ERP and Artificial Intelligence as Transformative Technologies for SME Sustainability: A Socio-Technical Systems Perspective

Small and medium-sized enterprises (SMEs) are a fundamental pillar of the global economy and play a critical role in advancing sustainable development. They constitute the majority of firms worldwide and contribute significantly to employment, productivity, and economic resilience, especially in emerging economies [1, 2]. Their size and systemic relevance make SMEs essential in meeting sustainability goals, including responsible production, environmental performance, and inclusive growth [3-5]. Despite their importance, SMEs face persistent challenges in translating sustainability goals into stable, measurable outcomes.

A substantial body of literature indicates that sustainability challenges in SMEs are structural and multidimensional. Financial constraints, high implementation costs, and uncertainty about returns on sustainability investments limit the intensity of adoption [6-9]. Technological barriers, including low digital readiness, fragmented systems, and limited access to appropriate technologies, constrain process efficiency and environmental performance [10-13]. Organizational challenges, such as weak strategic alignment, informal structures, and limited process formalization, further reduce the scalability and consistency of sustainability initiatives [14-17]. In addition, knowledge-related limitations, including low sustainability literacy and weak learning capability, hinder effective decision-making and adaptive capacity [18-21]. Institutional factors, such as regulatory complexity, weak stakeholder support, and limited inter-organizational collaboration, further constrain the integration of long-term sustainability [22-25]. These barriers interact cumulatively rather than independently, leading to persistent inefficiencies and inconsistent sustainability outcomes [16, 25-27].

In response to these challenges, digital technologies have increasingly been positioned as enablers of sustainability transformation in SMEs. Enterprise resource planning (ERP) systems provide capabilities related to process integration, data standardization, and operational control, enabling improved coordination, traceability, and resource management [28-30]. Artificial intelligence (AI), in contrast, offers advanced analytical capabilities, including pattern recognition, prediction, optimization, and adaptive learning, supporting more proactive and data-driven decision-making [31-33]. Empirical studies suggest that these technologies can improve operational efficiency and sustainability performance when appropriately implemented [34, 35].

However, despite increasing digital adoption, SMEs continue to experience recurring implementation challenges. Advanced technologies such as AI are often introduced without sufficient organizational readiness or a reliable, integrated data infrastructure. Existing studies frequently examine ERP and AI as independent or parallel technological solutions, with ERP research focusing on integration and control, while AI studies emphasize analytics and automation [29, 35, 36]. This separation obscures a critical structural issue: AI applications depend on consistent, integrated data environments, which are typically established through ERP systems, while ERP systems alone are insufficient to support adaptive, learning-oriented decision-making. Consequently, current literature provides a limited explanation of how these technologies interact within a structured sustainability transformation process.

More importantly, prior research often conceptualizes digital transformation as a discrete adoption event rather than a staged, conditional process [37, 38]. In practice, SMEs frequently adopt digital technologies in response to external pressures without sufficient alignment with organizational capabilities, leading to partial implementation, underutilization, and limited strategic impact. These patterns suggest that the effectiveness of digital technologies is not inherent to the technologies themselves, but depends on how they are sequenced and embedded within socio-organizational conditions.

From a socio-technical systems (STS) perspective, organizational outcomes emerge from the interaction between social subsystems (such as leadership, defined as the process of influencing organizational direction; organizational capability, referring to the collective skills and processes that enable organizations to achieve their goals; and decision-making structures, which are the systems that shape how organizational choices are made) and technical subsystems (such as digital infrastructures, defined as the technological backbone supporting digital activities, and analytical tools, which are software or systems supporting data analysis) [39-41]. However, while this perspective provides a strong theoretical foundation, its application in SME sustainability research remains largely technology-neutral, offering a limited explanation of how specific technologies structure transformation processes or interact as distinct capability layers.

This study addresses these limitations by developing a Technology Transformation Model. This model conceptualizes SME sustainability as a structured socio-technical process. It is grounded in STS theory and operationalized through ERP and AI capability layers: socio-organizational conditions serve as the enabling foundation, ERP provides an integrative infrastructure, and AI operates as an adaptive intelligence layer built on ERP-enabled data environments.

The novelty of this study lies in its technology-specific and dependency-based transformation logic. Rather than introducing new constructs, this study redefines how existing constructs interact within the context of SME sustainability transformation. Specifically, ERP and AI are conceptualized as sequentially dependent layers rather than parallel technologies, where the effectiveness of AI depends on ERP-enabled data integration, and ERP effectiveness depends on prior organizational readiness. This dependency-based and threshold-conditioned perspective provides a structured explanation for recurring digitalization challenges and uneven sustainability outcomes in SMEs.

Methodologically, this study uses a systematic literature review (SLR) to synthesize evidence on sustainability barriers, ERP and AI capabilities, and socio-organizational conditions. Based on these findings, a structured expert assessment evaluates the conceptual coherence and practical plausibility of the proposed model.

This study contributes to the literature in three ways. First, it advances STS research by introducing a transformation model that defines the interdependent roles of ERP and AI. Second, it extends SME digital transformation research by shifting from static adoption perspectives to a process-based understanding. Finally, it offers actionable insights for policymakers and practitioners by providing a logic for sequencing digital investments and aligning capabilities to support sustainability.

3.1 Research design

This study adopts an SLR to develop an integrative understanding of sustainability transformation in SMEs. The review is designed to identify, synthesise, and integrate three interrelated domains: (1) sustainability barriers in SMEs, (2) ERP capabilities in SME sustainability, and (3) AI capabilities in SME sustainability.

The SLR follows PRISMA principles to ensure transparency, replicability, and methodological rigor. Unlike traditional single-theme reviews, this study employs a multi-theme integrative approach to construct a process-based explanation of how SMEs transition from sustainability barriers to digitally enabled sustainability outcomes.

Figure 1. PRISMA flow diagram

3.2 Data source and retrieval procedure

The primary data source was the Scopus database, selected for its comprehensive coverage of peer-reviewed literature in sustainability, information systems, and operations management. Search and retrieval were conducted using Publish or Perish, which facilitated:

1. structured query execution,
2. metadata extraction,
3. and dataset management.

The search was limited to publications between 2020 and 2026, reflecting the most recent developments in ERP and AI adoption in SMEs.

3.3 Search strategy and query formulation

The search was conducted in January 2026 using three separate queries applied to TITLE-ABS-KEY fields in Scopus.

Theme 1: Sustainability Barriers in SMEs

(("sustainability barrier*" OR "barrier* to sustainability" OR "sustainability challenge*" OR "constraint* to sustainability" OR "obstacle* to sustainable practice*") AND ("SME" OR "SMEs" OR "small and medium enterprise*" OR "small business*" OR "medium enterprise*") AND ("sustainability" OR "sustainable development" OR "environmental performance" OR "social performance" OR "economic sustainability"))

Theme 2: ERP and SME Sustainability

(("enterprise resource planning" OR ERP OR "ERP system*" OR "ERP implementation") AND ("small and medium enterprise*" OR SMEs OR "small business*") AND ("sustainability" OR "sustainable development" OR "environmental performance" OR "social performance" OR "economic sustainability" OR ESG))

Theme 3: AI and SME Sustainability

(("artificial intelligence" OR AI OR "machine learning" OR "deep learning" OR "data analytics") AND ("small and medium enterprise*" OR SMEs OR "small business*") AND ("sustainability" OR "sustainable development" OR "environmental performance" OR "social performance" OR "economic sustainability" OR ESG))

All queries were filtered using:

1. Publication year: 2020–2026
2. Language: English
3. Document type: articles and conference papers

The initial search yielded:

1. 200 records (Theme 1)
2. 54 records (Theme 2)
3. 200 records (Theme 3)

Total initial dataset: 454 records.

3.4 De-duplication and data cleaning

All records were exported and merged into a single dataset. Duplicate removal was conducted using:

1. DOI matching,
2. title similarity,
3. author and year verification.

After de-duplication, the dataset was reduced to approximately 410 unique records.

3.5 Screening and selection process

A three-stage screening procedure was applied:

Stage 1: Title Screening

Articles were excluded if they:

1. did not involve SMEs,
2. lacked sustainability context,
3. or were purely technical.

Stage 2: Abstract Screening

Articles were excluded if they:

1. focused only on large enterprises,
2. lacked organisational relevance,
3. or did not relate to ERP, AI, or sustainability barriers.

Stage 3: Full-text Screening

Full-text evaluation ensured:

1. conceptual relevance,
2. analytical depth,
3. and explicit linkage to sustainability outcomes.

3.6 Inclusion and exclusion criteria

Inclusion Criteria

1. Peer-reviewed Scopus-indexed publications
2. Explicit SME relevance
3. Coverage of at least one thematic domain
4. Linkage to sustainability outcomes.

Exclusion Criteria

1. Non-English publications
2. Editorials and non-academic documents
3. Studies focusing only on large firms
4. Purely technical studies without organisational context.

3.7 Core article selection criteria

A final filtering stage was conducted to identify the core analytical dataset (52 articles) based on:

1. Conceptual Relevance: Articles must explicitly address:

• sustainability barriers,
• ERP capabilities,
• AI capabilities in sustainability context.

2. SME Focus: Priority was given to SME-specific studies or those transferable to SMEs.
3. Mechanism-Oriented Contribution: Only studies explaining:

• why barriers occur,
• how ERP/AI operate,
• how outcomes emerge were retained.

4. Analytical Depth: Studies must provide empirical or strong conceptual contributions.
5. Non-Redundancy: Redundant studies were removed to maintain parsimony.
6. Integration Potential: Articles contributing to cross-theme integration were prioritised.

3.8 Final dataset composition

Table 1. Systematic literature review (SLR) final dataset

These 52 articles constitute the final analytical corpus (Table 1).

3.9 Data extraction and coding

A structured coding framework was applied across all selected articles. Each study was coded into:

1. Sustainability barriers
2. ERP capabilities
3. AI capabilities
4. Organisational conditions
5. Sustainability outcomes

Coding was conducted iteratively using constant comparison to ensure consistency and conceptual clarity.

3.10 Reliability and validation

To enhance robustness:

1. coding was iteratively refined,
2. discrepancies were resolved through analytical discussion,
3. categories were stabilised through repeated comparison until saturation.

3.11 Analytical synthesis approach

The synthesis followed a three-step approach:

1. Within-theme synthesis

• Barrier classification
• ERP capability identification
• AI capability mapping

2. Cross-theme integration: Linking organisational barriers, ERP systems, and AI capabilities
3. Model development

• Constructing a process-based transformation framework
• Explaining how SMEs transition from barriers to sustainability outcomes

3.12 Transparency and replicability

To ensure transparency:

1. full search strings are provided,
2. selection criteria are explicitly defined,
3. dataset composition is clearly reported,
4. and a PRISMA flow diagram (Figure 1) illustrates the selection process.

3.13 Expert-based conceptual assessment

To strengthen the conceptual robustness and practical relevance of the proposed Technology Transformation Model, an expert-based conceptual assessment was conducted following the SLR. This step aimed to evaluate the conceptual clarity, internal logical coherence, and practical plausibility of the model, particularly in representing the interdependent roles of socio-organisational conditions, ERP as an integrative infrastructure, and AI as an adaptive intelligence layer in SME sustainability transformation. This assessment serves as a structured conceptual appraisal rather than empirical validation, providing expert-informed support for the model’s integrity and applicability.

3.13.1 Expert panel composition

The expert panel consisted of 10 experts selected using purposive sampling to ensure balanced representation across the socio-technical spectrum of SME sustainability. The panel included:

1. 3 academic experts in sustainability, STS, and information systems,
2. 4 SME practitioners with direct experience in ERP and/or AI-enabled operations,
3. 3 policy and institutional experts involved in SME development and digital transformation programs.

All experts had at least five years of relevant professional experience and demonstrated familiarity with sustainability-oriented digital transformation in SMEs.

3.13.2 Assessment procedure

The assessment employed a single-round structured evaluation, appropriate for theory-driven conceptual model appraisal. Experts were provided with:

1. A visual representation of the proposed STS–ERP–AI model,
2. A concise explanation of each subsystem and interrelationship,
3. An evaluation instrument using a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Experts were asked to evaluate the model based on 7 criteria:

1. Does the proposed model appropriately place socio–organisational readiness as the primary foundation for SME sustainability transformation?
2. Is the conceptual role of ERP as an integrative infrastructure for process alignment, data consistency, and operational control clearly and logically defined in the model?
3. Is AI correctly positioned as an adaptive and analytical layer that builds upon ERP-enabled data structures rather than as a standalone technology?
4. Does the proposed transformation sequence from socio–organisational foundations to ERP adoption and subsequently to AI integration reflect a coherent and theoretically sound logic?
5. Do the interactions and feedback loops between social and technical subsystems adequately represent core STS principles?
6. Is the inclusion of external stakeholders and institutional support actors as contextual enablers conceptually relevant for SME sustainability transformation?
7. Does the model clearly link ERP–AI-enabled transformation to SME sustainability outcomes aligned with the Sustainable Development Goals (SDGs 8, 9, 12, and 13)?

The model was assessed using median and interquartile range (IQR) statistics. A model component was considered validated when:

1. the median score was ≥ 4.0, and
2. The IQR was ≤ 1.0, an acceptable consensus.

Qualitative comments from experts were used to refine terminology, clarify feedback loops, and improve the articulation of socio-organisational and technical interactions without altering the model's core structure.

Building on the results of the SLR synthesis and model assessment, the following discussion interprets the findings in relation to prior literature.

5.1 ERP and AI roles: Confirmed functions but repositioned structure

The findings of this study support prior research that identifies ERP and AI as important enablers of sustainability in SMEs. ERP has been consistently associated with process integration, data standardization, and operational control, enabling improved coordination and resource efficiency [29, 34, 52]. AI, in turn, contributes through predictive analytics, optimization, and adaptive decision-making, supporting more proactive sustainability practices [33, 58, 63]. These roles are also reflected in process-oriented studies, where integrated systems and analytical capabilities improve operational sustainability performance [70]. However, existing studies largely treat ERP and AI as parallel or functionally equivalent technologies. This assumption obscures a critical structural distinction. The present findings demonstrate that ERP and AI do not operate at the same analytical level. ERP functions as an infrastructural layer that stabilizes processes and data environments, whereas AI operates as an adaptive layer that depends on these structured conditions. This repositioning shifts the analytical focus from “what technologies do” to “how technologies are structurally organized,” providing a more precise explanation of their roles in sustainability transformation.

5.2 From complementarity to conditional dependency

The dominant view in the literature frames ERP and AI as complementary technologies that jointly enhance performance [29, 35]. While this interpretation is not incorrect, it is analytically insufficient. Complementarity assumes co-existence, but does not explain activation conditions. The findings of this study demonstrate that the relationship between ERP and AI is not merely complementary but conditionally dependent. Specifically, AI effectiveness depends on ERP-enabled data integration, while ERP effectiveness depends on socio-organizational readiness. This introduces a directional and staged relationship that is largely absent in prior work. Empirical evidence on digital culture further reinforces this interpretation, showing that technological systems yield outcomes only when aligned with organizational context and behavioral conditions [71]. By converting complementarity into dependency, this study advances the literature from a static to a process-based understanding of digital transformation.

5.3 Organizational readiness as a determinant, not a moderator

Prior studies consistently highlight organizational readiness, leadership, and human capability as important factors in digital transformation [15, 37]. However, these factors are typically treated as contextual variables or moderators. The present findings challenge this positioning by demonstrating that organizational readiness is a determinant that shapes the viability of the entire transformation pathway. This interpretation is consistent with research on organizational sustainability, which emphasizes the role of internal systems, culture, and human resource practices in enabling long-term performance [72]. It is also aligned with conceptual frameworks that identify internal capability alignment as a prerequisite for sustainable SME transformation [73]. The contribution of this study lies in elevating organizational readiness from a supporting factor to a structural prerequisite that conditions both ERP institutionalization and AI effectiveness.

5.4 Threshold logic: Explaining why digital transformation often fails

This realignment of organizational readiness sets the stage for a deeper exploration of one of the central contributions of this study: the introduction of a threshold-based interpretation of digital transformation. Prior research often assumes that technology adoption leads to improved sustainability outcomes, although empirical findings remain inconsistent [7, 16]. The present findings provide a clearer explanation of this inconsistency. Digital technologies do not generate outcomes automatically; they become effective only when minimum organizational and infrastructural conditions are achieved. This threshold logic is consistent with evidence showing that SMEs face significant challenges in implementing sustainability systems, including reporting, governance, and institutionalization barriers [74]. More importantly, this perspective explains a recurring empirical anomaly: why similar technologies produce divergent outcomes across SMEs. The answer lies not in the technology itself, but in whether the required threshold conditions are met. This reasoning also helps explain the variability reported across the literature, bringing us to the question of transformation sequencing.

5.5 Explaining variability in prior findings through transformation sequencing

The literature on digital transformation in SMEs frequently reports uneven and sometimes contradictory outcomes, particularly in relation to sustainability performance and operational improvement, where similar technologies yield divergent results across contexts [38, 69, 75]. These inconsistencies are commonly attributed to contextual factors such as resource constraints or digital maturity, yet they remain insufficiently explained at a structural level. The findings of this study suggest that such variability is not merely contextual but fundamentally conditioned by transformation sequencing. Specifically, sustainability outcomes depend on whether socio-organisational readiness, ERP integration, and AI capability development occur in an aligned and sequential manner. When this sequence is disrupted, digital technologies operate below their functional potential, resulting in fragmented or unstable outcomes. By introducing sequencing as an explanatory mechanism, this study reframes previously fragmented findings into a coherent transformation logic, shifting the analytical focus from whether digital technologies work to under what conditions and in what sequence they generate meaningful sustainability outcomes.

5.6 Advancing socio-technical systems toward technology-specific modelling

STS theory emphasises that organisational outcomes emerge from the interaction between social and technical subsystems, including organisational capabilities and technological infrastructures [40, 41]. While this perspective has been widely applied in SME sustainability research, its operationalisation often remains technology-neutral, treating the technical subsystem as a broad and undifferentiated category. As a result, prior studies tend to overlook how specific technologies contribute differently within transformation processes, limiting the ability to explain how ERP and AI shape sustainability outcomes in distinct yet interconnected ways. This study extends the STS perspective by introducing a technology-specific and dependency-based interpretation of transformation. ERP is conceptualised as an integrative infrastructure that enables data consistency and process stability, while AI functions as an adaptive layer that builds upon ERP-enabled data environments to support predictive and optimisation-oriented decision-making. This layered configuration suggests that digital transformation in SMEs is not an additive process of technology adoption, but a structured and conditional pathway in which technological effectiveness depends on sequential capability development. By reframing ERP and AI as interdependent rather than parallel technologies, this study provides a more precise explanation of how sustainability outcomes emerge under varying organisational and infrastructural conditions.

The authors would like to express their sincere gratitude to Universitas Islam Indonesia (UII) and Institut Teknologi Garut (ITG) for their institutional support and academic environment that facilitated the completion of this research. The support provided by both institutions was essential in enabling the development, refinement, and validation of the conceptual framework presented in this study.