Artificial Intelligence Applications for Cleaner Production and Sustainable Development in Southeast Asia: A Systematic Review and Future Research Directions

Abstract

Artificial intelligence (AI) has reshaped various aspects of human lives, particularly through its capabilities to address complex sustainability challenges. Despite the rapid expansion of AI applications, their contribution to cleaner production and sustainable development remains underexplored, especially in developing nations. In Southeast Asia (SEA), where AI adoption has grown substantially across environmental, economic, and social dimensions, research that examines its role in cleaner production outcomes remains fragmented. In view of this gap, this study conducts a systematic literature review (SLR) of AI applications related to cleaner production and sustainable development by examining relevant themes, application areas, and sustainability dimensions addressed by AI, while evaluating the maturity of AI methodologies, alignment with cleaner production outcomes, and integration with circular economy and resource efficiency goals. Moreover, it investigates the barriers and challenges that constrain AI application and offers future research directions to advance AI deployment for cleaner production and sustainable development across SEA countries.

1. Introduction

Sustainability broadly refers to the fulfillment of human basic needs without compromising the needs of future generations. It is commonly viewed as a condition where humans live harmoniously and responsibly with the environment through the responsible use and efficient use of natural resources [1]. Furthermore, this approach reflects a balanced integration of environmental, economic, and social sustainability considerations, which is widely characterized as the Triple Bottom Line (TBL). These three dimensions are interdependent and should be in synergy to achieve the long-term sustainability goals.

As the world confronts major dilemmas of the 21st century such as climate change, continuous deterioration of natural resources, and rising global population, it echoes an urgent call toward sustainability. Recent reports on climate change indicate a forecasted increase of 2.6 °C in global temperature due to insufficient efforts and weak climate commitments from countries. In addition, the continued fossil fuel emissions bring the world into a new and catastrophic era of extreme weather events compared to the pre-industrial period [2]. Likewise, the degradation of natural resources presents an alarming situation, which is marked by the Earth Overshoot Day, where the human consumption of natural resources has already exceeded the capacity of the planet to regenerate them within the same year [3]. Another pressing concern is overpopulation, in which recent statistics show that the global population is anticipated to reach about 10.3 billion by 2080, which has equally negative effects for food security and the loss of biodiversity [4]. Consequently, these issues signify the need for concrete action, particularly in promoting cleaner production and advancing sustainable development.

In response to these pressing sustainability issues, the United Nations (UN) continuously exerts efforts to confront both present and future sustainability challenges. This commitment is demonstrated through the continuous efforts to achieve the 17 UN Sustainable Development Goals (SDGs), which serve as a universal blueprint for action to eradicate poverty, protect the environment, and ensure peace and prosperity by 2030 [5]. Aligned with the three dimensions of sustainability, the SDGs emphasize that progress in a specific dimension influences outcomes across other dimensions, underscoring their interdependent nature. This phenomenon can be realized through cleaner production and sustainable development, which act as critical drivers in addressing and resolving timely sustainability issues.

Within the broader context of sustainable development alongside advancements brought by Industry 5.0, cleaner production has emerged as an essential strategy for improving environmental performance while maintaining economic competitiveness and social well-being. Cleaner production refers to the implementation of environmental methods to processes, products, and services to achieve efficiency while reducing the threats to society and the environment [6]. This approach entails the efficient management of resources and energy, development of smart technologies (e.g., artificial intelligence, blockchain technology, big data analytics, etc.), and the strategic formulation of policies that have an eventual sustainable impact. This also enhances the organization and coordination of supply chains across sectors and industries. Hence, cleaner production initiatives play a pivotal role in contributing to sustainable development. Meanwhile, sustainable development integrates human-centric values with technological advancements aimed at building resilient and efficient industries [7]. This framework encompasses different aspects of human–machine interactions, circular economy models, and renewable energy integration that are directed toward reducing waste, decreasing energy consumption, and minimizing environmental impact. Together, cleaner production and sustainable development represent essential pathways for addressing contemporary sustainable challenges.

Among the emerging technologies that drive this transformation, artificial intelligence (AI) demonstrated profound and far-reaching impacts on business, governments, society, and even sustainability at large. AI refers to computational systems capable of performing tasks that typically require human cognitive abilities, including learning, reasoning, and decision-making [8]. Over the past decade, AI applications have expanded rapidly, which has generated substantial impacts to various sectors such as healthcare, agriculture, finance, logistics, and manufacturing. In the context of sustainability, AI offers new opportunities to accelerate progress toward the achievement of SDGs, as it facilitates systems thinking while enabling data-driven insights and decision-making as well as improving operational efficiency [9]. Given these capabilities, AI becomes a key enabler of advancing cleaner production and supporting sustainable development objectives.

In this study, cleaner production is defined as a preventive and process-oriented strategy that aims to reduce environmental impacts across the entire production lifecycle through resource efficiency, emissions minimization, and waste reduction. Unlike outcome-based sustainability monitoring, cleaner production emphasizes operational interventions, including lifecycle assessment (LCA), embodied carbon quantification, process optimization, and closed-loop material flows [10,11]. Accordingly, AI applications are relevant to cleaner production as they support measurable improvements in production efficiency and reductions in environmental burdens, and not just by solely predicting sustainability-related indicators.

The transformative potential of AI has become evident in Southeast Asia (SEA), where it has gradually evolved from promise to practice. Recent developments indicate that the AI sector in SEA reached approximately 4 billion US$ in 2024, which is expected to grow fourfold by 2033 [12]. This notable growth highlights the strong potential of AI in accelerating the digitalization of the region. Despite these recent developments, SEA continues to face persistent sustainability challenges amid rapid economic development brought by industrialization and urbanization, which encompass massive deforestation and biodiversity loss, high vulnerability to climate change, severe levels of pollution, high reliance on non-renewable energy, and socio-economic and governance constraints [13]. Despite the growing global interest in the intersection of AI and sustainability, existing literature focusing on cleaner production and sustainable development remains limited, especially in the context of the SEA region. From this perspective, this study aims to address this gap through a systematic literature review approach. Particularly, this study seeks to answer the following research questions:

• What are the prevalent themes, application areas, and sustainability dimensions (environmental, economic, social) addressed by AI research in SEA?
• To what extent are AI methodologies employed to improve cleaner production and sustainable development outcomes in SEA?
• What are the barriers, challenges, and future research directions for AI applications for cleaner production and sustainable development in SEA?

This systematic review contributes by examining the regional characteristics of SEA and the role of AI application for cleaner production and sustainable development in the region. The paper is organized into major sections structured as follows: Section 2 provides an overview of AI for sustainability in SEA; Section 3 describes the methodology; Section 4 presents the results; Section 5 discusses the implications of the findings; and Section 6 provides the conclusions and limitations.

2. Overview of Sustainability in SEA

SEA has emerged as one of the fastest-growing regions for AI adoption, driven by rapid industrialization coupled with growing investments in emerging technologies and the establishment of regional technology hubs and innovation centers [12]. Governments across the region have begun institutionalizing national AI strategies aimed at enhancing industrial efficiency and productivity, improving public service delivery, and strengthening long-term economic competitiveness [14]. These initiatives have demonstrated the role of AI as a strategic enabler for cleaner production and sustainable development.

The diffusion of AI in the SEA region is driven by its young and promising potential workforce capable of maximizing the use of advanced technologies, dynamic startup ecosystems, and growing infrastructure development due to massive interest in building datacenters [15,16]. These developments are notable in several SEA countries such as Singapore, which has become a regional center for AI investments and innovation. Likewise, the strategic locations of Vietnam and Malaysia have also attracted investors for AI research and manufacturing automation and the establishment of cloud computing facilities [17,18]. Meanwhile, the Philippines and Indonesia have also experienced notable growth in AI use and applications, particularly in sectors such as manufacturing, agriculture, logistics, and public administration, where AI is deployed to improve operational efficiency and service delivery [19,20].

Despite these positive trends, AI adoption across SEA remains uneven due to differences in technology readiness among each country, as well as their financial capacity and human capital development. While more digitally advanced countries have demonstrated growing AI ecosystems, other countries struggle to continue due to various constraints. These regional asymmetries have important implications for sustainability outcomes, as they influence the extent to which AI can be effectively leveraged to support cleaner production, circular economy, and industrial systems.

Consequently, this systematic review of AI applications in SEA is necessary to assess their alignment with sustainability dimensions, understand the maturity of AI methodologies, and identify the barriers and challenges for their broader adoption to evaluate the current and future directions for AI in advancing cleaner production and sustainable development across the SEA region.

3. Methodology

Systematic literature review (SLR) is widely employed in the physical and medical sciences research, and its application has extended to the computer science and information systems research in recent years to synthesize existing evidence, minimize reporting bias, and provide a transparent and reproducible research process [21]. In this study, an SLR was conducted by integrating Harzing’s Publish or Perish (PoP) software (version 8.19.5300) with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework.

PoP software was employed to capture the interdisciplinary and region-specific nature of AI applications for cleaner production and sustainable development in SEA based on predefined keyword combinations and search strategies, as illustrated in Figure 1. The software extracts and analyzes citation data from Google Scholar to ensure broader coverage of emerging and applied research that may not yet be fully represented in curated citation databases [22], particularly from developing economies in SEA. Moreover, PoP is designed to allow individual researchers to identify influential studies and map research trends, which effectively assess research impact and research output [23]. Accordingly, database retrieval was complemented by inclusion criteria, systematic screening, and post-retrieval quality assessment so that only relevant and methodologically sound studies were retained for synthesis.

In parallel, the PRISMA framework was adopted to guide the identification and screening of publications, which enhances the transparency and quality of the review process and provides a reference point for other researchers in the field [24]. This review followed the PRISMA 2020 guidelines (Supplementary Materials) to ensure a transparent and reproducible research process [25]. Although the review protocol was not registered in a public registry, a structured review protocol defining the search strategy, eligibility criteria, screening procedures, and data extraction was developed prior to the literature search to ensure methodological rigor and consistency.

3.1. Search Strategy and Study Identification

The researchers initially employed a set of keywords such as “artificial intelligence”, “cleaner production”, “sustainable development”, and “Southeast Asia” through the PoP software. To ensure comprehensive coverage and the retrieval of relevant papers, the search strings were expanded using Boolean operators (AND/OR). The complete list of the Boolean search strings executed in the PoP software is included in Appendix A. During the initial run of the software, 117 potentially relevant publications were identified.

3.2. Screening and Eligibility Assessment

After identifying papers using the PoP software, the PRISMA screening process was applied. First, duplicate records were removed, resulting in 96 unique publications. Subsequently, inclusion and exclusion criteria were applied. Only studies published in English and focused on SEA countries were considered; hence, 85 papers were subjected to title and abstract screening. Moreover, only journal publications and conference proceedings were considered, which reduced the eligible publications to 75 papers for full-text assessment. After a thorough full-text reading, only 62 publications were eligible and included in the final systematic review. The flow of the systematic review is demonstrated in Figure 2.

3.3. Quality Assessment

To ensure rigorous SLR implementation, each study underwent a formal quality assessment based on four criteria: (1) clarity of research objectives and AI application; (2) appropriateness of AI methodology; (3) presence of model validation, performance evaluation, or empirical testing; and (4) evidence of support to cleaner production or sustainability outcomes. Studies were not excluded based on quality score; instead, the researchers performed quality assessment to contextualize findings and interpret the maturity of AI methodologies across the literature [26].

3.4. Coding and Extraction

A structured coding protocol was developed to ensure consistency and reproducibility in data extraction and synthesis. Each paper was coded across four dimensions:

• AI methodology (e.g., regression-based models, supervised and unsupervised learning, deep learning, optimization-based approaches, explainable AI, NLP, and hybrid models).
• Functional role of AI categorized as predictive (e.g., forecasting, estimation, and risk prediction), prescriptive (optimization, decision support, and process improvement), and governance-level (e.g., policy analysis, regulation, ethical assessment).
• Application areas.
• Sustainability dimensions (environmental, economic, or social sustainability).

The mapping of the sustainability dimensions was operationalized based on their cleaner production or sustainable development outcomes based on a previous study [27].

3.5. Coding Validation

Initial coding was performed by two researchers and validated by co-authors who have expertise in AI and sustainability. Although formal inter-rater reliability statistics were not calculated, disagreements were resolved through iterative discussion to ensure conceptual consistency [26,28]. This collaborative process enhanced the robustness of the coding scheme while accommodating the interdisciplinary nature of the reviewed studies.

4. Results

Data visualization and analytical synthesis were conducted through Google Colab using Python-based workflows to enable transparent and reproducible analysis tailored to the scope and size of the dataset. This approach supported frequency distribution analysis, cross-tabulation of AI methods and sustainability dimensions, and comparative visualization of application areas, methodologies, and functional roles of AI.

4.1. Descriptive Analysis

The SLR process resulted in a total of 62 research articles that were critically analyzed following the PRISMA screening and eligibility assessment, as illustrated in Figure 3. Based on the PoP results, publications addressing AI-driven cleaner production and sustainable development in SEA began to grow from 2017 onwards. A gradual increase in publication volume can be observed in recent years, which indicates the growing academic interest in the intersection of AI and sustainability. Despite this increasing trend, it can also be noted that the notable growth in publications is observed from 2024 onwards, which reflects that research in this area is still in its initial but rapidly expanding stage.

With the growing interest in AI research within developing economies of the SEA region, Figure 4 demonstrates the distribution of reviewed studies across ASEAN member countries. Based on the descriptive results, Indonesia accounts for the largest share with 26 studies, followed by Malaysia (10) and Vietnam (9), which reflects their increasing engagement in AI research for sustainability context. Thailand and the Philippines each contributed five publications, while Singapore accounted for four studies. On the contrary, Brunei, Myanmar, and Laos had single publications each. Notably, no relevant research articles were retrieved from Cambodia and Timor-Leste based on the search strategy and inclusion criteria.

4.2. Prevalent Themes of AI Research in SEA

This section synthesizes the prevalent themes of AI research related to cleaner production and sustainable development in SEA. Based on the systematic coding, four recurring themes emerged: AI governance, climate change adaptation, operational efficiency, and social delivery, as illustrated in Figure 5.

4.2.1. AI Governance

The rapid evolution and technological volatility of AI require established regulatory frameworks, ethical guidelines, and mechanisms to build public trust, which makes AI governance a central research theme across several SEA countries, such as Vietnam, Malaysia, Indonesia, and Singapore. Existing studies underscore the challenges of aligning AI deployment with ethical accountability and societal values, particularly in centralized and single-party-political systems. For instance, research conducted in Vietnam reveals that although centralized governance facilitates rapid AI adoption, it likewise exposes limitations related to ethical flexibility, accountability, and public trust, particularly in areas such as facial recognition and public health [29]. These findings highlight tensions between state control and individual rights protection.

Furthermore, comparative studies reveal misalignment between global AI governance frameworks and national priorities in SEA. For example, research that examines the Global Partnership on Artificial Intelligence (GPAI) indicates that while Singapore aligns closely with international ethical standards, non-member countries such as Malaysia, Thailand, and Vietnam tend to prioritize human capital development and digital infrastructure expansion [30]. Moreover, structural barriers to global participation, such as restrictive membership criteria, language limitations, and unequal representation, were identified. Another comparative study explores the AI policies between Singapore and France, which highlights the importance of a polycentric governance approach that allows for localized AI implementation while adhering to a shared global ethical principle [31]. In another comparative study examining Indonesia, Malaysia, and the Philippines, the findings highlight differences in AI adoption patterns and sustainability priorities across the three countries [32]. Despite this flourishing institutional commitment to AI governance within the region, the reviewed studies suggest that governance frameworks are still at an early stage, which requires substantial support for further development.

4.2.2. Climate Change Adaptation

AI applications for climate change adaptation also emerged as a prevalent research theme, which is evident in environmentally vulnerable SEA countries such as the Philippines, Thailand, Malaysia, and Indonesia. Recent studies commonly employ machine learning models for environmental monitoring, climate forecasting, and disaster risk management. For example, research in Thailand applies machine learning models to examine the compounding influence of adverse weather conditions on air quality deterioration [33]. In Malaysia, predictive models are utilized to forecast trends in both precipitation patterns and surface air temperature fluctuations, which offers data-driven insights to support urban planning and climate risk mitigation strategies [34].

Likewise, AI-based climate modeling is employed in the Philippines to predict the impacts of future climate scenarios on rice yields, which offers a strategic tool for policymakers and farmers to better understand the complex and non-linear relationship between changing weather patterns and food security [35]. In contrast, a bibliometric study in Indonesia maps and visualizes climate change adaptation using AI and identifies a critical gap in the practical integration of advanced technologies, despite their significant potential for enhancing predictive accuracy and disaster preparedness [36]. In general, these studies show that AI research in climate adaptation remains largely predictive and diagnostic, with limited evidence of prescriptive or operational climate resilience strategies.

4.2.3. Operational Efficiency

AI-driven improvements in operational efficiency represent an emerging theme across manufacturing, logistics, and service sectors in SEA. An empirical study from Malaysia demonstrates the use of AI for demand forecasting, route optimization, warehouse automation, and process analytics that contribute to productivity and cost reduction [37]. Evidence from logistics and digital platform companies in Indonesia further suggests that AI adoption improves overall operational performance and competitive advantage, although measurable sustainability outcomes are not consistently established [38].

Further, in the context of small and medium enterprises (SMEs), particularly in the Philippines, qualitative findings indicate that AI is primarily leveraged for economic performance and operational efficiency, while its integration into environmental and social sustainable strategies remains in its nascent stages [39]. The reviewed studies highlight that although AI adoption offers clear efficiency gains, persistent challenges in the operational use of AI, such as high implementation costs, data security concerns, and shortages of digital skills, limit its broader contribution to cleaner production.

4.2.4. Social Delivery

Social delivery emerges as a distinct theme in the SEA region, which reflects the role of AI in enhancing public services, healthcare accessibility, workforce development, and social inclusion. In Vietnam and Indonesia, studies reveal that AI applications in recruitment systems, public administration, and regulatory governance have improved the overall efficiency and service delivery [40,41]. However, these benefits are constrained by limited technical expertise, data privacy concerns, and ethical governance gaps. One study in Indonesia highlights the adoption of a hybrid AI governance model, which integrates a principles-based approach with specific guidelines for various sectors, which is identified as a crucial step toward aligning AI deployment with social well-being and the attainment of the SDGs [42].

In healthcare, AI is examined as a viable tool to address systemic service delivery challenges. A review in the Philippines highlights the role of AI-powered screening tools and digital mental health platforms amid the vulnerability of the country to psychological stressors of climate change [43]. Similarly, research in Malaysia emphasizes the essential use of AI chatbots, virtual therapy platforms, and predictive analytics to mitigate shortages of healthcare professionals and enhance service accessibility in underserved communities [44]. Beyond healthcare, a meta-synthesis study in the Philippines provides a comprehensive overview of policy guidelines and implementation strategies in integrating AI into educational settings to foster inclusive and equitable learning environments [45].

4.3. Application Areas of AI in SEA Countries

AI applications in SEA span multiple sectors and industries and play a critical role in advancing cleaner production and sustainable development. Beyond economic development, these applications contribute to environmental protection and social well-being. Figure 6 shows the AI application areas across SEA based on the reviewed studies.

4.3.1. Public Governance and Smart Cities

AI applications in public governance and smart city initiatives are evident in Malaysia, Singapore, and Indonesia. In Malaysia, AI is embedded into its national modernization strategies such as the Putrajaya Smart City initiative and MyDigital blueprint, illustrating its role in optimizing urban planning, automating public services, and enhancing crisis response capabilities [46]. Predictive AI models are employed to optimize traffic flow, improve waste management efficiency, and enhance energy use in urban systems. In the public sector, machine learning is also applied to streamline administrative processes, such as automated tax filing and permit processing, thereby improving service accessibility and transaction efficiency. Likewise, another study underscores the important role of AI in advancing waste valorization and circular economy practices by improving waste sorting, optimizing recycling processes, and enhancing resource recovery to support high-quality waste management planning [47].

Singapore represents a regional benchmark for integrating AI with green technologies, which transforms the dense urban center into a global model for smart and sustainable cities [48]. AI revolutionizes municipal efficiency through eco-friendly transportation systems, sensor-equipped waste management, and real-time energy monitoring systems, which have a significant impact on carbon footprint reduction and cost efficiency. Moreover, this also covers the predictive capability of AI in assessing the maintenance needs of municipal assets, such as roads, water treatment facilities, and public transit systems.

In Indonesia, AI applications in public governance are linked to green social innovation, although evidence suggests that their effectiveness remains heavily dependent on a skilled workforce and education and training [49].

4.3.2. Healthcare and Public Health

AI integration with big data and digital health platforms has a substantial impact on healthcare delivery across several SEA countries. In Indonesia, a bibliometric review indicates a strategic shift from urgent pandemic responses toward long-term healthcare system optimization aligned with the goals of Health 5.0 [50]. The study also highlights the synergy between AI, big data, and telemedicine, which bridges geographical distance while enhancing diagnostic accuracy.

In Singapore, evidence of advanced clinical AI applications demonstrates substantial public health benefits, like the development of SELENA+, which analyzes retinal photographs using a deep learning system to detect eye conditions of patients and BRAIN, which analyzes the profile of the Singaporean population for holistic healthcare management [51]. Comparable applications are observed in Vietnam for their AI-enabled COVID-19 contact tracing systems [29] and in Malaysia through predictive analytics for hospital resource allocations such as beds, ventilators, and testing kits [46]. Further contributions to public health are the application of AI in revolutionizing mental health services in Malaysia and the Philippines [43,44].

4.3.3. Supply Chain and Logistics

AI is extensively applied in logistics and supply chain management to improve demand forecasting, route optimization, and warehouse automation. In Indonesia, predictive models integrate historical trends, weather data, traffic conditions, and consumer shopping patterns to facilitate on-time delivery scheduling and resource allocation, thereby reducing idle capacity and operational costs [52]. Furthermore, warehouse automation supported by AI and IoT technologies (e.g., intelligent picking, automated sorting, and precise inventory management) has improved inventory accuracy and reduced human error, which enhances monitoring of sensitive goods such as pharmaceuticals or temperature-controlled products.

In Vietnam, machine learning combined with simulation techniques is employed to estimate fuel consumption and optimize supply chain configurations, leading to greenhouse gas emissions reduction. Across the region, both large enterprises and SMEs leverage AI to enhance organizational efficiency and competitiveness [53].

4.3.4. Manufacturing and Industry

AI adoption in manufacturing is linked to Industry 4.0 initiatives, particularly in Malaysia. Empirical evidence from Malaysia highlights that while awareness of AI technologies is evolving among manufacturing firms and public-listed companies, actual implementation remains in its infancy stage [37]. Moreover, behavioral and legal challenges such as employee resistance and limited regulatory frameworks continue to hinder the transition toward intelligent manufacturing systems [54].

Likewise, AI is applied to SMEs to support recommender systems, inventory management, human resource management, logistics, and workflow optimization [39]. Although the benefits of AI are numerous and its impact can be felt across business areas in SMEs, it remains far-reaching. Additional applications include ride-sharing matching strategies and traffic routing, utilizing low-energy transport options such as bicycles to reduce delivery-related energy intensity, and strategic AI modeling to support sustainability-oriented decision-making. In the tourism sector, generative AI is employed as a tool to enhance itinerary planning, cost efficiency, and user experience [55]. Meanwhile, in Indonesia, AI is applied in True Cost Accounting (TCA) for energy projects through future scenario analysis and corporate decision-making in line with the national net-zero emission targets [56].

4.3.5. Agriculture and Food Security

Given the region’s predominantly agricultural landscape, the growing AI applications in agriculture and food security have become evident in recent years. In Indonesia and the Philippines, machine learning algorithms are applied to crop yield forecasting, climate risk assessment, and nutrient optimization to improve agricultural productivity and food security [35,42]. For example, the comparative analysis of machine learning models provides implications on the integration of granular data on soil quality, pest incidence, water usage, and farming practices. Moreover, AI-enabled aquaculture systems that integrate sensors and analytics provide real-time monitoring of fish health and environmental conditions. Additionally, generative AI applications in the Philippines support eco-friendly product design in SMEs, particularly in textile and woven industries involving local communities, social partners, and designers, which highlights AI role in natural resource conservation and inclusive development [39].

4.3.6. Finance and Banking

In the finance and banking sector, AI is increasingly applied to improve the accuracy and transparency of ESG reporting in Indonesia and Malaysia. A study conducted in Indonesia explores the intersection of knowledge management and the integration of ESG factors through machine learning, which reveals that leading banks tend to prioritize environmental sustainability, with eventual positive impacts on their revenues and financial performance [57]. In the case of Malaysia, AI strengthens audit quality and improves the reliability of climate-related disclosures [58]. Further research on Indonesia’s Islamic banking sector examines the application of AI in supporting the digital transformation of financial services while aligning with ethical and sustainability principles [59].

4.3.7. Cultural Preservation and Education

AI applications in cultural preservation and education represent an emerging yet underexplored dimension of sustainable development in SEA. Empirical investigation in Indonesia highlights the role of AI in e-learning applications and the sustainability of regional languages, which enhances accessibility and engagement [60]. Meanwhile, a study in the Philippines highlights the application of generative AI in empowering learners and fostering an inclusive language classroom [61]. In a case study in Thailand, researchers developed an educational recommendation system which utilizes AI chatbots that foster strategic learning while streamlining communication in the university [62].

4.4. Extent of AI Methodologies for Cleaner Production and Sustainable Development in SEA

The extent of AI methodologies to support cleaner production and sustainable development in SEA varies considerably across countries and sustainability dimensions. This section synthesizes the current methodological landscape of AI applications across the region and evaluates their alignment with cleaner production outcomes.

Figure 7 demonstrates that the AI methodologies employed across SEA are diverse, but at the same time unevenly distributed. Indonesia and Vietnam account for the widest range of AI techniques, which may be attributed to a higher publication share while reflecting broader methodological experimentation. In Indonesia, commonly employed techniques range from traditional supervised learning methods (e.g., regression models, Random Forest, and SVM) to more advanced approaches, such as clustering algorithms and explainable AI (e.g., SHAP Analysis). These methods are frequently applied to environmental monitoring, emissions analysis, biodiversity protection, and ESG assessment. Likewise, Vietnam shows the extensive use of ensemble learning and neural networks, often integrated with optimization algorithms (e.g., Particle Swarm Optimization), for environmental resilience and resource management.

Thailand also demonstrates notable AI methodologies, particularly in the application of deep learning and boosting algorithms (e.g., ANN and LightGBM) applied to environmental estimation tasks, such as PM2.5 concentration monitoring and energy efficiency assessment, which underscores the country’s focus on employing AI for climate-related and energy-oriented research. In contrast, Malaysia, the Philippines, and Myanmar primarily rely on supervised machine learning and regression models, particularly in studies related to corporate governance, climate analysis, and public sector. Meanwhile, the use of NLP and topic modeling in Brunei reflects a growing interest in AI governance and policy analysis, which demonstrates a shift in understanding qualitative data through machine learning and text mining.

In general, the extent of AI methodologies in SEA suggests a gradual transition from simple statistical modeling toward ensemble-based, hybrid, and explainable AI approaches. However, it is important to emphasize that the higher number of applied AI methods does not necessarily equate to greater methodological maturity or stronger cleaner production outcomes.

Table 1. Extent of AI Methodologies in SEA.

Year	Country	AI Methodology	Cleaner Production Outcome	Sustainability Dimension	Source
2024	Brunei	NLP, Topic Modeling	Legal and Regulatory Implications	Social	[30]
2020	Indonesia	Multinomial Logistic Regression, Support Vector Machine and Naïve Bayes	Biodiversity Protection	Environmental	[63]
2022	Indonesia	Computer Vision, YOLO V5	Health and Safety Enhancements	Social	[64]
2023	Indonesia	Gradient Boosting Regressor, Random Forest Regressor, AdaBoost Regressor	Emissions Reduction	Environmental	[65]
2024	Indonesia	Ensemble Machine Learning Classifier (EMLC) with the Synthetic Minority Oversampling Technique (SMOTE)	Community Well-being	Social	[66]
2025	Indonesia	Support Vector Regression, Random Forest, XGBoost	Biodiversity Protection	Environmental	[67]
2025	Indonesia	DBSCAN, NLP	Revenue Generation	Economic	[57]
2025	Indonesia	Gaussian Mixture Model (GMM), Random Forest classification, SHAP Analysis	Competitive Advantage	Economic	[68]
2025	Indonesia	Random Forest Regressor	Ethical and Equity Considerations	Social	[69]
2024	Laos	Interpretable Machine Learning with Remote Sensing Imagery	Enhanced Compliance with Environmental Regulations	Environmental	[70]
2019	Malaysia	Support Vector Machine (SVM)	Ethical and Equity Considerations	Social	[71]
2022	Malaysia	KNN, Naive Bayes, Decision Tree	Revenue Generation	Economic	[72]
2024	Malaysia	GPR, Ensemble of Trees, Neural Networks	Waste Minimization	Environmental	[47]
2025	Malaysia	Support Vector Regression (SVR), Random Forest Regression (RFR), Linear Regression (LR)	Biodiversity Protection	Environmental	[34]
2022	Myanmar	Random Forest (RF), Support Vector Machine (SVM), Classification and Regression Trees (CART)	Biodiversity Protection	Environmental	[73]
2024	Philippines	ANN, Random Forest (RF), Gradient Boosting (GB) Linear Regression (LR)	Biodiversity Protection	Environmental	[35]
2018	Thailand	Bootstrap ANOVA, Multiple Linear Regression	Cost Reduction	Environmental	[74]
2023	Thailand	ANN	Efficiency Improvements	Environmental	[75]
2023	Thailand	IDW-ANN	Biodiversity Protection	Environmental	[76]
2025	Thailand	Light Gradient Boosting Machine (LightGBM), SHAP Analysis	Emissions Reduction	Environmental	[33]
2022	Vietnam	Multivariate Adaptive Regression Splines (MARS), Random Forest Regression (RFR), Lasso Linear Regression (LLR)	Biodiversity Protection	Environmental	[77]
2022	Vietnam	SVM with Social Ski Driver Optimization (SSD), Fruit Fly Optimization (FFO), Sailfish Optimization (SFO), and Particle Swarm Optimization (PSO)	Biodiversity Protection	Environmental	[78]
2023	Vietnam	Catboost, Support Vector Machine, and Extratrees	Water Conservation	Environmental	[79]
2023	Vietnam	Radial Basis Function Neural Networks–Search and Rescue Optimization (RBFNN–SARO), Radial Basis Function Neural Network–Queuing Search Algorithm (RBFNN–QSA), Radial Basis Function Neural Network–Life Choice-based Optimizer (RBFNN–LCBO), Radial Basis Function Neural Network–Dragonfly Optimization (RBFNN–DO)	Biodiversity Protection	Environmental	[80]
2024	Vietnam	ANN	Cost Reduction	Economic	[53]

summarizes the AI methodologies applied across SEA countries and maps their corresponding cleaner production outcomes and sustainability dimensions.

In Brunei, NLP and topic modeling techniques are employed to examine the global AI governance landscape, which emphasizes the need for more inclusive and context-sensitive institutional frameworks to accommodate the dynamic and unique developmental goals of the SEA region, thereby contributing to social sustainability [30]. Although it contributes indirectly to cleaner production processes, legal and regulatory implications condition the adoption of AI for environmental and industrial practices.

Indonesia employs the most diverse AI methodologies for cleaner production outcomes. Supervised learning techniques (e.g., logistic regression, SVM, Naïve Bayes, Random Forest, Gradient Boosting, and XGBoost) are applied to predict earthquakes from historical data [63], identify factors contributing to greenhouse gas emissions in the palm oil industry [65], and predict daily maximum temperatures to manage the diverse microclimates of the country [67], which have a direct impact on biodiversity protection and emissions reduction. Additionally, computer vision and YOLO v5 are utilized for automated hazard detection in power plants [64], which enhance operational safety. Clustering and NLP techniques applied to ESG assessments and sustainability reporting in the banking industry improve revenue generation [57], while explainable AI methods (e.g., GMM, Random Forest, and SHAP analysis) support digital transformation and competitive advantage among MSMEs [68]. Moreover, ensemble classifiers with SMOTE for sentiment analysis of public tweets regarding UN SDGs [66], while Random Forest models are used to investigate the socio-economic drivers of gender inequality [69], which promote community well-being and ethical and equity consideration, respectively.

In Laos, interpretable machine learning with remote sensing imagery is applied to assess the spatiotemporal impacts of a large-scale infrastructure project such as the China-Laos Railway, which demonstrates the positive spatial spillover effects and its role as a catalyst for regional economic growth [70]. Moreover, this supports cleaner production pathways through enhanced compliance with environmental regulations.

Malaysia demonstrates a more direct link between AI applications and operational cleaner production outcomes. Gaussian Process Regression, ensemble trees, and neural networks are utilized to predict municipal solid waste generation [47], which provides quantitative inputs for waste minimization strategies and circular economy development. Further, classification models (KNN, Naïve Bayes, and Decision Tree) are applied to sustainable tourism demand forecasting (e.g., national parks and wildlife centers) [72], which supports resource-efficient management of protected areas. Additionally, regression-based and ensemble methods are applied to forecast precipitation trends and surface air temperatures [34], which are essential for sustaining cleaner production systems for overall climate change analysis. Meanwhile, another study employed supervised machine learning to examine the nexus between governance, financial performance, and social inclusivity to enhance corporate decision-making [71].

In Myanmar, machine learning techniques (e.g., Random Forest, SVM and CART) are employed to map land-use and land cover changes in gold mining districts, which reveal significant reductions in forest cover alongside a notable expansion of agricultural and mining lands [73]. These findings provide a diagnostic basis for enforcing biodiversity protection and mitigating industry impacts. Likewise, ANN, Random Forest, Gradient Boosting, and Linear Regression are employed in the Philippines to understand climate change impacts on rice crop yields [35], which supports food security and resource-efficient agricultural practices.

Thailand exhibits a strong support for cleaner production outcomes through AI applications. For instance, statistical and machine learning techniques, such as Bootstrap ANOVA and Multiple Linear Regression, are employed to investigate the energy efficiency of photovoltaic rooftop systems [74], which generate actionable insights for potential investors on long-term energy performance and emissions reduction. Moreover, ANN is combined with quantum chemistry techniques to construct catalytic materials databases to enhance the efficiency and sustainability of gas and coal-fired power plants [75], which represents a process-level cleaner production intervention. Another study proposed a hybrid IDW-ANN framework, which aims to further enhance localized climate modeling [76], thereby supporting adaptive infrastructure and energy planning that reduces resource inefficiencies and climate-related losses.

In Vietnam, AI applications for cleaner production outcomes span both operational efficiency and resource optimization. For example, ANN is combined with discrete event simulation to model fuel consumption in aquaculture supply chains [53], which enables logistics optimization and emissions reduction. Further, various AI methods such as MARS, Random Forest, and Lasso Linear Regression are used to estimate land-use change and mitigate inefficient urban expansion [77], while hybrid AI-metaheuristics models are used to quantify flood risk due to the expansion of impervious residential surfaces and increased rainfall [78]. Aside from this, machine learning and remote sensing using Catboost, SVM, and extra trees are utilized to assess inundation susceptibility [79], and deep learning techniques are used to assess the impact of climate change and land-use change on landslide susceptibility [80], thereby contributing to biodiversity protection and water conservation and larger cleaner production outcomes of reducing disaster wastes and resource loss.

Generally, AI applications in SEA demonstrate a growing emphasis on cleaner production outcomes, particularly in emissions reduction, energy efficiency, waste management, and resource conservation.

4.5. Cross-Tabulation of AI Methodologies and Sustainability Dimensions

The cross-tabulation of AI methodologies across sustainability dimensions is visualized in Figure 8.

Based on the results, AI applications towards environmental sustainability are evident in SEA countries, which utilize nearly every AI method. Environmental sustainability refers to the protection and efficient management of natural resources to ensure ecological integrity and long-term environmental viability [81]. In the context of cleaner production and sustainable development, this dimension emphasizes emissions and pollution reduction, resource efficiency, climate resilience, and circularity [82]. Across SEA, environmental sustainability constitutes the most dominant dimension addressed by AI applications. In particular, AI models, such as regression models, Random Forest, and neural networks are most frequently applied to predictive and diagnostic tasks, which include flood susceptibility assessment, land-use change analysis, and temperature prediction. The application of optimization algorithms and hybrid AI approaches also indicates that research toward this dimension is transitioning to prescriptive AI solutions, especially in areas related to energy efficiency, emissions reduction, and resource management.

Economic sustainability reflects long term resilience, productivity, and competitiveness without compromising environmental and social sustainability [83]. In the context of cleaner production, it is characterized by resource-efficient value creation, cost optimization, and sustainable business performance. In SEA, AI is viewed as a strategic enabler of economic sustainability, especially in emerging markets constrained by structural and resource limitations [84]. This review indicates that economic sustainability studies adopt a more exploratory approach, which employs clustering and explainable AI techniques to analyze ESG insights in banking and assess digital transformation performance among MSMEs. Evidently, the “Other AI/ML” category is the most frequently observed, which signifies the development of hybrid models designed to capture complex economic systems in emerging markets. Despite these strategic and managerial insights, their link to measurable cleaner production remains indirect in most cases.

Social sustainability focuses on improving quality of life, equity, inclusion, workforce development or even cultural preservation [85]. In the intersection of AI, cleaner production and sustainable development, it encompasses equitable access to services, human capital empowerment, and the preservation of social systems and cultural integrity [86]. Compared with environmental and social dimensions, AI applications addressing social sustainability in SEA are relatively lower in frequency and exploratory. In most cases, NLP and text mining are commonly employed to analyze unstructured data related to gender inequality, corporate social responsibility, labor dynamics, and AI governance. This demonstrates the role of AI in interpreting human behavior and policy impacts within the dimensions of sustainable development. However, most AI applications toward social sustainability focus more on governance and lack direct integration with cleaner production processes.

4.6. Methodological Maturity and Functional Roles of AI Applications

AI applications are classified according to their functional roles, namely, prescriptive, predictive, and governance-level AI, as summarized in

Table 2. Synthesis of AI Functional Roles, Methods, and Sustainability Dimensions.

AI Functional Role	AI Methodologies	Sustainability Dimension	Methodological Maturity
Predictive AI	Deep learning, Random Forest, SVM, regression models	Environmental	Moderate to High
Prescriptive AI	Hybrid AI-optimization, decision support systems	Environmental, Economic	Moderate
Governance-Level AI	NLP, Explainable AI	Social, Economic	Low

. This classification allows for the assessment of methodological maturity and the extent to which AI contributes to cleaner production.

The analysis indicates that predictive AI dominates the literature, which accounts for most of the reviewed studies. These applications primarily focus on forecasting environmental indicators, such as emissions, climate variables, and land-use change, which also classify sustainability-related patterns and estimate risk or impact metrics. Despite the essential and valuable contribution of predictive models for decision-support systems, they do not directly translate into operational interventions for cleaner production.

Meanwhile, prescriptive AI applications that support optimization, decision-making, or process improvement remain comparatively limited. Based on the literature, these studies typically employ hybrid AI–optimization frameworks or decision-support systems aimed at improving resource allocation, energy efficiency, or logistics performance. These applications show greater alignment with cleaner production principles; however, their methodological maturity remains moderate due to limited empirical validation and sector-specific implementation.

Governance-level AI applications, such as policy analysis, regulatory monitoring, and ethical or institutional assessment using NLP or explainable AI, are the least represented and remain largely exploratory. Although these applications do not directly optimize production processes, their role is essential in shaping regulatory readiness and institutional trust for larger diffusion of cleaner production technologies.

Given all these, in terms of methodological maturity, the majority of the studies relied on regression-based models and conventional machine learning approaches. Although recent studies have increasingly employed ensemble learning, deep learning, hybrid architectures, and explainable AI methods, this transition is uneven across countries and application areas. Thus, diversity in methodologies does not necessarily imply higher maturity but reflects fragmentation, where diverse algorithms are applied without standardized validation.

Further, in terms of model evaluation practices, this review reveals substantial variability. While some studies report cross-validation, many lack comparative benchmarking and robustness checks (e.g., reliance on single dataset evaluation, limited details on dataset size). These limitations constrain reproducibility, thereby restricting the operational deployment of AI solutions for cleaner production across sectors and national contexts.

4.7. Barriers and Challenges to AI Applications in SEA

Despite the growing adoption of AI across SEA, its effective deployment for cleaner production and sustainable development is hindered by a range of interrelated barriers. Based on the synthesis of the reviewed studies, four major categories of challenges are identified—(1) technological infrastructure and data ecosystems, (2) human capital and workforce readiness, (3) governance, policy, and ethics, and (4) financial and economic constraints—as illustrated in Figure 9.

4.7.1. Technological Infrastructure and Data Ecosystems

Countries in the SEA region, particularly developing economies, face persistent challenges related to limited technological infrastructure and data ecosystems. For instance, SMEs in the Philippines report unreliable connectivity and insufficient IT infrastructure as major constraints, which limit their capability to collect, store, and process customer data for data-driven AI sustainability applications. Similarly, the availability and quality of data remain critical prerequisites for the development of robust AI models for high-resolution environmental data [39]. Further, another challenge includes limited historical climate data, which has compelled researchers to rely on downscaling techniques rather than direct historical observations [76], which reduces model robustness for cleaner production planning.

In Indonesia, the implementation of TCA for energy and environmental projects is impeded by fragmented data sources across government agencies, private sector records, and environmental monitoring systems that operate in silos [56]. Similar issues emerged in the palm oil industry, where emissions data are often private and difficult to obtain [65]. In the banking sector, legacy systems remain incompatible with modern AI and big data analytics, emphasizing limited data accuracy and interoperability [87].

In Vietnam, limited transparency and public disclosure of AI-related policy documents undermine institutional legitimacy and public trust [29], thereby constraining data sharing for scaling AI solutions for cleaner production.

4.7.2. Human Capital and Workforce Readiness

The adoption of advanced AI technologies for cleaner production necessitates a workforce equipped with specialized technical skills; however, evidence shows a critical shortage of AI talent across the region. In Malaysian manufacturing, insufficient expertise in data analytics is identified as a major barrier, especially in automated forecasting that supports continuous production optimization [37]. Many companies report the absence of AI specialists, which limits their capacity to maximize innovation through AI.

Indonesia faces a similar challenge, as skilled professionals often migrate from the public to the private sector to receive higher compensation. Hence, this issue constrains innovation capacity within government units and poses a significant impediment in realizing the effective use of AI for public service improvement and sustainability initiatives [14].

Aside from skill shortages, digital literacy gaps, and workforce resistance further limit AI adoption. In Indonesia’s Islamic banking sector, resistance among senior employees has become evident, which includes chatbots and digital verification tools, due to their fear of job displacement [59]. In Vietnam, HR departments face significant challenges due to the lack of technical skills and expertise among their teams, alongside their concerns about data privacy, cybersecurity risks, high implementation costs, and the misalignment with existing organizational processes [40]. These barriers hinder the integration of AI into production workflows necessary for cleaner production.

4.7.3. Governance, Policy, and Ethics

Governance, policy, and ethics further constrain full AI implementation for cleaner production, as regulatory fragmentation remains evident across the region. In Indonesia, despite the substantial progress in establishing ethical guidelines and national AI strategies, gaps persist in AI-specific legislation, enforcement mechanisms, and institutional oversight [36,88]. The absence of a unified AI law has contributed to fragmented governance, particularly in complex scenarios involving autonomous AI decision-making, algorithmic bias, or distributed accountability in supply chains [89]. Moreover, there is also existing divergence between national and local governments that impedes the formulation and implementation of AI policies, as local AI proposals are often rejected due to limited capacity and resource constraints [14]. Similar issues are evident in Malaysia, where AI lacks harmonization across agencies, which limits data sharing while maintaining confidentiality and compliance [32]. Another study highlights the importance of national policies to improve efficiency in public sectors, safety and management alongside information technology literacy and readiness of population toward an improved urban service [90].

Ethical concerns and trust deficits also pose substantial barriers. The intersection between digital transformation and corporate social responsibility in SMEs emphasizes that although advanced technologies such as AI enhances competitiveness, enterprises should be vigilant with their social and ethical obligations [91]. Likewise, disparities in public trust are evident between urban and rural populations, with lower trust among rural communities attributed to limited exposure and understanding of AI systems [29]. Such trust deficits reduce social acceptance and limit the deployment of AI-driven systems essential for cleaner production implementation.

4.7.4. Financial and Economic Constraints

Financial barriers remain a crucial constraint, particularly given the significant economic divide between developed and developing nations across the SEA region. For instance, in Indonesia, micro and ultra-micro enterprises (MUMEs) report that the high cost of AI tools prohibits them from adoption, which also restricts participation in the digital economy aside from other inherent challenges including low levels of digital literacy and a lack of understanding of AI’s benefits [92,93]. This has widened the gap between large enterprises with sufficient capital and small-scale counterparts that are unable to participate in AI-driven sustainability initiatives.

In Malaysian manufacturing, high investment costs and long payback periods discourage AI adoption, as decision-makers often perceive uncertain long-term benefits relative to research, development, and implementation costs [37]. A similar scenario is observed in the Philippines, where SMEs prioritize immediate financial stability over long-term investments in AI capabilities for cleaner production [39]. In Laos, large-scale infrastructure projects have also contributed to uneven regional development, which is heavily influenced by the leading industries of parent cities, further reinforcing economic disparities [70].

4.8. Future Research Directions

Based on the synthesized evidence, the application of AI for cleaner production and sustainable development in the SEA region has started to thrive. While the reviewed literature highlights existing challenges and barriers to AI implementation, this study also reveals opportunities and research directions for scholars, policymakers, and practitioners in advancing AI in supporting sustainability outcomes across SEA. The future research directions emerging from the reviewed studies are illustrated in Figure 10.

4.8.1. Advancements of AI Methodologies

The extent of AI methodologies employed in SEA encompasses all three sustainability dimensions and a wide range of industries, which have gradually transitioned toward hybrid and advanced approaches. Future research should therefore prioritize the advancement of AI methodologies by moving beyond basic predictive models toward adaptive and prescriptive AI frameworks that directly support cleaner production decision-making.

This transition is already evident in other SEA countries. In Thailand, interpolation techniques are combined with machine learning models (e.g., IDW-ANN) to increase spatial resolution and overcome data scarcity. Likewise, in Indonesia, spatially adaptive forecasting frameworks integrate clustering techniques (e.g., K-Means) with machine learning models such as SVR, Random Forest, and XGBoost to predict daily maximum temperatures across the country. Future studies should further expand the application of explainable AI techniques, such as SHAP, to enhance interpretability, informed decision-making, transparency, and improved predictive accuracy, which are useful for digital transformation assessment, environmental monitoring, and sustainability reporting.

Moreover, greater emphasis should be directed toward deep learning architectures, such as Long Short-Term Memory (LSTM) networks, to better handle spatiotemporal dependencies and sequence noise in climate-related forecasting tasks, such as temperature and rainfall prediction in Malaysia. Another promising research direction involves the integration of bio-inspired optimization algorithms (e.g., Fruit Fly Optimization, Particle Swarm Optimization, and related metaheuristics) with machine learning models to improve flood and landslide susceptibility predictions in climate-vulnerable regions such as Vietnam. These advances are critical for transitioning AI applications from diagnostic tools to operational systems that enable cleaner production planning and risk mitigation.

4.8.2. Data Quality and Technology Enhancement

The effectiveness of AI applications for sustainability and cleaner production relies heavily on the availability of high-quality and massive volumes of data. In view of this, future research should focus on improving both data quality and technological infrastructure across SEA.

One priority area is the utilization of higher spatial resolution satellite imagery (e.g., Sentinel-2A) combined with object-based classification methods to enhance land-use and deforestation analysis. Another crucial direction involves the development of national standards for TCA, particularly in Indonesia, by leveraging emerging technologies such as blockchain and big data analytics to improve data traceability and transparency in energy and production systems. Additionally, future research should promote the transition toward automated monitoring systems to reduce human error and enhance real-time data availability, as demonstrated by AI-enabled air quality monitoring systems for PM2.5 concentrations in Thailand. These technological enhancements are essential for supporting continuous improvement and evidence-based cleaner production strategies.

4.8.3. Governance and Policy Frameworks

Robust governance and policy frameworks are critical enablers of the long-term and responsible deployment of AI for cleaner production and sustainable development. Future research should examine the effectiveness of sector-specific versus horizontal regulatory approaches in governing AI applications across environmental, economic, and social dimensions of sustainability.

Further exploratory studies are necessary to assess the alignment of global ethical frameworks promoted by GPAI and regional priorities in SEA, which often emphasize human capital development over AI ethics. Research should also investigate the strategies in the ASEAN regional guidelines and global standards to facilitate cross-border cooperation, interoperability, and data sharing. Moreover, studies should also focus on the role of regulatory sandboxes and risk-based governance models, like in Indonesia, in balancing innovation with public protection and ethical standards. Beyond formal regulation, future research should explore community-led and participatory AI-development models, particularly those that have subsequent impact on the preservation of cultural diversity and protection of indigenous peoples to ensure AI systems remain socially embedded and responsive to local needs.

Longitudinal studies that examine the evolution of public trust in AI over time are also necessary, particularly in such as in the context of Vietnam. Additionally, comparative analyses between innovation-driven governance models, like in Singapore, and ethics-centric or authoritarian models, such as in France and Vietnam, would also enrich understanding regulatory effectiveness, social acceptance, and cleaner production outcomes.

4.8.4. Sector-Specific Applications for Sustainability

Given the diverse economic structures and sustainability challenges across SEA, future research should focus on sector-specific and context-sensitive AI applications. For instance, the investigation of abandoned properties in Vietnam could be extended to assess their long-term economic effects and their contributions to regional sustainability in supporting circular economy transitions.

In healthcare, future research should explore the role of AI for mental health interventions such as empathic chatbots and predictive analytics, like the case of the Philippines and Malaysia, to address climate-related psychological stress, while ensuring inclusivity and avoiding widened digital divides. In agriculture, integrating granular datasets on soil quality, water usage, and pest incidence can significantly improve crop yield forecasting and climate resilience strategies.

Behavioral and social sustainability also warrant greater attention. Future studies should examine the influence of AI in influencing greener lifestyles and sustainable consumption among Generation Z consumers, particularly in Indonesia. Moreover, longitudinal research on the impact of AI on job displacement and the effectiveness of reskilling programs is essential for assessing long-term social sustainability implications, as noted in the Philippines and Vietnam.

5. Discussion

This systematic review synthesizes evidence from 62 studies to examine the role of AI in supporting cleaner production and sustainable development across SEA. This review critically investigates AI applications across sectors and sustainability dimensions, methodological maturity, and functional roles of AI, and its alignment to cleaner production outcomes.

First, the findings confirm that AI applications in SEA are mostly employed to address environmental sustainability, particularly in climate forecasting, pollution monitoring, land-use analytics, disaster risk management, and energy system monitoring. This pattern can be attributed to the high vulnerability of the region to the impacts of climate change, such as flash floods, haze pollution, deforestation, and extreme weather phenomena, and the relative availability of environmental datasets compared to industrial or supply chain data. However, the review reveals that most environmental applications remain diagnostic and predictive, which support monitoring, early warning, and scenario analysis, rather than directly enabling cleaner production outcomes. Prescriptive AI applications, such as optimization algorithms for emissions reduction, material efficiency, or circular production, remain limited. This suggests that AI for cleaner production in SEA appears to be at a transitional stage, where it is often employed to understand and predict environmental problems rather than systematically address waste emissions at the process or system level. Likewise, the contribution of AI in realizing social and economic sustainability is also in early stages, which underscores the need for integrated frameworks that address economic viability and social well-being alongside environmental goals.

Second, the review highlights significant disparities in the extent of AI methodologies employed across SEA countries. Based on the studies included in this review, Indonesia, Vietnam, and Thailand employed a wide range of AI techniques while addressing multiple sustainability dimensions. Although fewer studies were identified for Singapore, evidence from case studies demonstrates advanced AI deployment, particularly in urban systems and governance-related applications. In contrast, countries with less developed research and industrial capacities tend to apply AI in a more exploratory way, often focusing on governance and policy analysis using qualitative or NLP-based approaches. Meanwhile, no studies were retrieved from Cambodia and Timor-Leste, which demonstrates regional development gaps. These findings suggest that the maturity of AI applications for cleaner production in SEA appears to be uneven, which may be attributed to industrial differences, research capacity, data availability, and national priorities. Hence, this implies the need for coordinated regional strategies within ASEAN countries to reduce fragmented governance and support knowledge transfer.

Third, a notable contribution of this review is related to the gradual transition in SEA from conventional machine learning toward more advanced, hybrid, explainable AI approaches. The growing application of ensemble learning and optimization algorithms offers increasing scholarly attention in addressing complex sustainability challenges in the region. Moreover, the integration of clustering with regression or the combination of machine learning and bio-inspired optimization is essential for the SEA context due to the presence of data scarcity, spatial heterogeneity, and non-linear environmental processes. More importantly, the application of explainable AI for cleaner production supports transparency and interpretability, which is important among policymakers, stakeholders, and the public, especially in applications related to environmental regulation, ESG reporting, and resource management. Nevertheless, the review also shows that the validation of cleaner production outcomes (e.g., quantified energy saving or emission reductions) through advanced algorithms remains limited and highly context-specific; thus, this reinforces the need for closer integration between AI development and industrial performance assessment.

Fourth, the findings reveal barriers and challenges that constrain AI adoption for cleaner production and sustainable development in SEA. Limitations in technological infrastructure and data ecosystems remain prominent, especially in countries with developing economies, which restrict the deployment of robust AI models capable of supporting life-cycle analysis, real-time process monitoring, and production optimization. Human capital and workforce readiness, particularly among SMEs, further limit adoption, despite the pivotal role of SMEs in regional economies and sustainability transitions. In addition, fragmented governance and the absence of unified AI policy frameworks also hinder the wider adoption of AI in SEA, which undermines trust and slows diffusion across sectors, and hinders the integration of AI into supply chains, sustainability reporting, and cleaner production practices.

Importantly, the review also identifies emerging but underdeveloped AI applications in waste valorization, circular economy, and resource recovery. Several studies report the integration of AI into waste classification, recycling optimization, material flow analysis, and decision-support for reusing resources. These AI applications indicate a significant potential in transforming waste streams into value-added resources while reducing the reliance on non-renewable resources to support circulation production systems. However, most applications from the reviewed studies are at the initial stage, with limited deployment and scarce use of indicators for waste recovery and lifecycle analysis towards circular economy. Hence, this gap underscores the need for future research to integrate AI in waste management, life-cycle assessment, and material recovery, especially for developing SEA economies.

Finally, this review offers important theoretical and practical implications. From a theoretical perspective, this review extends the cleaner production and sustainable development literature by elucidating the shift in AI research from environmental monitoring toward more advanced predictive and prescriptive sustainability strategies. The study highlights that the current AI applications for cleaner production and sustainable development landscape in SEA are shaped by national capacity, governance structures, and human capital readiness. Consequently, the study underscores the importance of developing AI sustainability frameworks that support the application of explainable and hybrid AI models, promote regulatory sandboxes for green AI investigation, and strengthen regional collaboration among ASEAN nations to reduce disparities in maximizing AI application. In terms of practical implications, this study offers industries, especially SMEs, to leverage interpretable AI solutions to improve resource efficiency, reduce emissions, and enhance operational resilience. At the same time, it confirms that cleaner production requires future algorithm-specific, industry-level empirical studies that account for production scale, technological configuration, automation level, and life-cycle energy and emissions performance. Generally, the review serves as a benchmark study for advancing AI applications that support waste valorization, circular economy implementation, and broader cleaner production and sustainable development pathways in SEA.

6. Conclusions and Limitations

AI offers substantial potential to advance cleaner production and sustainable development across the SEA region. This study conducted an SLR of 62 studies examining AI application across the environmental, economic, and social sustainability dimensions within the diverse regional context of SEA. To realize the objectives of the study, this study integrated Harzing’s PoP software with the PRISMA framework to identify prevalent research themes, application areas, the extent of AI methodologies as well as their maturity and functional roles, barriers and challenges, and future research directions for the continuous advancement of AI application in the region. The findings indicate that a large portion of AI applications in SEA are directed toward environmental sustainability. While these present meaningful contributions to environmental protection, most AI applications for cleaner production and sustainable development outcomes remain diagnostic and predictive, with limited evidence of using AI toward prescriptive and hybrid approaches for production systems and supply chains. Moreover, substantial disparities exist across SEA countries, which reflect uneven digital infrastructure, research capacity, and policy readiness. To fully realize its potential for cleaner production and sustainable development, AI applications must evolve toward more advanced applications through hybrid modeling, optimization algorithms, and explainable AI approaches that enhance transparency and trust. These findings further offer actionable insights for policymakers and practitioners by identifying priority sectors, methodological trends, and structural barriers that influence AI application for sustainability in emerging economies. In general, this study underscores that AI holds significant potential to support cleaner production and sustainable development in SEA; however, this requires coordinated advancements in data ecosystems, human capital, governance frameworks, and regional collaboration among ASEAN member states.

Despite its contributions, this study has several limitations that should be acknowledged. First, the literature review is primarily based on Harzing’s PoP software, which is sensitive to keyword selection and indexing practices; consequently, relevant studies employing alternative terminology may not be captured during the retrieval process. Second, the distribution of reviewed studies across SEA countries is uneven, which may reflect disparities in research output, data availability, and publication visibility. This limits the generalizability of the findings, particularly for underrepresented countries such as Cambodia and Timor-Leste, for which no eligible studies were retrieved. Third, although formal inter-rater reliability statistics were not computed, all screening and coding decisions were collaboratively validated by the authors to enhance consistency and reduce subjective bias in the classification of AI methodologies, application areas, and cleaner production outcomes. Finally, as a systematic review, this study synthesizes reported evidence rather than empirically validating cleaner production outcomes, and the assessment of AI effectiveness is therefore dependent on the rigor and scope of the original studies. Future research may address these limitations by conducting reviews on specific sectors or countries, expanding database coverage, and empirically examining emerging AI techniques and governance frameworks that continue to shape cleaner production and sustainable development in the SEA region.