2. Challenge 1—Lack of Clarity in Return on Investment
According to Bouanba, Barakat, and Bendou
[10][18], in a study conducted on Moroccan logistics companies, 80% of the sample of companies do not use Artificial Intelligence strategies to improve supply chain performance. One of the causes for this high percentage is the issue with “budgeting,” as not all companies are willing to make a significant investment in AI due to the belief that there will be no return on their investment. As a result, these companies lose competitiveness and face disadvantages in terms of labor costs, time, and money. Deiva, Ganesh, and Kalpana
[11][19] classify and discuss specific challenges of AI for a successful implementation of this technology, and economic challenges are presented as a barrier to adopting this technology. Small and medium-sized enterprises face challenges in embracing the benefits of Artificial Intelligence due to financial constraints. Additionally, an analysis conducted by the International Data Corporation (IDC) reveals that only 30% of organizations have achieved a 90% AI implementation rate, highlighting the low success rate as another significant obstacle to high investment. Supporting this perspective, Li et al.
[12][20] investigate the relationship between Artificial Intelligence (AI), big data (BD), and advanced digital technologies (ADT) and reveal their potential integration in the design and implementation of intelligent energy management systems (SEM). According to the study, AI can be used for intelligent energy management, including energy generation forecasting, demand forecasting, demand-side management (DSM), optimized energy storage operation, energy theft detection, predictive maintenance and control, energy price forecasting, weather-related energy forecasting, and building energy management. However, to achieve these objectives, some challenges are present, such as quantifying the relationship between AI integration and economic benefits.
3. Challenge 2—Organizational Culture
Hradecky et al.
[13][21] conducted a study on the application of Artificial Intelligence in the exhibition sector of the events industry, resulting in the finding that the European exhibition industry is slowly adopting AI, with some factors acting as motivators and inhibitors of this adoption. Among these factors is the organizational culture, where within the organizational dimensions, the support of top management is crucial for making crucial decisions and creating an environment conducive to innovation. However, it becomes evident that there is a lack of vision and progressiveness among CEOs regarding the adoption of AI, as the implementation of this technology is often left out of the strategic plan of many companies. Top management, in many cases, does not see AI as a way to reduce costs and drive strategic advancement, showing insecurity on their part concerning Artificial Intelligence, as they remain distant from this tool. According to the study by Rejeb et al.
[14][22], there is a need to understand AI in the agri-food sector and ensure improvements for this industry, as it is one of the main contributors to the economy of any country and faces challenges such as climate change, unprecedented technological innovation, and increasing demands for sustainability, traceability, and transparency. However, for the implementation of AI applications to achieve productive and strategic benefits, such as task automation, profitability, and improved quality and safety of food, some challenges are listed by the authors. Among the challenges addressed are the organizational barriers to AI adoption, with a perception of reluctance to adopt new technologies in this sector, creating uncertainties about the value of AI in this industry. Wellsandt et al.
[15][23] propose an approach regarding the predictive maintenance and the growing need for the adoption of technologies that automate this process to predict and prescribe maintenance actions. The study suggests the interaction between predictive maintenance systems through an intelligent digital assistant, where this assistant is Artificial Intelligence. However, the adoption of hybrid augmented intelligence in a predictive maintenance system faces challenges, including the need for convincing managers who have budgetary control to understand the benefits of cost and time reduction and increased quality, among others, and who are receptive to this technology and capable of justifying the investment. If this challenge is not overcome, there is a risk of compromising the fulfillment of AI expectations. According to Bouanba, Barakat, and Bendou
[10][18], in their study on Moroccan logistics companies, a significant challenge in managerial choices is the persistence of the traditional management method. According to the authors, it was concluded that most managers are still rooted in traditional managerial strategies in their decision-making processes, ignoring AI technologies in agile innovations to improve supply chain performance, demonstrating that the concept of implementing these technologies is still not popular in most Moroccan logistics companies.
4. Challenge 3—Acceptance of AI by Workers
Ahmad et al.
[16][24] conducted a study on seven different energy fields and their various applications, including electricity production, energy supply, electric distribution networks, energy storage, energy savings, new materials and energy devices, energy efficiency, nanotechnology, and energy policy and economy. Within the context of energy supply, utilities offer the possibility of providing their customers with renewable and affordable electricity while also promoting more efficient energy use among customers. In this scenario, there are many challenges regarding the versatility of developing AI applications and improving the quality of data training for ML algorithms. Among these challenges is the resistance of human workers who are sometimes associated with challenges such as a lack of trust, as the unpredictability in AI performance causes concerns. According to Johnk, Weissert, and Wyrtki
[17][25], employees need to perceive AI as a tool and understand its applications. This context is part of the process of AI acceptance by workers since they can view this technology as an ally in performing their functions. By acquiring adequate knowledge on how to work with AI, employees will know how to utilize the technology to propose solutions within the organization, understand the possibilities of use, where and how it should or should not be applied, and have appropriate expectations regarding the results expected from AI. Wellsandt et al.
[15][23] emphasize the importance of convincing factory floor and office employees to adopt a Hybrid Augmented Intelligence Assistant called DIA (Digital Intelligent Assistant) to obtain organizational benefits. The use of information, advice, recommendations, or actions provided by DIA is essential for employees to benefit from the assistance. However, it is crucial that DIA can enhance the individual tasks of employees, avoiding usability issues or lack of reliability; otherwise, it will not provide significant benefits. Failure to meet employees’ basic needs, such as workplace privacy protection and non-monitoring of individual performance, will result in greater distrust and lower acceptance of DIA by employees. Therefore, to achieve effective hybrid augmented intelligence, measures that increase the assistant’s reliability are necessary.
5. Challenge 4—Data
Baduge et al.
[18][26] study the application of AI in architectural design and this sector covers the entire building lifecycle, from the conceptual phase to its completion. It highlights that AI has great potential, but it is necessary to consider the associated factors and challenges to fully harness its potential. Obtaining a high-quality dataset is a crucial consideration to address the problem at hand. This will allow exploring all the capabilities of algorithms and obtaining effective results in the field of architecture. Deiva Ganesh and Kalpana
[11][19] point out that the absence of adequate standard information in the system hinders the construction of a successful AI framework. Adequate work with data is necessary for the use, inspection, or storage of large volumes of data in AI applications. To perform these tasks, it is essential to have new and efficient technologies. If there is no concern about this data organization structure, the performance of AI applications can be compromised. Javaid et al.
[19][27] divide the study into five categories on the necessary conditions for AI adoption, emphasizing the following data-related needs: data availability, data quality, data accessibility, and data flow. According to the study, data availability is a crucial aspect of training and the effectiveness of Artificial Intelligence (AI) models in generating accurate predictions. Experts emphasized that the nature of data has a significant impact on AI preparation. Structured data, such as those arranged in two-dimensional relational structures, are more convenient for implementing conventional AI systems. In contrast, unstructured data, such as sound, visual, or graphical records, are essential in advanced AI applications, such as object detection. Regarding data quality, it is addressed that for models to function well and be trained effectively, high-quality data are required, and these data need to be suitable for specific use. It is highlighted that organizations sometimes struggle with the quality of historical data, needing to improve data preparation, data processing, and data quality assurance. Concerning data accessibility, it is necessary, according to the study, that this access is fast and practical, suggesting that organizations facilitate this access by centralizing data and allowing access to authorized personnel from various sources so that AI specialists responsible for dealing with AI can easily manage and use this material for their specific purposes. The need for a good data flow is also addressed, which plays a crucial role for AI specialists in transferring data from their source to their appropriate use. When this data flow is automated and smooth, it becomes easier to implement and maintain AI-based systems as they can process data continuously. This continuous approach allows AI systems to be updated and improved over time, providing more accurate and up-to-date results. By ensuring an optimized data flow, it is possible to maximize the potential and effectiveness of AI systems.
Jöhnk, Weissert, and Wyrtki
[17][25] address the implementation of AI in agriculture, where solutions using Artificial Intelligence enable farmers to enhance production efficiency, improve crop quality, and reduce product time to market. The application of AI-based techniques, such as hyperspectral imaging and 3D laser scanning, plays a crucial role in monitoring and maintaining the health of crops
[20][28]. These advanced AI-driven technologies allow for the collection of a large volume of accurate data, providing detailed information on the health status of plantations, and facilitating analysis and decision-making. However, building efficient AI systems depends on the availability of a substantial amount of data for training and accurate predictions. Although spatial data are widely accessible, especially in agricultural areas, obtaining temporal data is more challenging. Most data related to crops are available only once a year during the cultivation period. Additionally, maturing data infrastructure takes time, resulting in a time-consuming process of building robust Machine-Learning models. Soori, Arezoo, and Dastres
[21][7] conducted a study on Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) in robotics applications and highlight the challenges related to data collection. In the field of robotics, the challenge of obtaining large volumes of high-quality data for training AI and ML algorithms is one of the main obstacles faced. Data collection, labeling, and annotation can be time-consuming and costly, and the presence of noise and biases in the data can compromise the accuracy and reliability of models. This challenge is particularly relevant in the field of robotics, where data acquisition can be complex and subject to uncertainties and interferences. Efficient strategies need to be developed to address these difficulties and ensure the quality of data used in AI and ML systems for robotics. In Wellsandt et al.’s
[15][23] study on Hybrid Augmented Intelligence in predictive maintenance with intelligent digital assistants, Conversation-Driven Development is a process in which employees contribute data from real conversations to train an assistant. Over time, the assistant becomes more accurate and capable of predicting correct responses. However, if the solution is not well accepted, there will be fewer conversations and opportunities to train the assistant, resulting in a downward spiral where employees avoid using it, and developers cannot improve its reliability, further reducing its usage.
6. Challenge 5—Data Protection
According to Ahmad et al.
[16][24], intelligent manufacturing involves an interconnected network for the exchange of knowledge between production and machining units. Network communication, mainly through the Internet, requires data authentication and information at various global identity points and encryption. Ensuring the integrity of the system and the process is essential for the development of network structures and smart manufacturing. Additionally, data protection is a significant challenge currently, especially considering the growing role of AI in the energy sector. The energy industry faces vulnerabilities and risks of cyberattacks and data theft across its infrastructure. Before addressing data as a fundamental part of the national economy, it is crucial to ensure cybersecurity as an essential protective measure. Deiva Ganesh and Kalpana
[11][19] point out in their research that systems in AI approaches differ from humans because they cannot understand the nuances of situations and derive appropriate meaning from them. Due to their limitations in understanding the inputs and outputs they process, these systems become vulnerable to unpredictable errors. This barrier, in turn, facilitates cyberattacks and risks that can affect application domains. To prevent the misuse of information systems, it is necessary to implement standard regulations, ethical guidelines, and appropriate policies. According to Hradecky et al.
[13][21], in the study on Artificial Intelligence (AI) and organizational readiness to adopt AI in the exhibition sector of the event industry, regarding data management and privacy, researchers point out that large exhibition spaces are in direct connection with local authorities and policymakers. AI technologies rely on a large amount of data, and policymakers are committed to protecting the privacy of their customers. This collaboration between exhibition venues, authorities, and policymakers is essential to ensure the responsible and safe use of AI technologies. Jöhnk, Weissert, and Wyrtki
[17][25] highlight the inappropriate use of results obtained through data used in AI technologies that can cause irreparable damage. Thus, to avoid unethical AI outcomes, organizations must be aware of biased learning and input data used. It is important to recognize that unconditional reliance on biased AI results can lead to accountability for discrimination, even unintentionally. To enhance the ethical readiness of AI, measures, and protocols must be established to prevent discrimination, thus reducing the associated liability risks. Li et al.
[12][20] emphasize that the deployment of Artificial Intelligence (AI) faces significant challenges, especially regarding data security and cybersecurity. These challenges are particularly relevant when considering the application of AI in various sectors, including the use of AI models in renewable energy sources, such as wind, solar, geothermal, hydropower, oceanic, bioenergy, hydrogen, and hybrid energy. The successful integration of AI in these contexts requires careful measures to ensure data protection and the prevention of cyber threats, aiming to ensure the efficiency and operational safety of these innovative technologies. In Rejeb et al.’s
[14][22] study, it is addressed that data privacy protection is a critical challenge in AI-based agri-food systems due to the absence of adequate privacy standards. The responsibility for decisions made by AI and the need for regulations on data processing and analysis are relevant issues to be addressed. Robust measures and specific regulations are indispensable to ensure privacy protection and promote the trust of those involved in these AI systems.
7. Challenge 6—Skilled Workforce
According to Ahmad et al.
[16][24], AI process automation technologies represent a valuable opportunity for organizations, as they facilitate the workflow of frontline employees. However, it is essential to recognize that significant challenges still need to be overcome. Among these challenges, the need for trained employees capable of dealing with the complexities of Artificial Intelligence stands out. Additionally, the qualification of employees to work effectively with Artificial Intelligence is a crucial aspect to be addressed. Confronting these challenges is fundamental for organizations to fully leverage the benefits provided by AI process automation. According to Deiva Ganesh and Kalpana
[11][19], the widespread adoption of AI technologies in smart infrastructure-based networks presents significant challenges related to network connectivity and monitoring. The development of AI systems requires advanced approaches that improve the interaction between humans and computers and facilitate the flow of information. However, the absence of adequate infrastructure and a skilled workforce has been an obstacle to the effective adoption of this technology in most organizations. Jöhnk, Weissert, and Wyrtki
[17][25] emphasize that employee qualification through upskilling emerges as a fundamental strategy for the effective implementation of AI projects. This process involves acquiring interdisciplinary skill sets, ranging from knowledge in statistics, data management, and analysis to data engineering and expertise in specific domains. Given the scarcity of qualified AI specialists in the job market, investing in employee training becomes an imperative need for organizations. This approach aims to fill the skills gap and promote efficient and sustainable use of AI in organizations.
8. Challenge 7—Infrastructure
According to Ahmad et al.
[16][24], in the context of challenges in energy systems and Artificial Intelligence, the deployment of intelligent production technology faces challenges in synchronizing between old and new equipment, especially due to the incompatibility of communication protocols. To support the new machines in the energy sector, adopting a new protocol to replace the old ones is necessary. This approach allows for coordinated operation between existing and new equipment, resulting in reduced equipment replacement and favoring the economic efficiency of the system. Deiva Ganesh and Kalpana
[11][19] emphasize in their study that the lack of adequate intelligent infrastructure for information acquisition and storage, along with limited network assistance and high investments, has led many organizations to struggle in transforming their practices efficiently. Successful implementation of AI systems requires advanced approaches that enhance the interaction between humans and computers, thus facilitating the flow of information. However, the insufficiency of suitable infrastructure has emerged as a significant challenge in the widespread adoption of this technology by most organizations. According to Hradecky et al.
[13][21], connectivity plays a fundamental role as infrastructure to enable the implementation of AI in the context of the exhibition sector. The adoption of AI in exhibitions is closely related to the technological capabilities that allow for the creation of autonomous networks. Javaid et al.
[19][27] highlight some limitations of using AI in agriculture. For the authors, the lack of integrated and accessible solutions that effectively incorporate AI in agriculture represents a significant obstacle to the widespread adoption of this technology in the sector. Most farmers face time constraints and digital skill limitations to explore AI solutions autonomously. To seamlessly integrate AI into agriculture, it is necessary to integrate these new solutions into the existing systems and legacy infrastructure already used by farmers. This integration is essential to ensure the successful and efficient adoption of AI in agriculture. According to Jöhnk, Weissert, and Wyrtki
[17][25], creating a modular IT infrastructure is essential to enable the integration of new AI applications, as well as to support intensive data training related to AI and testing procedures. In this context, organizations focus on developing three essential resources in IT infrastructure for AI: data storage resources capable of generating and storing large volumes of information, network resources for rapid access, processing, and data transportation, and scalable computing power resources to handle the workload demands of AI. In the study by Li et al.
[12][20], in the context of a Smart Energy Management System (SEM), establishing adequate infrastructure is essential to handle the vast amount of data transmitted in real-time by various connected devices. This infrastructure must be able to efficiently process and analyze the received data to extract relevant information for system monitoring and control. Rejeb et al.
[14][22] conclude in their study that robotic applications often require real-time processing, a task that can be computationally intensive and require specialized hardware. Additionally, to deal with large volumes of data, build models, and make real-time predictions, Artificial Intelligence/Machine Learning/Deep Learning systems require considerable processing power. This need represents a challenge for robotic applications, as robots are limited by energy and computational capacity constraints. Soori, Arezoo, and Dastres
[21][7] point out in their study on the infrastructure of AI deployment in the agri-food sector that there are four categories of obstacles that are considered the main challenges of Artificial Intelligence (AI) in this field. One of these categories is related to AI’s technological constraints, such as connectivity, power supply, bandwidth, security, data validation, and integrity, network latency, response time, flexibility, and the need for big data in AI model training.
9. Challenge 8—Economic Factor
Deiva Ganesh and Kalpana
[11][19] highlight that small and medium-sized organizations face difficulties in adopting advances in Artificial Intelligence (AI) due to the investment factor. Only 30% of organizations have achieved a 90% implementation rate of AI, emphasizing the low success rate as a significant barrier to the required investment. Hradecky et al.
[13][21] note that, in the context of AI adoption, the predominant perception among participants is that larger organizations are seen as faster adopters, with greater potential and financial resources to incorporate new technologies, including AI. These larger companies can absorb the risks and initial costs associated with AI implementation and are known for their technological innovation. On the other hand, some participants observed that smaller companies demonstrate an innovative approach to AI adoption, leveraging their agility, quick responsiveness, and willingness to take risks despite budget constraints. Although larger companies have financial and human resources, they do not always excel in terms of agility and may face challenges related to slowness in decision-making processes. According to Jöhnk, Weissert, and Wyrtki
[17][25], the life cycle of an AI application encompasses the step of adapting AI systems according to an organization’s specific context and data. However, AI adoption is a time-consuming and costly process. Moreover, implementing AI requires organizations to invest in building specialized knowledge and overcoming initial uncertainty regarding AI resources and their value.
10. Challenge 9—Ethical and Social Structures
Jöhnk, Weissert, and Wyrtki
[17][25] address the ethical issues of AI implementation and emphasize that ethics in the application of Artificial Intelligence (AI) encompass the development of new methods to prevent unethical outcomes that may arise from biased learning or distorted input data. It is of paramount importance for organizations to be ethically prepared, as unquestioning trust in biased AI results can lead to accusations of discrimination, even unintentionally. According to Rejeb et al.
[14][22], the implementation of automation and robotics in the agri-food industry has raised social concerns due to the possibility of replacing workers with machines. Despite the high costs of equipment and specialized labor, investments in these technologies can be advantageous due to the reduced workforce required, which offsets the high initial costs. However, the reduction in human intervention may pose significant challenges in terms of employment patterns. According to Soori, Arezoo, and Dastres
[21][7], the use of AI and robotics entails significant ethical and social challenges. Concerns arise regarding the impact of automation on the job market, as well as the potential for AI systems to exhibit bias and reinforce existing inequalities. Moreover, there are apprehensions about the potentially harmful use of robots, such as in military or surveillance applications, raising issues related to privacy and security. These concerns demand a careful approach and regulatory measures to ensure a responsible and ethical implementation of these technologies.
Encyclopedia
