The emergence of artificial intelligence (AI) technologies has compelled many industries including the manufacturing sector to adjust with regard to how they operate through what is termed Industry 4.0. In recent times, traditional industries have been compelled to integrate digital technologies as a tactic to sustain their competitiveness in the market environment. Since the early 1970s, these emerging industries have experienced significant trends aimed at increasing the competitiveness of their production processes (Aydalot & Keeble 2018). Various trends and advancements have emerged over the last few decades, transforming the manufacturing field and shaping how businesses operate today. These developments are being driven by technology, which according to Shai, Bakama and Sukdeo (2020) remains critical to the success of any business, regardless of its size or the products or services it offers. Furthermore, it plays a pivotal role in the success of any organisation by increasing efficiency and enabling innovation, regardless of industry or sector (Gaglio, Kraemer-Mbula & Lorenz 2022).

Technology as a crucial driver of rapid change has revolutionised the way modern business is carried out (Shai et al. 2020). Organisation for Economic Co-operation and Development (OECD 2021) recognises technology as a transformative force to economies worldwide reforming how businesses design, market and sell their goods and services; however, despite the decline in broadband costs and increased internet access via low-cost mobile phones in the mid-to-late 2000s, the digitisation gap between high-income and low-income countries remains wide. This is evident in some developing countries such as South Africa, where the digital divide is widening (OECD 2021).

In the manufacturing sector, Industry 4.0 describes an ongoing process involving industrial automation, digitalisation, and vertical and horizontal value chain integration (Vaidya, Ambad & Bhosle 2018). In the manufacturing sector, the phenomenon aims to improve factory productivity (Maisiri & Van Dyk 2021), improve working conditions (Mayer & Oosthuizen 2021) and stimulate the development of new business models (Oztemel & Gursev 2020). In the assembly lines, technologies such as 3D printing have enabled rapid prototyping, reduced manufacturing lead times and facilitated product customisation (Enrique et al. 2022). Collaborative and autonomous robots improve production efficiency and safety by collaborating with human workers or performing tasks autonomously (Liu et al. 2022). In addition, big data analytics have enabled businesses to extract valuable insights from vast amounts of data, leading to better customer service and better decision-making (Müller, Kiel & Voigt 2018). Wearable devices such as smart glasses and smart gloves are being deployed to improve job performance and provide real-time guidance to workers (Todde et al. 2022). Industry 4.0 reduces unit costs and time (Büchi, Cugno & Castagnoli 2020). It also makes it easier to meet individual customer needs and enables product customisation (Calabrese et al. 2023).

Over the years, the world has evolved towards a digital future, with Industry 4.0 technologies being seen as the way forward (Oztemel & Gursev 2020). The AI is a crucial component of Industry 4.0, driving digital transformation across industries (Trong & Kim 2020). It is one of the most prominent technological concepts alongside blockchain, Internet of Things (IoT) and cloud computing (Reier Forradellas & Garay Gallastegui 2021). The AI can be characterised in its most basic form as a concept that relies on accessible data sets and computational techniques to address emerging problems (Du-Harpur et al. 2020). More explicitly, AI enables a system to understand external inputs accurately, extract key lessons from data and achieve specific goals through flexible changes based on those lessons (Ahmad et al. 2021).

The integration of AI technology is transforming the South African manufacturing industry (Magwentshu et al. 2019). In the aspects such as quality, productivity and workforce, both positive and negative influences have been felt. Although the influence of AI on manufacturing processes, productivity development, quality control and supply chain management (SCM) is recognised worldwide, there are still challenges such as resistance to change and workforce upscaling associated with their integration in the industry (Varshney 2020). Thus, there is a need for solutions that handle these challenges, especially in the South African manufacturing context where there is a lack of insight into its influence (Tjebane, Musonda & Okoro 2022). Furthermore, the limited information available on this topic hinders informed decision-making and strategic planning necessary for successful AI implementation in the industry (Phaladi et al. 2022). The aforementioned disposition on AI integration warrants the need for more studies that delve into exploring the terrain for the optimal benefit of the manufacturing industry. Thus, the rationale for this study enquiry sought to afford practical improvement that would enhance the integration of AI in the manufacturing industry and assist corporate policymaking (Epizitone 2022).

The objective of this study was to analyse the associations among quality, productivity, supply chain management and workforce transformation in the presence of AI in the South African manufacturing industry using multiple correspondence analysis (MCA):

The contributions of the study to the field of knowledge and application are threefold, as outlined below:

The rest of this chapter is concisely organised as follows. Section Literature review provides the literature review. Section ‘Materials and methods’ describes the research methods and materials. Section Result and discussion presents the results of the study and discussion. The article is then concluded in Section Five with a brief remark.

This study aims to gain a thorough perspective for understanding the intricacies of AI implementation within manufacturing companies in South Africa. The study adopted the following three major theories:

The Technology Adoption Theory: This is based on the diffusion of innovation theory and the technology acceptance model (TAM), states that perceived utility, usability, organisational preparedness and outside influences all have an impact on how AI technologies are combined and integrated in manufacturing. (Khan et al. 2023). These conceptual frameworks provide insight into why and how organisations adopt new technologies, thereby enriching the understanding of the AI adoption process in South African manufacturing companies.

The Organisational Change Theory: Deriving from organisational transformation theories such as Lewin’s change management model and Kotter’s eight-step model, emphasises the need to monitor organisational changes that arise from integrating AI (Madanchian & Taherdoost 2022). It emphasises the importance of leadership, communication, employee engagement and educational efforts to facilitate a successful transition (Firican 2023). Understanding these theoretical principles can help decipher the implications of AI for workforce dynamics and organisational ethics in manufacturing facilities in South Africa.

The Resource-Based View (RBV): From the firm sets that organisations accomplish a competitive edge by capitalising on their particular assets and capabilities (Lubis 2022). When connected to the domain of AI adoption, this system proposes that activities prepared with developed innovative framework, gifted workforce and strong authoritative competencies are in a better position to exploit the points of interest of AI in terms of proficiency, quality assurance and administration of the supply chain (Abrokwah-Larbi & Awuku-Larbi 2023). This perspective can enhance our evaluation of the varying implications of AI adoption in the South African manufacturing industry.

Through the integration of viewpoints from these theoretical systems, this study develops a comprehensive understanding of the influence of AI on the manufacturing industry in South Africa. It observes the consequences of joining AI into operations on productivity and measures, the reshaping of workflow, the upgrade of supply chains and the utilisation of AI developments to accomplish a competitive edge.

Artificial intelligence offers several potential avenues for improving business output. For example, machine learning advances have led to inexpensive and improved predictive analytics, complete mechanisation of tasks (such as self-driving cars) and new associations that can be combined to generate new ideas and know-how, improve access to knowledge and generate a number of innovations (Cockburn, Henderson & Stern 2019). At a conceptual level, AI, as demonstrated by Aghion, Antonin and Bunel (2019), is an additional input in a company’s production process that can affect the performance of a company through its impact on the invention of new ideas and technologies and its usefulness in solving complex problems. According to Brynjolfsson, Rock and Syverson (2018), increasing investment in AI technology can improve productivity and other types of factor inputs, so AI ought to be considered as an added immaterial capital in the firm’s manufacturing functions. Effective use of AI technology creates immaterial assets such as data sets, company-specific human skills and the establishment of new business processes. As with any new technology, the productivity benefits of AI technology may not be immediately apparent and may take time to materialise, as organisations will need to adopt a variety of processes and invest in complementary assets to realise the full productivity potential of AI (Brynjolfsson et al. 2018).

Artificial intelligence in the manufacturing industry has led to significant productivity gains in South Africa. Manufacturers have significantly improved operational efficiency, process optimisation and decision-making capabilities through AI technology (Gwagwa et al. 2020). Artificial intelligence-powered automation simplifies repetitive, mundane processes and releases employees to focus on more strategic, value-added activities (IT News Africa 2018). Artificial intelligence enables accurate demand forecasting, inventory management and production planning using predictive analytics and machine learning algorithms, reducing waste and improving consumer-demand responsiveness (TechInsight360 2019). In addition, data analytics based on AI have provided key insights into consumer behaviour and market trends, enabling companies to change their products and marketing strategies to gain a competitive edge (IT News Africa 2018).

Artificial intelligence has become a critical factor in the corporate world, revolutionising and streamlining how companies operate. Quality control is an important aspect that cannot be ignored if a company wants to be successful and meet customer requirements (Villalba-Diez et al. 2019). The AI provides assistance to companies to improve operational efficiency, accuracy, shorter cycle times and better operational compatibility (OECD 2021). The AI has the ability to transform South African manufacturing by improving quality at every stage of production (Ade-Ibijola & Okonkwo 2023). This enables manufacturers to reduce errors, improve productivity and deliver products that meet or exceed customer expectations.

Artificial intelligence has had a substantial impact on quality standards in the South African manufacturing industry. By using AI technology, manufacturers have improved quality control processes, optimised production flows and improved overall product quality (Gaglio et al. 2022). Automated AI-driven quality control systems, including computer vision and image recognition, enable real-time defect detection and prevention, minimising human error and ensuring only high-quality products will reach the market (Trakadas et al. 2020). The AI-driven predictive maintenance systems proactively address equipment failures, reduce downtime and ensure consistent production quality (Gaglio et al. 2022).

In addition, the ability of AI to analyse large data sets and optimise manufacturing processes has increased process efficiency and improved product quality and consistency (Gupta et al. 2022). Real-time quality monitoring by an AI-driven system ensures continuous tracking and intervention in the event of deviations, ensuring compliance with strict quality standards (Park, Phuong & Kumar 2019). In addition, AI-powered insights from customer feedback and market trends have made it easier to optimise product designs and create products that meet customer preferences and expectations (Rathore 2023). As such, the integration of AI into the South African manufacturing industry represents a major advance in improving quality throughout the production lifecycle, driving customer satisfaction and market competitiveness.

Artificial intelligence has a major influence on SCM. The AI gives supply chain managers greater visibility across their systems, allowing them to make better decisions and provide better customer service (Toorajipour et al. 2021). Logistics companies can benefit from the ability of AI in monitoring and predicting package movements at scale (Rahimi & Alemtabriz 2022).

Artificial intelligence in the South African SCM has been integrated to transform operations with the AI technology’s advanced capabilities. Recent studies have shown that introducing AI into SCM has significantly improved efficiency, accuracy and decision-making (Toorajipour et al. 2021). Furthermore, the data on the impact of AI on SCM in South Africa highlight important developments and opportunities (Gwagwa et al. 2020). For instance, Accenture reports that AI technology could boost productivity in the South African logistics sector by up to 30% (Accenture 2019:12). This increased productivity results in reduced costs, optimised processes and increased customer satisfaction.

South African manufacturers are leveraging AI-powered technology to streamline their supply chain operations. For example, by using predictive analytics and machine learning algorithms to forecast demand, manufacturers can better anticipate market needs, reduce inventory costs and minimise production waste (Mhlanga 2023). This enables manufacturers to achieve higher customer satisfaction while maximising operational efficiency.

The introduction of AI in South Africa is bringing about major changes in the workforce, reshaping the employment landscape and requiring a paradigm shift in workers’ skills and capabilities (Magwentshu et al. 2019). Artificial intelligence integration has presented great challenges and abundant opportunities to the South African labour market. On the bright side, AI could automate repetitive tasks, allowing employees to focus on their more strategic and creative aspects and improving job satisfaction and overall productivity (IT News Africa 2018). In addition, AI-powered tools and systems have streamlined work processes, increasing efficiency and productivity across industries (IT News Africa 2018). In addition, the introduction of AI is creating new jobs in areas related to AI development, maintenance and data analysis, providing opportunities for South African workers in areas of high demand (Gwagwa et al. 2020). In addition, the rapid data analysis capabilities of AI enable employees to make more informed decisions, leading to better business and operational outcomes (Principa 2019). However, this change is not without its challenges as AI automates certain tasks, the risk of job losses increases.

Within the framework of AI adoption, it has become apparent that there is a need to manoeuvre around the issue of potential unemployment and provide retraining for new roles. Bridging the skills gap is paramount, workers need to acquire digital literacy, data analysis and critical thinking skills to adapt to the evolving labour market (Varshney 2020). Ensuring inclusiveness and reducing inequalities in AI integration are also pressing concerns, requiring equal access to AI education and technology to prevent certain people from being left behind (Gwagwa et al. 2020). As AI adoption increases in South Africa, ongoing reskilling and upskilling efforts will be essential to preparing the workforce to thrive in an AI-driven economy (Magwentshu et al. 2019). Furthermore, exploiting the potential of AI while adhering to ethical standards requires a careful reflection of ethical considerations around privacy, AI algorithm bias and impact of AI on human rights (Gwagwa et al. 2020). In summary, the adoption of AI in South Africa presents both unprecedented opportunities and daunting challenges, requiring a proactive and comprehensive approach to workforce transformation to realise the benefits of AI while mitigating its potential downsides.

The data in this research article were obtained from qualitative research publications presented in reputable journals and conferences. Studies were selected for their relevance to the research topic and for the purpose of multiple correspondence analysis aimed at examining the impact of AI on multiple parts of an industrial process. The articles were found by searching keywords ‘Impact of Artificial Intelligence on the South African Manufacturing Industry’ on ScienceDirect, Scopus, Springer, Web of science, and Google Scholar. Search results were filtered by publication date and article type (review articles and research articles). Twenty-three articles covering AI and different aspects of the manufacturing process in South Africa (Supply chain management, productivity, workforce transformation, quality control) were selected. Only publication from the year 2018 to the year 2023 were included in the study to ensure that the study incorporates the most recent, relevant and highest quality research, reflecting the latest technological advances, industry practices and policy contexts. As depicted in Table 1, an inclusion and exclusion criterion was formulated as they are important when selecting data from publications as they help to define the quantity and quality of evidence relevant to the research question. It also helps to eliminate biases and confounding factors that could affect the outcome of a study.

The first search using the search parameters returned a total of 964 articles. After applying pre-defined selection criteria (inclusion criteria), 36 articles on Industry 4.0, automation, AI and manufacturing topics were registered. These 36 articles were screened to determine whether they were relevant to the research context. The quality of the research articles was then assessed and 23 articles were identified for MCA.

The data were encoded to assist in categorising and organising the data during collection from peer-reviewed articles based on the variables of quality, productivity, workforce transformation and supply chain management. The data were classified into quality, productivity, workforce transformation and supply chain management. This is illustrated in Table 2.

As shown in Table 2, five aspects of each of our variables were listed, which will be identified from the research articles, which best describe each variable. When conducting secondary data analysis, researchers adhere to ethical principles and guidelines. These include demonstrating respect for the rights and intentions of the original authors, validating the origin of the data, avoiding plagiarism, guaranteeing the confidentiality and anonymity of the data and disclosing any limitations and biases present in the data.

The Statistical Package for the Social Sciences (SPSS) programme is loaded with the data. By allocating numerical quantifications to each variable’s categories, optimal scaling makes it possible to apply standard operating procedures to the quantified variables. Based on the procedure’s optimising criterion, the ideal scale values are assigned. These scale values have metric features, in contrast to the original labels of nominal or ordinal variables. Maximising the spread of categories is the goal of optimal scaling. To get appropriate scaling in SPSS, the following actions were taken:

Select Dimension Reduction under Analyse, then Optimal Scaling, on the resulting window ‘All variables, multiple nominal’ was chosen. Single set was selected and a two-dimensional solution was chosen, and the analysis variables were chosen, consisting of Quality, Productivity, Workforce transformation and Supply chain. Under the options tab, Normalisation method was selected as ‘Variable principal’. Under the Output tab, Discrimination measures, correlation of transformed variables tables were selected (Epizitone & Olugbara 2020).

The goal of MCA is to arrive at a solution where objects in the same category should be drawn closely together, and objects in different categories should be shown widely apart, according to the MCA solution. Every object is as near as feasible to the relevant categories’ category points. When variables group objects into the same subgroups within the same categories, they are said to be homogenous (Olugbara et al. 2021). In order to maximise the overall spread for a one-dimensional solution, MCA allocates appropriate scale values (category quantifications) to each category of each variable. Multiple correspondence analysis discovers a second set of quantifications for a two-dimensional solution that are unrelated to the original set. Additionally, MCA provides scores for things; the averages of the item scores within a category serve as the quantifications for that category (Epizitone & Olugbara 2020).

Table 3 shows the cross-tabulation of all the variables, where each cell represents the frequency of occurrence of the corresponding categories. For example, the cell (S01, Productivity) has a value P1, P3, indicating that Bhagwan and Evans’s study reported efficiency improvements and reduction in waste or rework as the impact of AI on productivity. The cell (S18, Quality) has Q1, Q2, Q5, indicating that Sutherland’ study shows that AI impacts quality through quality control processes, quality assurance measures and quality improvement initiatives. The cell (S08, SCM) has nothing in it, indicating that Matenga Murena and Mpofu’s (2020) study did not report any impact of AI on supply chain management. The cell (S05, Method) has a value of case study, indicating that Daniyar et al.’s study used case study as the research method.

The extracted data were then converted into quantitative data on a scale of 0 (absolutely no impact) to 5 (very high impact), for example, study ID S03; productivity is 2 (low impact), and quantity is 0 (absolutely no impact), SCM is 1 (very low impact) and workforce transformation is 3 (moderate impact).

The data were analysed using MCA. The goal is to minimise the dimensionality of the data by representing rows and columns as points in low-dimensional Euclidean space, with distances reflecting their relatedness. Multiple correspondence analysis can be seen as a generalisation of principal component analysis (PCA) for categorical data or as an extension of correspondence analysis (CA) for more than two categorical variables (Olugbara et al. 2021). Data from the articles is organised into a tabular format and converted into an indicator matrix or Burt table, enabling the application of MCA on the manipulated dataset. An indicator matrix is a binary matrix where each row represents a person, and each column represents a variable category. The Burt table is a symmetric matrix containing all the two-way crosstabs of the variables. The MCA can extract the principal axes that explain the greatest variation in your data and display them graphically as maps of people and categories. These maps can show underlying data structures and patterns, such as groupings of people with similar profiles or correlations between variables in different categories.

Multiple correspondence analysis was conducted using SPSS in order to gain insights into the impact of AI technologies on quality, productivity, workforce transformation and supply chain management practices in the manufacturing sector in South Africa. Based on the MCA study, a two-dimensional MCA solution was determined to be the most appropriate. The first and second dimensions are eigenvalues of 2.773 and 2.619, inertia of 0.693 and 0.655, and Cronbach’s alpha of 0.853 and 0.824, respectively (Table 4). A commonly accepted lower bound for Cronbach’s alpha value is 0.70, although exploratory studies allow lower values. In this case, both alpha values are above the threshold of 0.70 (Hussey et al. 2023).

We employ a heterogeneous configuration to obtain a two-dimensional image of our data, our methodological approaches were performed with this limitation in mind, discrimination measures (Table 5 and Figure 1) display that all the values obtained from dimension one are above the value of ≥ 0.5 (Table 5), and in dimension two only two values were below the value 0.5 that is quality with a distinction of 0.416 and 0.403 (workforce transformation). The most discriminant variables for dimension one hierarchically were workforce transformation (0.848), quality (0.716) and productivity 0.689, with supply chain being the least discriminant at 0.520. The most discriminant variables for dimension two hierarchically were productivity (0.975) and supply chain (0.825) (Table 5). The most discriminant variables in both dimensions are productivity and supply chain management.

In dimension one (Table 6), correlations of transformed variables were as follows: quality is correlated significantly with productivity (r = 0.430, p < 0.001); quality is correlated significantly with workforce transformation (r = 0.586, p < 0.001); quality is correlated with supply chain management (r = 0.317, p < 0.001) and productivity is correlated with workforce transformation (r = 0.380, p < 0.001). Productivity is correlated significantly with SCM (r = 0.488, p < 0.001). Workforce transformation has a low correlation (r = 0.289, p < 0.001) with SCM.

Eigenvalues in a PCA context serve as a representation of the variability present within the data along the principal components. They play a crucial role in indicating the quantity of information that each principal component encapsulates (Karakuzulu et al. 2023). Larger eigenvalues are directly correlated with principal components that hold more significance in the analysis. Furthermore, high eigenvalues are indicative of the fact that the particular principal component is capable of explaining a substantial proportion of the overall data variance (Bao et al. 2022). These eigenvalues, being a fundamental aspect of PCA, hold a strong association with the various dimensions or components within the analysis.

In the realm of PCA, Dimension 1 boasts an eigenvalue of 2.250, which stands out as the dimension that elucidates the highest amount of variance within the data set. On the other hand, the subsequent dimensions possess comparatively lower eigenvalues, such as Dimension 2 with an eigenvalue of 0.844, Dimension 3 with a value of 0.497 and Dimension 4 with a value of 0.409. These eigenvalues serve as crucial indicators shedding light on the extent of variance that each dimension is able to capture. It is noteworthy that higher eigenvalues are indicative of components that hold more significance in the analysis. For instance, in the context of this research article, Dimension 1, focusing on AI adoption, is likely to have a substantial impact on the mentioned variables, whereas the influence of the other dimensions is relatively minor.