Reviewing the Roles of AI-Integrated Technologies in Sustainable Supply Chain Management: Research Propositions and a Framework for Future Directions

Abstract

In the post-pandemic era, the uncertain global market and rising social-environmental issues drive organizations to adapt their supply chain strategies to more dynamic, flexible models, leveraging advanced technologies like AI, big data analytics, and decision support systems. This review paper aims to examine the current research on AI-integrated technologies in sustainable supply chain management (SSCM) to inform future research directions. We adopted bibliometric and text analysis, targeting 170 articles published between 2004 and 2023 from the Scopus database following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol. We confirm that AI-integrated technologies have demonstrated the capability to enable SSCM across various sectors. We generated ten future research topics using the Latent Dirichlet Allocation (LDA) method and proposed 20 propositions. The results show that AI-integrated technologies in supply chain processes primarily address sustainability, focusing on environmental and economic issues. However, there is still a technological gap in tackling social issues like working conditions and fair dealing. Thus, we proposed a dynamic framework of AI in SSCM to help researchers and practitioners synthesize AI-integrated technologies in SSCM and optimize their supply chain models in future directions.

1. Introduction

The globalization of manufacturing and trade has increased the complexity of supply chains [1]. This complexity increases the challenge of managing the supply chain, and therefore, it requires a better understanding of the challenging issues involved in global supply chains. In recent years, with the increasing awareness of social and environmental pressure caused by the public and governments, optimizing sustainable supply chain performance to reduce the negative effects on the environment, economy, and society has been on the agenda [2]. These adverse impacts are mainly from end-to-end supply chain activities required for sourcing, manufacturing, distribution, and logistics [2,3].

Along with economic development causing severe issues such as the greenhouse effect, abnormal climate, and environmental degradation, reducing carbon emissions has become frequently discussed since industries’ unplanned and irresponsible actions are potential threats to sustainability [4]. The growing size of the carbon trading market and the enlargement of carbon business mechanisms necessitate increased emphasis on sustainable activities involving supply chain management (SCM) to avoid environmental and social threats [5,6]. SSCM should address issues within the supply chain activities from suppliers to customers, not only in operation and market fields [3]. For instance, CO₂ emission has been considered in a green, sustainable supply chain network design [5]. However, these activities and processes require continuous optimization and improvement to create a more responsive and robust supply chain [6]. Recently, artificial intelligence (AI) has been explored in different fields and industries, where significant interest in supporting the broad adoption of AI has been demonstrated [7,8,9], including in SCM. AI, defined as the capability of machines to communicate with and imitate the decision-making of humans, is being used to solve problems with higher accuracy and speed [10]; thus, supply chains powered by AI can independently carry out tasks and make decisions without the need for human involvement [11].

Here, the term “AI-enabled” in systems refers to systems that possess capabilities in one or more of the following areas: problem-solving, knowledge representation, reasoning, planning, learning, perceiving (which includes computer vision), acting (such as in robotics), natural language processing, and communicating [12] (p. 85). This definition emphasizes the range of functionalities that AI can bring to systems, enhancing their ability to perform tasks that typically require human-like cognitive processes. Kassa (2023) [13] identified “prediction, automated reasoning, clustering, decision-making, decision support, and optimization” as the most frequently utilized AI capabilities in AI and supply chain resilience. Trends to achieve sustainability from AI-integrated technologies involve robotics [14], blockchain [15,16,17,18], Internet of Things (IoT) [19,20,21,22], big data analytics [23,24,25,26,27,28,29], decision support system [30,31,32,33,34,35,36,37], etc. In addition to these technologies, AI-related algorithms in SSCM include machine learning [14,38,39,40,41,42], fuzzy logic [42,43,44], the AI-based K-means clustering technique [45], deep reinforcement learning algorithms [45], Bayesian networks [13] XGBoost [39], genetic algorithms [46], artificial neural networks [14,44,47], gene expression programming [39], etc. These AI capabilities and functions empower SSCM to provide information and services by distributing information to make sustainable value [48]. AI capability is also associated with resourced-based theory. Mikalef and Gupta (2021) [49] (p. 2) defined the concept of AI capability, in a context of organizational and resource context, as “the ability of a firm to select, orchestrate, and leverage its AI-specific resources,” which results in increased creativity and performance of a firm. Consequently, AI has also revolutionized business methods for companies’ success [50]. Both AI’s technological and business attributes have made the supply chains more critical and are crucial for SSCM

In this regard, academia has played a crucial role in promoting AI innovation research in fields relevant to supply chains and sustainability. For example, Lee (2023) proposed a sustainable real-time fashion system (RTFS) that facilitates direct trade between customers and companies, effectively eliminating unnecessary intermediate distribution processes, which saves energy in the delivery process [51]. Ebinger and Omondi (2010) explore the existing digital approaches as instruments to facilitate supply chain transparency as an element to enhance SSCM [52]. Their proposed framework builds an ecosystem where AI-based multi-agent systems’ operations tools facilitate the interaction of stakeholders in the supply chains. Some review articles also ascertain the current state of green SCM practices in addressing sustainability issues in the renewable energy sector (e.g., [6,53]). Muthuswamy and M. Ali (2023) [54] uncover the complex interplay between SSCM and machine intelligence, revealing the challenges encountered in integrating machine intelligence with sustainability and opportunities that enhance demand forecasting and inventory management to foster sustainable sourcing practices and reduce waste. Ref. [6] reviews the research trends and proposed propositions in terms of AI in SSCM. These contributions indicate that AI adoption in SSCM is increasing and becoming broader in the scholarly discourse [6,55]. Regarding AI in SSCM, from a business perspective, AI can fulfill different roles within and along a firm’s innovation processes [56]. In this context, comprehensive academic papers have recently been published on AI technology and innovation in business and management, and the management of technology and innovation domains, such as in Technological Forecasting & Social Change, Technovation, Journal of Product Innovation Management, Journal of Cleaner Production, and Journal of Business Research, to name but a few. By reviewing the literature, it is apparent that AI and innovation are focus points in business. This focus is usually accompanied by ecosystem research on the innovation management process within the company [57]. Hence, sustainable ecosystems could be another research direction in AI’s business innovation activities. In similar studies, for example, Naz et al. (2022) [6] conducted structural topic modeling to review AI’s roles in sustainable supply chains (SSCs) and proposed a framework for future directions. However, they focused on propositions for AI addressing sustainable issues in supply chains, rather than holistically classifying each topic of AI in SSCs, and how AI is applied to the procedure to be used in the management of SSCs. Our systematic review explores how AI-integrated technologies address environmental, economic, social, and management issues in SSCM.

This research examined the current state of AI in facilitating SSCM through a bibliometric analysis of 170 journal articles and reviews published from 2004 to 2023, sourced from the Scopus database. After evaluating these research topics and key concepts through topic modeling for text analysis, we constructed the future propositions for building AI-enabled SSCM to inform the following research questions:

RQ 1: What are the current trends concerning applied AI technology and SSCM?

RQ 2: What are the latent topics regarding AI technology and SSCM?

RQ 3: What AI-integrated technologies are involved in sustainable activities to address sustainable issues in supply chain processes?

RQ4: What are future research directions in AI-integrated technologies and SSCM?

To answer these questions, we conducted bibliometric and text analysis to identify critical scholarly trends and emerging topics in the literature. We also examined how research in this area has evolved over the last two decades, providing an agenda for future research.

The remainder of the paper is organized as follows: Section 2 introduces a theoretical background. Section 3 presents the methodology employed for data extraction and review approaches. Section 4 details the bibliometric and text analysis of the selected articles. Section 5 discusses our findings and proposes a research framework to inform the future research agenda.

2. Theoretical Background

2.1. Supply Chains and Supply Chain Management (SCM)

Argumentation involving supply chains and SCM can be traced to the late 1990s. A supply chain was initially defined as the process from raw materials to the end customer, including the members, such as producers, product assemblers, wholesalers, retailer merchants, and transportation companies, involved in the supply chain [58,59]. Mentzer et al. (2001) [60] synthetically reviewed the definitions of the supply chain from tangible (set of firms) and intangible (network of organizations) perspectives. They defined a supply chain as “a set of three or more entities directly involved in the upstream and downstream flows of products, services, finances, and information from a source to a customer” [60]. Kranz (1996) [61] argues that SCM is a delivery process from suppliers’ suppliers to customers’ customers. Moreover, a management philosophy can direct supply chain members to create value and optimize customer satisfaction through innovative solutions. Min et al. (2019) [62] further explained the evolution of SCM based on the research of Mentzer et al. (2001) [60] in their articles. They highlighted the critical market and technological changes that have emerged in SCM. Building on the insights of Min et al. (2019) [62], there is a growing emphasis on technology-driven SCM in academia and the advancement of practice in various fields. Therefore, this paper aims to review the literature surrounding AI-enabled supply chain innovation to build a basis for AI and SSCM, exploring the gaps toward informing the future research agenda.

2.2. Supply Chain Operation Reference (SCOR)

The Supply Chain Operations Reference (SCOR) model, developed by the Supply Chain Council (https://cscmp.org/, accessed on 12 December 2023), acts as a strategic planning tool, enabling senior managers to simplify supply chain intricacies, and is poised to become the standard for advancing the next generation of SCM, as rooted in industrial practices [63]. This step-by-step procedure delineates a unique framework encompassing five processes (plan, source, make, deliver, and return), structured into three hierarchical levels: process types, process categories, and decomposed processes [64]. It serves as an evaluative instrument for SCM, facilitating an understanding of the operational processes within a business entity and pinpointing the key attributes instrumental in achieving customer satisfaction [65]. A supply chain encompasses the management process from the supplier’s suppliers to the customer’s customers, intending to fulfill customer demands. In forming the hierarchical process model, APICS (2017) provides four major sections of the SCOR model Version 12.0: (1) performance, (2) processes, (3) practices, and (4) people; and four levels: level (1) major processes, level (2) process categories, level (3) process elements, and level (4) improvement tools/activities. The term “level” means the span of processes. A level 1 process spans multiple levels, a level 2 determines the capabilities within the level 1 processes, a level 3 process focuses on detailed activity, and a level 4 process develops standard process descriptions of activities within the level 3 processes (APICS, 2017). The SCOR in level 1 offers a standard process approach for communication among supply chain partners, integrating with the six distinct elements—plan, source, make, deliver, return, and enable—to effectively meet customer demands [63,66,67]. Thus, we employed the level 1 process as the theoretical basis to assist scholars in understanding how AI is involved in each process of SSCM.

2.3. Sustainable Supply Chain Management (SSCM)

The concepts of “sustainable” and “sustainability” were first stated in the report “Our Common Future” by the Brundtland Commission (World Commission on Environment and Development) in 1987, highlighting the need for development that meets the needs of the present without compromising the ability of future generations to meet their own needs [68]. This focuses on integrating economic growth, environmental protection, and social equity [69]. Bossel (1999) [70] developed the concept of sustainability. Ref. [70] argues that it is a “dynamic concept”, and therefore defined sustainable development as evolving societies, environments, technologies, cultures, and values; a sustainable society should support and enable these changes, allowing for ongoing, viable, and energetic growth. A sustainable organization, therefore, should integrate stakeholder perspectives, ensuring that decisions address necessities, ecological integrity, resource efficiency, and community empowerment while integrating environmental, economic, and social objectives [71]. Carter and Rogers (2008) [72] align the conceptual frameworks of sustainable development with SCM to ensure that sustainability goals are integrated into supply chain strategies. Accordingly, SSCM was defined as “the strategic, transparent integration and achievement of an organization’s social, environmental, and economic goals in the systemic coordination of key inter-organizational business processes for improving the long-term economic performance of the individual company and its supply chains” [72]. This definition focuses on the dimensions of risk management, transparency, culture, and strategy and the interrelationships among them [72], while Seuring and Müller (2008) [73] highlight SSCM as the resource integrations and cooperation among all stakeholders in the SCM process for sustainable development’s three dimensions, environmental, social, and economic, to satisfy stakeholders’ requirements. This aligns with the definition that concerns the ecological and social systems influenced by a focal firm’s internal and supply management operations [74].

3. Methodology

We adopted bibliometric analysis of the 170 selected studies to achieve the first objective. Then, these studies were systematically reviewed through topic modeling to define future topics regarding this field of study, which achieved the second objective. Based on the topics of future research, we summarized AI-integrated technologies in SSCM processes among the selected studies. We proposed a framework for building innovation ecosystems, which addressed the third objective.

3.1. Inclusion and Exclusion Criteria

Inclusion and exclusion criteria assist researchers in selecting relevant articles for their reviews [75]. All selected articles were articles and reviews because empirical articles contribute to theory building and grounding [76], and review papers synthesize and analyze a large number of research articles, offering a comprehensive overview, background knowledge of a particular field, and future directions [75]. Given the breadth of our research, conference papers were excluded [77], and book chapters, conference reviews, books, editorials, and short surveys were excluded. Additionally, the searched databases consist exclusively of journals published in English between 2004 and 2023 in online academic journals (see

Table 1. The inclusion and exclusion criteria.

Inclusion Criteria	Exclusion Criteria
Journal articles and reviews.	Reports, white papers, conference articles, and chapters in books.
Subjects: engineering, business, management, and accounting; social sciences; decision sciences; economics, econometrics, and finance; arts and humanities; multidisciplinary.	Computer sciences; environmental science; mathematics; agricultural and biological sciences; energy; chemical engineering; physics and astronomy; biochemistry, genetics, and molecular biology; materials sciences; chemistry; psychology; earth and planetary sciences; medicine; immunology and microbiology; veterinary; pharmacology, toxicology and pharmaceutics.
Published from 2004 January to 2023 December.	Outside the selected time period.
Written in English.	Non-English papers.

).

3.2. Data Sampling Process

To execute the search process efficiently, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol was used [77,78]. The PRISMA protocol has four phases: identification, screening, eligibility, and inclusion. Figure 1 shows the details for each stage.

In the identification phase, we searched the Scopus database using keywords such as artificial intelligence, supply chain, and sustainable supply chain management. We also used Boolean operators to refine and design search strings, such as artificial_intelligence AND sustainable AND supply_chain OR supply_chain management. The Boolean retrieval model for information generates queries using Boolean expressions, where terms are combined using the operators AND, OR, and NOT [79].

Table 2. Keyword-based retrieval criteria.

Search Phase	Keyword and Search Strings
Step One-Keyword Search	Artificial_intelligence, sustainable supply_chain management
Step Two-Search String	“Artificial_intelligence” OR “blockchain” OR “machine_ learning” OR “digitalization” OR “big data analytics” OR “big data” OR “digital technologies” OR “artificial_neural_ network” OR “blockchain technology” OR “artificial_intelligence research” AND “supply_chain”

shows the keyword-based retrieval criteria. The first step is a keyword search performed on the title, abstract, and keyword sections of articles, and the second step is to search strings where a list of keywords is used in this phase. Then, the search results were screened, and related studies were identified based on the inclusion criteria (see Figure 1). A total of 519 hits were identified according to the filtering criteria.

Figure 1 presents the PRISMA flow diagram as an algorithm, depicting the systematic process used to identify, screen, assess eligibility, and include studies in a literature review. Initially, 519 articles were identified from various databases based on criteria such as keyword retrieval, timeline (2000–2023), journal type (articles and reviews), and language (English). These articles were screened, resulting in 267 assessed based on titles, keywords, and abstracts. Then, 228 full-text articles were evaluated for eligibility using domain and keyword inclusion/exclusion criteria. Ultimately, 170 studies met the criteria and were included in the synthetic analysis, ensuring a comprehensive and rigorous review of the relevant literature.

3.3. Bibliometric Study

Bibliometric analysis is a “rigorous method for exploring and analyzing large volumes of scientific data” [80] used to reveal development trends in specific research areas [81]. Bibliometric analysis was typically applied in our systematic literature review to evaluate the research trends and publish characteristics of AI in SSCM to address our first research question. It uses a systematic analysis technique to identify the most used keywords, affiliations, and related academic works, which benefits establishing a network of co-authors, countries, and institutions [6]. In particular, it helps researchers evaluate the present significance of research trends using these indicators. Many prior review studies conducted bibliometric analysis in the fields of AI and sustainable supply chain resilience [82,83], AI and circular SCM [84], AI, blockchain, Internet of Things, and supply chain in sustainability [85], information technologies (AI, big data, and machine learning), and AI, SSCs, and third-party logistics [86], and AI, SSC, and sustainable development [6,87]. Therefore, we conducted the bibliometric analysis as an appropriate technique to examine the current studies of AI-integrated technologies in SSCM.

3.4. Text Analysis Using Topic Modeling

Topic modeling is a machine learning technique used to uncover latent themes within a text dataset, representing the text data as a collection of topics and the associated vocabulary distribution for each topic [88]. It has become a widely used and notable approach to identifying the main themes in a large and unstructured dataset [88,89]. This systematic review employed Latent Dirichlet Allocation (LDA) topic modeling, which is one of the most popular methods used for text analysis [90]. The LDA method, first introduced by Blei et al. (2003) [91], was applied to help generate the number of topics. “Each topic is itself a probability distribution of words referred to as a topic model”, and “words can belong to multiple topics, and documents can contain multiple topics” [92]. It automatically organizes extensive document archives by identifying latent topics based on word co-occurrence patterns and estimates document-topic and topic-word distributions to extract primary topics within a corpus, making it useful for text classification and information retrieval [93]. Specifically, LDA assumes documents are mixtures of topics, and topics are mixtures of words, aiming to uncover the hidden topic structure that best explains the observed data. The reasons why we employed LDA modeling as an approach for the systematic literature review are as follows. First, it has the ability to automatically discover latent thematic structures within large and unstructured datasets (N = 170 of our selected studies), as quickly performed by the R language. Second, LDA can uncover the literature’s latent topic structures, revealing the research field’s main issues and trends.

The number of topics was preselected by tweaking the k value =10 in LDA using R. The way of finding the most appropriate number of topics in LDA is arguable. Here, we used a parallel computation method to evaluate the performance of topic models under different numbers of topics [94]. Specifically, ten-fold cross-validation and perplexity were employed as evaluation metrics. The dataset was randomly divided into ten subsets for cross-validation, ensuring the generalization ability and robustness of the evaluation. The candidate’s number of topics ranged from two to twenty, with an interval of two. For each candidate topic number, the model was trained on nine training sets, and the perplexity was calculated for the remaining validation set to assess the model’s predictive performance. The graph allows for an intuitive comparison of the model’s performance under different topic numbers, enabling the selection of an optimal topic number that effectively captures the underlying structure of the data while avoiding unnecessary model complexity. The elbow point at k = 10 indicates that perplexity decreases with increasing k up to this point and becomes stable beyond this point. Thus, we confirmed the ten emerging topics.

In addition, there are two important metrics, Frequency-Exclusivity (FREX) and Lift, in LDA. FREX balances the exclusivity and frequency of terms, and considers how often a term appears in a particular topic and how unique that term is to the topic compared to others [95]. Lift measures the ratio of the term probability in the topic to its probability in the corpus [88]. These metrics helped us better understand and interpret the generated topics, ensuring they were distinct and meaningful.

4. Results and Analysis

4.1. Bibliometric Analysis

4.1.1. Trend of Published Articles

Figure 2 depicts the changing trend of published articles related to AI in SSCM from 2004 to the end of 2023 in the selected journals. It is apparent that in the initial decade, the number of published articles hardly increased from 2004 to 2017. Between 2018 and 2020, there was a gradual increase in growth, but after that, there has been a doubled and redoubled increase in the number of publications (20 in 2021, 44 in 2022, and 62 in 2023) with the increasing attention on AI. Considering the COVID-19 period in the last three years, the application of AI in the field of SSCM has been attracting growing attention in this uncertain circumstance. This also indicates that transdisciplinary articles tend to come out in academia, and the focus on AI technology has attracted the attention of scholars in practical fields, such as SSCM in the innovation industry.

Figure 3 suggests a broad range of publication outlets, showing that studies on AI in SSCM were published in various subject areas. Research and publications in AI in SSCM covered several subject areas over the selected period. Among these subject areas, the most significant number of articles regarding the application of AI in SSCM was published under the two dominant disciplines, namely, engineering (34.8%) and business management and accounting (25.2%), corresponding to more than 50% of the total articles. The other notable subject areas from the selected studies are social sciences (17.9%) and decision science (16.9%). Economics, econometrics, finance, and arts and humanities have comparatively few studies in AI and SSCM studies.

4.1.2. Country with Citation Statistics

The review identifies the status of country contributions in the research field of AI in SSCM through the productive and influential effect from 2004 to 2023. Figure 4 illustrates the top five countries ranked by the number of published articles, citations, and total mean citations with the affiliated country of the first author. As is evident from Figure 4, China is the most productive country in AI in SSCM research, with 36 published articles, and the United States follows next, with 34. India and the United Kingdom contributed 33 and 30 articles, respectively, while France published 19. There are two primary reasons why China is poised to become the most prolific country in the field of AI. The first reason is the government’s active intervention and promotion of AI policies. Studies indicate that the circulation of AI policies saw a significant surge in 2019, which subsequently led to substantial corporate transformations and garnered extensive attention from the academic community [96]. Under the strategic direction of various AI policies, national research initiatives have increasingly focused on establishing projects in the realm of AI. Consequently, a significant number of Chinese scholars have made considerable contributions to this field. This trend is consistent with Figure 2, which shows a marked increase in publications beginning in 2019. Similarly, the United States, as a leading competitor with China in AI [97,98], also exhibits this characteristic.

Figure 4 also illustrates the citation statistics of the top five countries that cited the articles, suggesting that the United States, the United Kingdom, and China received a large number of citations. Although China has the highest number of published articles, the number of citations is almost half that of the United States. Thus, the United States is the most influential country in this field. Moreover, while the UK (n = 30) produced a lower number of published papers than the US, China, and India, they have higher total citations than China and India, with higher average cited published papers, which indicates the journal maintains high-quality standards and that the authors possess the highest influence within their respective research domain. France has 19 published papers, but the average of cited articles is ranked second, indicating the significant impact in this research domain. This also proves that the dominant AI in SSCM is in North American and European countries, like the UK and France. Still, Asia countries like China and India contribute significantly to AI in SSCM globally.

4.1.3. Journal Statistics

Table 3. Most influential publications in the research area of AI in SSCM.

Journals	No. of Articles	Total Citations	h-Index
Sustainability (Switzerland)	25	1022	185
Journal of Cleaner Production	12	536	246
International Journal of Production Research	11	1617	186
Energies	6	108	152
Annals of Operations Research	5	84	125
International Journal of Production Economics	5	454	118
Technological Forecasting and Social Change	5	577	157

depicts that Sustainability (Switzerland) is in first position by publishing 25 articles in AI and SSCM, with 1022 citations. Journal of Cleaner Production and International Journal of Production Research rank second and third in the number of published articles. Particularly worth mentioning is that the Journal of Cleaner Production has the highest h-index (n = 246), and the International Journal of Production Research has the most extensive total citations (n = 1617), which indicates a significant impact and influence within AI and SSCM. While Technological Forecasting and Social Change only has five published articles in this selected field, they have many more citations than the Journal of Cleaner Production, Energies, Annals of Operations Research, and the International Journal of Production Economics.

4.1.4. Keyword Statistics

The VOSviewer software version 1.6.20 obtained 1217 keywords from the selected articles. The minimum number of occurrences of a keyword is five; thus, 59 meets the threshold. Figure 5 visualizes the co-occurrence and annual overlay of the keywords. The circle’s size and color in the figure refer to the keyword’s importance (occurrence) and the average publication year (2018–2022), respectively. For example, SCM and decision support systems are the two keywords of importance that follow AI and sustainable development. Moreover, the terms colored yellow and light green (e.g., data analytics, digital storage, machine learning, Internet of Things, digital technologies, blockchain) represent the keywords that frequently occurred in the two years that can be considered emerging research focuses. This also illustrates the strength of keyword relations based on the closely located cluster presented. For example, it can be observed that sustainable development and interoperability are strongly related to supply chain and manufacturing. Likewise, sustainable development and artificial intelligence are closely related, as are life cycles and artificial intelligence. In contrast, the weak co-occurrence of keywords such as construction industry is at the outermost point. These visualizations can help identify the most recent and related topics for the research of AI in SSCM.

4.2. Text Analysis Using LDA Modeling

The systematic review applied LDA to the selected 170 documents and generated the latent topics by investigating these texts based on the similarity and frequency of words. The text from the title, abstract, and keywords was input for the LDA to generate latent topics. The “topicmodels” package was utilized for topic modeling due to its flexibility and efficiency in handling large-scale text datasets [99]. It offers implementations of topic models such as LDA, along with functions for model fitting, topic inference, and result visualization. The analysis process began with preprocessing the text data, which included converting text to lowercase, removing punctuation and numbers, eliminating stop words, and trimming excess whitespace to reduce noise and enhance model accuracy. Subsequently, the preprocessed text was transformed into a document-term matrix, preparing it for LDA model fitting. After selecting the model with the lowest perplexity, LDA fitting was performed. The model results were visualized using “ggplot2” and “tidytext”, highlighting the highest-weighted words within each topic along with their beta values, thereby revealing the underlying topic structure in the text data. Finally, we obtained the results as shown in

Table 4. FREX and Lift of Generative Keyword.

Topics	Keywords	FREX	Lift
1 Supplier Selection in SSCM using decision support systems	Decision, supply_chains, supplier, support, systems, selection	continuous, continues, oscillators, agile, Denizli, focuses, triple, detection, PLS, multiproduct	ambiguity, amalgamates, robots, Asia, effectual, messaging, radical, engaged, seamless, kingdom
2 Optimizing Global SSCM through AI-integrated Big Data	Data, supply_chain, management, digital, artificial_intelligence, sustainability	many, moderating, conflicting, sample, dearth, exhibited, UCS, exploration, grown, listed	radically, regulation, animal, discovery, excellence, metric, required, managing, optimization, globalized
3 Food Sustainable Supply Chain Optimization	Supply, sustainable, digital, chain, food, management, technologies	inventory, learn, pinning, scenario, descriptive, empower, trade, environment, entire, exhibit, makings	previously, prominent, SCM, horticulture, feasibility, integration, precision, involvement, live, hybridized
4 Smart Manufacturing in Supply Chains	Supply, chain, sustainable, industry, manufacturing, artificial, intelligence, data	match, local, put, sectors, dissemination, faced, trends, footage, methodologically, analyze	readers, public, SMAC, largest, generally, manufacturer-supplier, radio, model-based, reduced, bank
5 Renewable Energy in Supply Chains	Industry, supply, analysis, sustainable, energy, data, artificial_intelligence, sustainability	implementing, influences, scarce, secondly, depletion, competencies, twenty, enabled, profitable, Malaysia	pivot, post-fuel, summarized, first-stage, almost, realities, involving, shocks, indirectly
6 Transportation Optimization and Environmental Management in Supply Chains	Supply, sustainability, food, chain, research, industry	inductive, intersection, phase, seaports, country, derivatives, twenty, forecasts, maximizing, merge	philosophies, powerful, scalable, medium, behaviors, endless, sensor, meta-ontology, ranking, get
7 Decision Support in Supply Chains	Supply, system, sustainable, manufacturing, environmental, product	known, mapping, RR, scanners, data-intensive, Europe, trying, crippled, cost-effectiveness, beyond	problem-solving, ration, situational, effect, blending, interplay, remotely, ArcelorMittal, API, CMS
8 Sustainable Logistics Management and System Integration	Smart, energy, supply, management, systems, logistics	letter, mere, pure, SDG, design-science, deployment, unnecessary, goal, programming, pre-COVID, loss	protective, redesign, six, insufficiently, fundamental, formally, Scopus, optimizing, sense, implementing
9 Data-driven Risk Management in Disruptions	Supply_chain, research, review, management, industry, literature	Finland, formation, parameters, scan, edge, firm, treatment, emergence, package, bitcoin	metrics, monitor, sampling, long-lasting, hotspot, micron-controller, routing, improves, rule, cationic
10 Environmental Management in Construction Supply Chains	Supply_chain, construction, risk, sustainability, management, environmental	holds, Indian, paths, sales, deep, evidence, according, deeply, massively, loss	path, portfolios, satisfaction, demand-markets, establishing, lithium-ion, algorithm, essence, real, hegemony

.

Table 4 lists all the summarized topics, and FREX and Lift. As previously stated, FREX in topic modeling is used to identify words that are both frequently occurring in a specific topic and relatively exclusive to that topic compared to others; Lift is a metric used to evaluate the importance of a word to a specific topic by comparing the word’s probability within the topic to its overall probability in the entire corpus [93]. For example, words with the highest probability that generated topic label 1 are “decision”, “supply”, “supplier”, “support”, “systems”, “sustainable”, “system”, “sustainability”, “selection”, and “chains”. The terms in topic 1 suggest a strong focus on the decision support systems involved in managing supply chains, emphasizing the importance of sustainability. Thus, based on the highest words, we summarized the topic in label 1 as “supplier selection in SSCM using decision support systems”.

The FREX metric in Table 4 shows a focus on continuous processes, agile methodologies, detection systems, and specific technologies. As we explained before, the appearance of these words in FREX indicates their high frequency of occurrence within topic 1 and their low frequency in other topics. The words that appeared in Lift metrics, such as ambiguity, amalgamates, robots, and effectual, further emphasize specific aspects and benefits of using decision support systems in SSCM, illustrating how these systems amalgamate technologies and methodologies to create seamless, effective, and agile supply chains. Therefore, the emerging research focus of AI and SSCM is to optimize decision-making and decision support systems, ensure sustainability, and strengthen supplier support through continuous improvement, agile methods, and advancements in robotics and effective communication strategies. According to these interpretations, Section 4.3 provides 20 specific propositions based on the results.

4.3. Propositions of Emerging Topics of AI in SSCM

Based on the previously generated emerging topics through LDA modeling, this section proposes 20 propositions for future research directions based on topic, FREX, and Lift.

4.3.1. Supplier Selection in SSCM

The ambiguity of the world drives digital transformation for resilient supply chains. The worldwide attention to sustainability paradigms focuses on the criteria of supplier selection [29]. Moreover, decision-making in management often involves complex factors and uncertainties, and AI technology can provide more intelligent and comprehensive decision support to select the appropriate suppliers. In this decision process, it is necessary to consider socio-economic and environmental factors. Thus, Tavana et al. (2023) [100] used fuzzy inference to evaluate suppliers from economic, social, circular, and Industry 4.0 perspectives. Ebinger and Omondi (2020) [52] leverage AI-integrated blockchain to trace and track the compliance of individual and sub-supplier products with sustainability certifications and standards. Therefore, we make the following propositions:

Proposition 1: AI technologies should be leveraged to help companies evaluate suppliers’ sustainability performance and optimize supplier networks.

Proposition 2: Agile supply chains should be built that efficiently respond to market changes and supply chain disruptions and maintain the supply chain’s continuity and suppliers’ stability.

4.3.2. Optimizing Global SSCM through AI-Integrated Big Data

The widespread application of AI and big data in SCM, particularly in decision support system processes and analyzing large volumes of data, improves decision-making efficiency and accuracy [48]. Driven by these technologies, SSCM needs to consider uncertainties from the external environment to respond to the demands of the global market. Therefore, real-time data analysis and predictive models can facilitate sustainable activities in the supply chain (e.g., [26,36,101]). For example, Neethirajan (2023) [101] highlights the transformative potential of AI-integrated technologies in improving the efficiency and sustainability of the dairy livestock export industry through enhanced traceability and real-time monitoring. It also addresses the associated challenges and ethical considerations, providing a strategic framework for successfully integrating these technologies into long-distance livestock transportation [101]. Thus, based on the above discussion, we make the following propositions:

Proposition 3: AI and big data should be applied to facilitate SSCM with precise decision-making of global supply chains.

Proposition 4: Globalization and sustainability requirements drive the innovative application of AI-integrated big data and digital technologies in SCM.

4.3.3. Food Sustainable Supply Chain Optimization

In the distribution network, the issues of delays and inefficiencies result in food waste and customer satisfaction in the delivery processes of supply chains [31]. Similar customer demands and logistics issues have gained attention in food sustainability [42,102,103]. Thus, AI-integrated technologies for reducing food wastage, real-time product monitoring, and reducing scalability issues have been proposed in SSCM. In addition to the mentioned AI technologies, such as the Internet of Things, robots, and XGBoost, other optimization techniques based on AI, such as machine learning, could be considered for predicting customers’ food demand to minimize food waste. Moreover, upstream parts of the food supply chain, such as agricultural production, significantly impact the environment and resource use. Thus, precision horticulture or agriculture can improve production efficiency and reduce resource waste and environmental pollution. For example, precision agriculture technology, such as machine learning and the Internet of Things, can optimize planting and breeding processes and improve production efficiency and resource utilization [104,105]. Hence, we make the following propositions:

Proposition 5: Using predictive analytics to accurately predict food demand, optimize inventory management, and reduce food waste in delivery should be applied in SSCM.

Proposition 6: Precision industrial technologies should be explored and applied to SSCM.

4.3.4. Smart Manufacturing in Supply Chains

The industrial supply chain impacts many global environmental challenges, particularly within the manufacturing sector [6], which drives several studies that have focused on how AI impacts sustainable manufacturing performance. For instance, Panigrahi et al. (2023) [106] proposed that AI chatbots positively influence supply chain visibility to enhance sustainable supply chain performance. Jamwal et al. (2022) [107] discussed the evolution of deep learning approaches. They proposed a deep learning-based framework in manufacturing, highlighting how deep learning-based approaches can enhance the sustainability performance of industries. Moreover, some studies have recommended that digital twins overcome social, economic, or environmental issues in the supply chain. For example, Kamble et al. (2022) [108] revealed that a digital supply chain twin should encompass both the entities and individuals involved across the entire supply chain rather than being confined solely to local manufacturing systems. Kombaya Touckia (2022) [109] proposed a digital twin design and simulation model for reconfigurable manufacturing systems to establish more flexible and agile manufacturing systems. Thus, we propose the following:

Proposition 7: AI-integrated technologies, such as chatbots and digital twins, facilitate agile manufacturing; therefore, different integrated technologies and algorithms for dynamic supply chains are needed.

4.3.5. Renewable Energy in Supply Chains

To reduce dependence on fossil fuels in the supply chain and improve sustainability, renewable energy that provides energy with little or no emissions in terms of greenhouse gases or pollutants has been selected as the most efficient alternative to fossil fuels [53]. Some suggestions related to Industry 4.0 are driven by key digital technologies, such as AI, IoT, and big data analytics (BDA). Regenerative energy that uses biological cycles to shift to renewable energy through AI, IoT, and BDA was mentioned in the “ReSOLVE” framework proposed by MacArthur et al. (2015) (cited [103]). Balaman et al. (2016) [30] proposed a bi-level decision support system to aid the modeling and optimization of multi-technology, multi-product supply chains and co-modal transportation networks for biomass-based (bio-based) production to select the most favorable supply chain configuration, and thus design a bio-based supply chain with a small environmental impact. Helal et al. (2023) [82] explored the broader use of machine learning through a systematic literature review and found AI has potential research gaps in decarbonizing energy systems. Thus, we make the following propositions:

Proposition 8: A data analysis platform should be established using AI technology to analyze and summarize supply chain data, and provide decision support.

Proposition 9: AI should be leveraged to analyze energy usage data to optimize the application of renewable energy and reduce carbon emissions.

Proposition 10: It is necessary to assess the biomass associated with supply chains, converting renewable energies at every stage of the supply chain.

4.3.6. Transportation Optimization and Environmental Management in Supply Chains

Supply chain disruptions are always associated with weather conditions, traffic accidents, etc., negatively impacting the entire supply chain [36]. Therefore, Karam and Reinau (2022) [36] presented a novel hybrid approach combining a simulation model, optimization algorithms, and a cost-effectiveness analysis by evaluating cost and the environmental effect of CO₂ emissions. Balaman et al. (2018) [30] proposed a hybrid solution methodology that integrates fuzzy set theory and the ε-constraint method, addressing the dual challenges of economic and environmental sustainability, providing a balanced solution that considers both financial and ecological impacts. Similarly, Tirkolaee and Aydin (2022) [110] developed a hybrid solution technique based on possibilistic linear programming and a Fuzzy Weighted Goal Programming approach to accommodate their suggested bi-level model to ensure the sustainability of the overall system in uncertainty. Mansouri et al. (2015) [111] focused on environmental sustainability and the trade-offs involved with economic and operational objectives. They examined the potential of multi-objective optimization (MOO) as a decision support to improve sustainability in maritime transport [111]. Thus, we propose the following:

Proposition 11: Transportation should be incorporated into environmental management and economic dimensions in the sustainable transportation model.

Proposition 12: Use AI algorithms to conduct environmental impact assessments, develop measures to reduce carbon emissions, and achieve sustainable transport.

4.3.7. Decision Support in Supply Chains

The focal company faces multiple tangible and intangible risks due to lower-tier suppliers’ noncompliance with sustainability standards and failure to meet the stakeholders’ expectations of extending sustainability to lower-tier suppliers whose products are produced under environmentally and socially unsustainable conditions [35]. To address this issue, Ref. [35] developed a model-driven decision support system using a Bayesian network that can assist operations managers in selecting the most effective sustainability governance approaches in a multi-tier supply chain, considering each situation, linking the decision support system to risk management in supply chains. Similar studies on risk management via decision support systems, for example, Ref. [37] proposed a multi-objective simulation-based decision support tool for wine’s transportation risk management. Thus, we make the following propositions:

Proposition 13: AI should be used for intelligent decision support and system optimization to ensure the prediction of suppliers’ sustainable performance.

Proposition 14: Develop an algorithm-based model to identify the low-tier suppliers’ sustainable products in every manufacturing process in supply chains.

4.3.8. Sustainable Logistics Management and System Integration

Optimizing the logistics process and system integration is the key to sustainable supply chain development due to the sustainability awareness in various economic sectors, which drives firms to use logistics outsourcing to reduce emissions with compliant logistics service providers. Ref. [86] established trends in technology and sustainability in logistics outsourcing, identifying key elements such as trust, cooperation, multi-selection criteria, and AI capabilities, highlighting the growing research focus on green logistics, reverse logistics, and the circular economy. Bhatti and Assemgul (2023) [112] investigated the impact of intelligent inventory systems on reverse logistics in the Saudi Arabian manufacturing industry and found that implementing AI-based inventory management systems improves operational and warehousing activities in reverse logistics addressing challenges and enhancing resource replacement. Stanisławski and Szymonik (2021) [113] targeted production companies that use logistic supply chains in their operations, highlighting the relevance and impact of intelligent systems, such as AI, blockchain, IoT, etc., on improving logistics efficiency and communication between companies and consumers; however, the target manufacturing companies do not believe that implementing intelligent systems provides significant market advantages. Based on previous discussions, we propose the following:

Proposition 15: Explore the synergy between AI and blockchain in creating sustainable logistics systems. Research could focus on implementing these technologies to optimize logistics management, ensuring efficiency and sustainability in supply chain operations.

Proposition 16: Research should explore combining blockchain and AI technologies to redesign supply chain systems, ensuring transparency and security. This direction could address the insufficiently formalized aspects of traditional logistics systems and optimize them for better performance and resilience.

4.3.9. Data-Driven Risk Management in Disruptions

The COVID-19 pandemic has severely impacted supply chains worldwide, affecting finance, lead time, demand, and production performance [114]. Bechtsis et al. (2022) [115] highlight the need for data-driven risk management paradigms to enhance supply chain security, resilience, and sustainability, especially in light of disruptions like the COVID-19 pandemic. The study proposes a new framework addressing gaps in the literature and practice, validated by a case study in the organic food supply chain, emphasizing the role of digital technologies in data monetization. Ref. [116] conducted a case study of top global businesses using supply chain risk management, demonstrating that AI, big data, and adaptive logistic regression classifier systems are vital for risk identification and assessment, evaluating organizations based on success rate, resilience, and customer value. Moosavi et al. (2022) [114] addressed the need for systematic literature surveys to extract disruption management strategies from influential research and provide managerial insights to ensure resilience and SSCs for critical products during pandemics. Based on these discussions, we propose the following:

Proposition 17: Create and deploy AI-driven risk management frameworks that leverage big data analytics and adaptive logistic regression classifiers to enhance supply chain resilience, security, and sustainability in response to disruptions.

Proposition 18: Investigate the role of digital technologies such as AI, IoT, and blockchain in managing supply chain disruptions, aiming to identify best practices and innovative strategies that enhance supply chain performance and ensure the resilient and sustainable supply of critical products during global crises.

4.3.10. Environmental Management in Construction Supply Chains

Numerous intricate sustainability components, such as environmental, social, and financial, should be considered for the long-term evaluation of SCM in the construction sector [117]. Ref. [118] proposed that the construction material industry is poised to adopt AI and environmentally and socially responsible practices in supply chains to address global challenges and market emergencies. Ref. [119] also believe that construction supply chains adopt sustainable practices can lower the environmental impact. However, the investigated results revealed influential causal issues that affect the adoption of AI in the construction supply chain in India, which are mainly trust, uncertainty, risk, and project cost. Yevu et al. [120] explored the application of digital twin technology in the prefabrication supply chain, focusing on enhancing smart construction and monitoring carbon emissions, revealing a significant research interest in emissions control, AI-based decision-making, and blockchain integration, alongside the practical use of technologies. Thus, we make the following propositions:

Proposition 19: Explore strategies to mitigate trust, uncertainty, risk, and cost challenges to facilitate broader adoption of AI and digital twin technologies in construction supply chains, enhancing both sustainability and efficiency.

Proposition 20: Investigate the integration of blockchain technology within digital twin frameworks to improve transparency and decision-making processes in the construction material industry’s supply chains, thereby supporting more sustainable and environmentally and socially responsible practices.

4.4. AI-Integrated Technologies with Sustainable Issues

To holistically mine texts based on the latent topics, we summarize fourteen AI-integrated technologies associated with eleven sustainable issues of five pillars of sustainability in recent SSCM areas from a matrix form in

Table 5. AI-integrated technologies in SSCM.

		AI-Integrated Technologies
			1	2	3	4	5	6	7	8	9	10	11
			Machine Learning	Big Data Analytics	Fuzzy	Internet of Things	Decision Support System	Robotic	XGBoost	Deep Learning	Explained AI	Digital Twin	Block Chain
SSCM- Environmental Issues	1	Resource waste	[19]	X	X	[102,103,124,125]	X	[103]	[126]	X	X	X	X
	2	Energy use and carbon emissions	[19]	[23,127,128]	[129]	[22]	[130,131]	X	X	X	X	[120]	X
SSCM- Economic issues	3	Cost control	X	X	X	X	[131,132]	X	X	X	[133]	X	X
	4	Supply chain disruption	[121,122]	[25]	X	[123]	[36]	X	X	X	X	[109]	[114]
	5	Resource consumption	X	[27]	X	X	X	X	[126]	X	X	X	X
SSCM- Social issues	6	Working conditions	X	[134]	X	[134]	X	X	X	X	X	X	[134]
	7	Fair dealing	X	X	X	X	X	X	X	X	X	X	[135]
	8	Corporate social responsibility	X	[120]	[136]	X	X	X	X	X	X	X	X
SSCM- Management issues	9	Product quality assurance	[104]	[48]	X	X	[110,133]	[137]	[126]	X	X	[109]	X
	10	Processes complexity	X	X	X	X	X	[137]	X	[107]	X	[137]	X
	11	Customers satisfaction	[138]	[48]	X	X	X	[139]	X	X	X		X
	12	Supplier selection	[138]	[29]	[100]	X	X	X	[126]	X	X	X	[52]
	13	Risk management	[122]	[48]	X	X	X	X	[126]	X	X	X	X
	14	SSCs’ transparency	X	[52]	X	X	X	X	X	X	X	X	[135,137]

. For example, the disruption of the supply chain (economic issues) can be addressed by machine learning [121,122], and it also can be solved by other AI-integrated technologies, such as big data analytics [25], Internet of Things [123], and digital twin [109]. This matrix table presents specific applications of AI-integrated technologies that enable SSCM that were not mentioned in previous topics. “X” stands for the current research gap between AI-related technologies and sustainable issues in supply chains; for example, big data analytics, fuzzy, decision support systems, deep learning, explained AI, digital twin, and blockchain integrated with AI have potential for use in research on addressing the issues of resource waste management (environmental issues).

Furthermore, we can combine the previous analysis of the latent topics in Section 4.2 (Table 4), such as the propositions on a decision support system that facilitate a sustainable supplier selection in the first topic, with the commonality found in Table 5. For example, the management issue of supplier selection has yet to be solved by decision support systems in our selected studies. Topic 5 (Renewable Energy in Supply Chains) concerns environmental issues. Thus, Proposition 8 and Proposition 9 suggest a data analysis platform to provide decision support in analyzing energy usage data to optimize the application of renewable energy and reduce carbon emissions. This aims to solve environmental issues in terms of “energy use and carbon” (Table 5, No. 2), and compared to AI-integrated technologies, explained AI has yet to be used to build a data analysis platform; therefore, future studies may consider how to use explained AI to address this question. These examples also demonstrate the accuracy and reliability of LAD modeling (Table 4) and the matrix framework (Table 5).

5. Discussion and Proposed Research Framework

This study aims to examine current research on AI-integrated technologies in the entire supply chain process to inform future research directions. To address the first research questions, a bibliometric analysis was employed to examine the trends concerning applied AI and SSCM. The trend of AI in SSCM can be summarized as follows: first, AI and SSCM have gained a substantial focus from academia, notably from powerhouses in AI development, and especially the United States, China, and Europe. Second, the high citation volumes are associated with studies focusing on AI’s role in addressing sustainable issues in SCM, reflecting the importance and recognition of these themes in the academic community. Third, the keywords analysis highlights this field’s interdisciplinary collaboration, application-oriented research, and technology-driven nature.

Our second research question addresses the latent topics in AI and SSCM. Therefore, we conducted text analysis and employed an LDA approach to explore the latent topics related to the significance of AI-related technologies in SSCM. Using LDA, we generated ten topics. According to the defined ten topics with FREX and Lift words, the study generated thematic topics using LDA, such as supplier selection in SSCM, optimizing resources for the circular economy, food SSC optimization, smart manufacturing in supply chains, renewable energy in supply chains, transportation optimization, environmental management in the supply chain, decision support in supply chains, sustainable logistics management and system integration, data-driven risk management in disruptions, and environmental management in construction supply chains. The impact of AI on these topics is considerable, and many studies have shown its importance.

To further understand AI in SSCM and address our third research question, we identified the AI-related or integrated technologies with sustainable issues using a matrix form. Eleven AI-related technologies that contributed to addressing fifteen sustainable issues in supply chain processes were presented. However, some AI integration technologies have not been fully explored to address sustainability issues in the supply chain, such as social issues, including fair dealing and work conditions.

In addition, the selected studies demonstrate AI’s capability to construct a sustainable supply chain. In the management processes, sustainable issues arise from activities involved in SCOR’s six elements. For example, planning processes involve strategies and methods for demand/supply planning and management; therefore, machine learning is used as a demanding forecasting tool to optimize resource usage and reduce waste, and make plans that prioritize resource efficiency and minimize environmental impacts, incorporating sustainability goals into overall supply chain strategies [37,105,140,141]. The current literature has provided insight into SCOR’s six elements across different sustainable activities with AI-integrated technologies or algorithms, as

Table 6. AI-enabled sustainable activities in SCOR’s elements.

SCOR	Sustainable Activities	AI-Integrated Technologies/Algorithms
Plan	Developing plans that prioritize resource efficiency and minimize environmental impact [116,142,143];	AI and Adaptive Logistic Regression Classifier (ALRC); decision support system; AI-push XGBoost; Artificial neural network.
	Utilizing forecasting tools to optimize resource usage and reduce waste [126,144];
	Incorporating sustainability goals into the overall supply chain strategy [37,42,140,141].
Source	Sustainable supplier selection [29,100,136,138,145,146,147,148];	Fuzzy Delphi; Decision support systems; Machine learning; Fuzzy logic; Fuzzy inference; Genetic algorithms; Robotic.
	Sustainable material sourcing [48,118].
Make	Reducing waste and emissions during production [22,120,130];	Machine learning; IoT; Blockchain.
	Utilizing renewable energy sources in production facilities [53].
Deliver	Optimizing routes to reduce fuel consumption and emissions;	Big data analytics; Decision support system; Blockchain.
	Using eco-friendly transportation options like electric or hybrid vehicles [15,30,36,110,111,149];
	Implementing smart logistics solutions to enhance efficiency [15].
Return	Establishing systems for the return, recycling, and disposal of products [142,150,151,152]	Machine learning; Big data analytics; AI-push; Blockchain; IoT; decision support system.
	Managing end-of-life products responsibly to minimize environmental impact [132,153].
Enable	Implementing lean principles to reduce excess inventory and waste [112,126];	Decision support systems; Big data; Fuzzy set; IoT; Swarm intelligence algorithms.
	Sustainable product and packaging [154];
	Working with suppliers to improve their sustainability practices [24];
	Tracking sustainability metrics such as carbon footprint, energy usage, and waste [34,148]

shows.

To address our fourth research question, this research proposed a framework (Figure 6) comprising ten emerging research themes for SSC and integrating AI-related technologies and algorithms into SCOR’s six elements to overcome the issues arising from the above sustainable activities. The research framework defines the advantages of incorporating AI to attain sustainability in the SCM processes. It includes ten latent topics of SSCM, with eleven sustainable activities previously shown in Table 6. The AI-integrated technologies play significant roles in addressing the issues of traditional supply chains and enabling sustainable activities in planning, sourcing, making, delivering, returning, and enabling processes based on SCOR. SSCM leverages AI-integrated technologies and algorithms to address sustainable issues through sustainable activities. Therefore, in this framework, the left and right boxes rotate dynamically with the dynamic sustainability in supply chains, and the emerging future directions would inform the agenda and may adjust dynamically.

6. Conclusions and Limitations

Globalization has had social and economic impacts on traditional SCM, and the problems arising in the global supply chain process have forced continuous innovation and optimization of SSCM. Thus, companies are facing pressures from government regulations, societal expectations, and competitive dynamics to integrate sustainability into their operations [6]. To address the issues of sustainability, AI-integrated technologies in supply chain processes have revolutionized traditional SCM processes. Ten emerging research themes were produced among the selected studies, which are supplier selection in SSCM, optimizing global SSCM through AI-integrated big data, food SSC optimization, smart manufacturing in supply chains, renewable energy in supply chains, transportation optimization, and environmental management in the supply chain, decision support in supply chains, sustainable logistics management and system integration, data-driven risk management in disruptions, and environmental management in construction supply chains, incorporated with 20 propositions. Our review explores the roles of AI-integrated technologies and algorithms in SSCs related to environmental, economic, social, and management issues; meanwhile, we discovered that there are still technical barriers, such as social issues including working conditions, ethics, and fair dealing, as well as challenges to AI technology in addressing sustainable supply chain issues. These gaps are also in line with the contents of the 20 propositions, and provide future research directions for researchers.

Practically, our proposed framework first underscores the significant role of AI-integrated technologies in dynamically addressing and resolving the issues inherent in traditional supply chains. By applying AI technologies, practitioners and engineers can continuously optimize processes such as planning, sourcing, making, delivering, returning, and enabling to resolve potential sustainable issues in the future. Secondly, this dynamic framework is one in which the sustainable activities (left box) and sustainable issues (right box) rotate dynamically with the evolving sustainability requirements of supply chains. This dynamic interaction ensures that SSCM can quickly adapt to emerging challenges and opportunities. Thirdly, SSCM utilizes AI-integrated technologies and algorithms to address existing sustainable issues and enable proactive and adaptive sustainable activities. This dynamic capability ensures that supply chains remain resilient and sustainable.

Similar to other review studies, this study has limitations. Firstly, the inclusion criteria limited the results of data sampling. For example, we adopted AI to search for keywords but ignored other relevant terms such as I4.0, digitalization, digitization, and digital transformation. Hence, future research needs to expand the search strings to review holistically. Secondly, conference proceedings were excluded from the selected studies considering the research breadth but would have some of the latest research trends [77] and may be useful to consider for studies interested in specific technologies within SSCM. Thus, future studies would include high-ranked conference papers for a broader diversification of relevant articles. Thirdly, we adopted the SCOR model to present sustainable issues and activities in the supply chain processes. However, each stage of this model is disrupted by technological innovation [6]. As previously explained, the SCOR model provides extensive sub-framework detail, and the other three levels need to be applied in this study. Last, in this uncertain context, AI in SSCM meets challenges and opportunities in different industries, and future studies will be needed to narrow down to one specific sector and provide a conclusive statement on how to measure challenges to survive disruptive innovation in SSCM.