly spend money on new and exciting tools and machines, for instance new tractors.From this, it seems like many farmers think that the charm of running an agricultural business is to be able to tailor and adapt the farm according to one’s liking.While some respondents like doing things very manually others like to develop their way of working consistently with new types of technology.To summarize the results of this interview study, the themes and topics are divided into what appears to be the demands or opportunities for AI in agriculture, as well as the barriers or hurdles that hinder the use of it.Furthermore, based on the contrastive responses and views of different groups of respondents,hydroponic farming the demands and barriers are differentiated by the respondent groups that all have distinct roles in the agricultural sector.
Table 2 shows an overview of the most important points from the interviews, divided over the different respondent groups.To begin with, the responses from farmer respondents show that there are many opportunities linked to the usage of AI and smart farming technologies in agriculture.Most importantly, according to them, new smart farming technologies have the potential of increasing their profitability, either by contributing to higher revenues or freeing time spent on some tedious tasks.On the other hand, the large initial costs to set up the technologies are identified as a barrier.However, if economical means allow for investing in such solutions, farmers believe that the investments will pay off in terms of profitability and competitiveness.Other factors that act as demands for smart farming technologies are their potential to be more sustainable and that they make farming more fun.Further barriers according to farmers are the complex solutions and lack of interoperability, as well as the poor prerequisites and opportunities of continuous education regarding technology in agriculture.Also, the fickle market makes smart farming risky to invest in for farmers.From a commercial enterprise point of view, there are many opportunities connected to smart farming, but also some critical barriers to overcome.The respondents of this group see potential in increased cooperation between companies as well as with farmers, business cases in providing Software as a Service and additionally to streamline logistics connected to agriculture.Nevertheless, data sharing and cybersecurity are seen as large hurdles to the use of these technologies.Respondents from research institutes also express a positive view on accelerated use of AI in agriculture.They believe such a development would result in more data collected by the farmers, which would decrease the time researchers themselves spend on gathering data.This would, according to the researcher respondents, lead to a faster and better research on agriculture.However, data sharing hinders, once again, the scientific development since high-paced research is hard to conduct without proper access to data from different sources.An additional identified barrier for smart farming is the mistrust from farmers that the scientifically developed solutions mirror a real agricultural demand and are not just developed for the sake of technology.
Finally, the respondents from governmental agencies claim that there is a great interest and demand for propagating smart farming technologies for national competitiveness as well as other economic reasons.Still, they are not sure how to position themselves in this transition, which slows down the process of digitizing the agricultural sector.This respondent group also views cybersecurity and data sharing as critical barriers to overcome.This paper provides a review of the main opportunities and hurdles for applying AI to agricultural businesses.By conducting a structured literature review and an interview study with 21 respondents from various parts of the agricultural industry, data has been gathered to get a holistic view on the use of smart technology in agriculture.The scope of the thesis is deliberately wide, focusing on three agricultural sectors: arable farming, milk production and beef production.Furthermore, the respondents are categorized by their role in the sector, ranging from governmental authorities, commercial enterprises, researchers as well as farmers.This broad view allows to acquire knowledge that ranges over several production sectors, as well as over several kinds of organizations with different views on the agricultural sector.Driving the farmers towards smart farming technologies are the needs for increased profitability, reduced workload and often a genuine curiosity for new technology.Surprisingly, all these aspects are not completely captured in the literature review.For example, there are studies about the impact smart farming can have on the relation between humans and animals on a farm, but they did not show in the literature review search.On the contrary, some expected drives for smart farming were not expressed by the respondents, such as the advantageous impact that smart farming can have on the environment through less nutrient loss.Instead, profitability stands out as the most influential factor which makes a clear business case an essential requirement connected to the propagation of smart farming technologies.
Since more and more agricultural products become available in the form of SaaS, allowing for sharing and renting equipment, the business case is changing for both farmers and machine producers, opening new possibilities.Nevertheless, for smart farming to really transform the agricultural sector, governmental agencies and commercial enterprises might need to take a more active role in the transition.Such aspirations are especially important to ensure that the governmental and societal demand for reduced emissions and increased sustainability is met in the technological shift.For the transformation to be successful, it is essential that the structures, allowing farmers to apply the smart farming technologies, are modern.One key requirement is that farmers have continuous and easy ways to acquire up-to-date knowledge of how to apply smart farming.Therefore, there is a need to ensure technical, agricultural education which is easily accessible through for example flexible, on-demand courses.Additionally, the smart farming techniques need to be modifiable to match the varying transparency and adaptability demands that different farmers have.Regarding how implementation and propagation of AI in agriculture might be hindered, this study identifies some factors that act as barriers.The most prominent one is how data is managed, which can be further specified to data sharing and ownership as well as cybersecurity.This is a complex question that as of now does not have a clear solution, neither technically nor legally.Here lays an important role for research institutes as well as authorities.However, there is a consensus among respondents that to transition the agricultural sector into a more data-driven and digital environment, the technical infrastructure must be secure.The solution must be able to guarantee that sensitive data is not available for intruders while at the same time guaranteeing access for the intended users.Furthermore, for the end users to be able to benefit from the digitalizing transition of the sector, the data models require a high degree of flexibility.This stems from the wide variety of machinery at farms as well as the varying level of technological interest and knowledge among the farmers.Moreover, an important aspect that slows down the process of implementing smart farming technologies and AI in agriculture is the economical dimension expressed by the respondents.
A large part of this are of course the high investment costs, but other economic aspects also play a part in this barrier.For example, the fickle market demands, the general low profitability in agriculture as well as the trend towards consolidation of farms all contribute to making investments full of risk.Other identified barriers that hinder the spread of AI in agriculture are some social factors, for example the concerns about technological over-dependency and insufficient end user trust towards technology.The lacking trust seems to stem from over-selling from developers of technology as well as a gap between the technology that is developed and the real market demands.As for the technical solutions that could potentially solve the demand for AI and smart farming technologies, there are many possible ways.In this study, findings show that a lot of the data and sensors types already exist.The problem that remains to be solved is to connect the input data to the output data by developing the datasets, and thereby closing the data cycle.Today, the dairy sector generally holds a closed and elaborate data cycle whereas generally the meat and arable sector have less developed data gathering and therefore less precise decision support tools.This is highlighted in both the interviews and the literature review, as high-resolution data allows for more precise and detailed decision support.Although, after a thorough process of data gathering from input to output, one can build models and evaluate which one of them performs best with some specified evaluating metrics.Additionally, a general problem and difficulty in building machine learning models is that models tend to take too many variables at the same time.The results show the importance of ‘starting small’ when building the models, i.e.using few input variables to begin with and then tune the model adding only one more variable at a time.It is also found that all possible use cases and technical solutions demand a high precision for classification model output as well as low prediction errors for regression models.Decision support in agriculture manages and affects core parts of the agricultural business, and therefore it is important that estimations and predictions are accurate.Interestingly, respondents from the arable sector express that they, as of now,hydroponic equipment accept higher levels of total error in the model.However, for future purposes and solutions with increased complexity, the total error must decrease which is likely to affect the bias- variance trade-off.A requirement for achieving precise supervised machine learning models, adapted to the local farm, will be easy pre-processing of the data.Thus, the data labeling process must either be simplified by developers or offered to the farmers as a service by consultants.
Technologically, the agricultural sector has developed for decades, but the shift towards smart farming techniques and data-driven agriculture might be one of the greatest transitions.Applied AI in agriculture has the potential to optimize and streamline agricultural activities in all sectors in agriculture.By data-driven decision support, and even tasks performed completely automatically, farmers hope to improve their output both in terms of quantity and quality, mitigate carbon emissions, decrease work time, and increase profits.For commercial enterprises and governmental agencies, the transition allows for updated supply chains and planning models, improving the agricultural industry on a macro-level.Still, several challenges remain unsolved, jeopardizing the speed of the transition.Here, there are important tasks for companies, authorities and research institutes.Nevertheless, with such strong incentives, the long-term trend towards increased usage of AI in agriculture is clear.The question is no longer if smart farming will continue to develop, but how the hurdles will be resolved, and which stakeholders will benefit from its radical transformative effects.Growing urban populations and the reduction of arable land, increase the need for productive, efficient, and environmentally friendly ways of agricultural production.For more than three decades, agriculture changed towards an increasing degree of automation.Today numerous digital solutions already exist to support farmers’ everyday lives.Examples can be found within the monitoring of crops and soils as well as for data analysis and storage including decision support.Most solutions today need a connection to farm external cloud systems where data and information are being received from and transferred to.Farmers are motivated to actively use this information technology to benefit from increases in farm input efficiencies, from decreases in negative environmental impacts as well as from automated operation documentation.However, farmers in Europe are diverse in their farm produces and many digital solutions only cover partly the activities within farms.This leads to the problem that farmers experience the lack of interoperability of different digital products.