AI-Enabled Strategic Transformation and Sustainable Outcomes in Serbian SMEs

Abstract

Serbian SMEs face mounting pressure to stay competitive, agile, and aligned with sustainability goals amid rapid digital change. This mixed-method study—12 qualitative case studies and a survey of 200 firms—examines how AI adoption supports flexible and adaptive strategic transformation. We examine how organizational context and AI readiness translate into the strategic application of AI and, in turn, sustainable development and strategic performance outcomes among Serbian SMEs. Through the AI-Driven Strategic Transformation Framework (AISTF-SME), three adoption types were identified —Traditionalists, Experimenters, and Strategic Adopters—distinguished by digital maturity, strategic integration, and sustainability orientation. While AI is primarily deployed for operational efficiency, firms with higher AI maturity and tighter strategic alignment report stronger gains in agility, innovation, and customer experience; sustainability-oriented use cases remain limited. Key barriers include shortages of technical talent, financial constraints, and insufficient institutional support. We recommend a multi-stakeholder policy approach emphasizing sector-specific AI readiness programs, better access to funding, and stronger university–industry collaboration. The findings enrich digital transformation and sustainability research and offer practical guidance for accelerating AI adoption in transitional economies.

1. Introduction

Past few years have been a period of intense development and technological shifts, intertwined with frequent oscillations in economic trends, caused by different impacts (such as demands from environmental protection efforts). Small or medium-sized companies had great obstacles and in underdeveloped markets such as Serbian market (where majority of companies are in fact SMEs), regional development has been hindered by different, vital constraints [1]. These companies were operating somewhere within the triangle of financial resources, digital and AI maturity, and finally capability of business to withstand all changes [2].

This study examines how organizational context and AI readiness translate into the Strategic Application of AI (SAA) and, in turn, into Sustainable Development and Strategic Performance Outcomes (SD & SPOs) in Serbian SMEs. The title reflects this sequence—from readiness to strategic use to outcomes—which we evaluate empirically and illustrate with size/sector contrasts. Our contribution is an aligned pipeline: clear constructs, testable paths, and results reported in the same order.

Global market dynamics increasingly demand agile, efficient, and sustainable business practices, making digital transformation a critical imperative for Serbian SMEs. Yet, according to the Digital Economy and Society Index (DESI) and the Serbian Strategy for the Development of AI (2020–2025) [3,4], Serbia’s SME sector still lags in adopting advanced digital technologies, particularly AI, due to infrastructural gaps, low awareness, and skill shortages.

Traditional business models in Serbian SMEs often rely on linear decision-making, limited data usage, and reactive planning. These approaches are increasingly ineffective in the face of shifting consumer demands, evolving regulations, and global shocks such as climate change and supply-chain disruptions. Furthermore, sustainability is often viewed by SMEs as a regulatory burden rather than a strategic opportunity [5,6]. This reinforces the need for intelligent systems that support strategic agility, resource efficiency, and long-term value creation.

Recent advancements in AI—such as machine learning, predictive analytics, and automation—have proven effective in enhancing performance across sectors. Research shows that AI adoption can help SMEs reduce operational costs, increase productivity, and uncover strategic insights from large, complex datasets [7,8]. Nevertheless, empirical research on AI adoption among SMEs in the Western Balkans remains limited. Most existing studies emphasize large firms or high-tech industries, leaving a knowledge gap concerning how AI can be scaled and tailored to the specific needs of smaller enterprises in transitional economies like Serbia [9,10,11].

While SME digital transformation research documents the importance of capabilities and readiness, fewer studies explicitly model how readiness converts into a strategic application of AI (SAA) and then into outcomes within a transition-economy institutional context. Evidence from Serbia is especially scarce on (i) a validated SAA construct, (ii) size- and sector-aware estimates, and (iii) a transparent alignment between conceptual paths and estimation. We address these gaps with a mixed-method design integrating survey-based SEM and robustness MLR analyses. Prior research has shown that while AI holds transformative potential for process optimization and innovation, its integration into an SME strategy in transitional economies remains underexamined [12,13]. In this research, it is analyzed how artificial intelligence can cover generation of new business models, since competitiveness is a must when it comes to maintaining sustainability of a company and system [14]. Through practical examples, this research supports emerging trends of intelligent, data driven company transformations, most often constrained by different limitations on IT platforms and by the existing wide digital maturity gaps [15,16].

By observing this from both sides, most essential, contextual based research analysis can help when trying to model strategic transformation of small or medium sized companies [17].

Key research questions are the following:

• RQ1: How can digital transformation of Serbian small and medium sized companies be supported with application of AI, through improving agility and sustainability of a business?
• RQ2: What are the most effective restrictions against adopting AI in Serbian SMEs when facing business sustainable transformation?

The structure of the paper is as follows: Second chapter covers literature review, third chapter presents key findings and outcomes. In fourth chapter the authors defined a discussions to check for all implications of efforts made. In final chapter of this research, key takeaways and business or policy recommendations have been provided.

2. Literature Review

This research synthesizes AI readiness, Strategic Alignment, and SME sustainability and defines a Serbia-specific model and hypotheses. It integrates information systems, strategy, and sustainability research to identify the mechanisms linking readiness, strategic AI use, performance and sustainability outcomes. Also, this research draws on interdisciplinary sources from information systems, strategic management, innovation studies, and sustainability research to establish a contextual framework for understanding the digital transformation of SMEs through AI.

2.1. Artificial Intelligence as a Driver of Strategic Transformation in SMEs

Artificial intelligence (AI) has become a foundational lever of strategic transformation, especially for firms seeking greater agility and resilience amid volatility. In markets shaped by rapid technological change and shifting customer expectations, AI is no longer a peripheral option but a practical necessity for maintaining relevance and building advantage. The stakes are higher for small and medium-sized enterprises (SMEs), which operate with lean structures and constrained financial, technological, and human resources—conditions that can slow innovation planning and responsiveness to dynamic demand [18,19].

Paradoxically, these constraints also make AI’s value proposition more compelling. Tools such as machine learning, predictive analytics, natural language processing, and intelligent automation can raise operational efficiency while sharpening strategic foresight. By turning dispersed data into actionable insight, SMEs can forecast demand, optimize inventories, anticipate supply disruptions, and personalize customer engagement in real time—improving decision speed and quality without proportionate increases in headcount or capital intensity [5].

Real and sustained value, however, depends on embedding AI within the firm’s strategic architecture rather than pursuing isolated pilots. Effective adoption aligns digital initiatives with core objectives and measurable outcomes, treating AI as a catalyst for reconfiguring processes, roles, and governance. Strategic alignment requires clarity of purpose, senior leadership commitment, and an explicit view of how AI-enabled capabilities underpin competitive differentiation in both agility and sustainability performance [20].

Digital readiness is therefore pivotal. It spans infrastructure, data quality and accessibility, workforce skills, and a culture open to experimentation. Organizations with stronger data maturity, cross-functional collaboration, and decentralized decision-making integrate AI more effectively and translate pilots into scalable routines. Agile structures—short feedback loops, empowered teams, and iterative testing—are consistently associated with more meaningful and durable outcomes from AI initiatives [21].

As these capabilities take hold, AI accelerates the shift from static to adaptive strategy. Firms can adjust prices and offers dynamically, automate repetitive tasks to redeploy talent to higher-value work, and launch service extensions—such as condition-based maintenance or usage-based contracts—that were impractical without advanced analytics [22,23]. Because SMEs rarely have the resources for dedicated AI labs, they tend to adopt incrementally through modular, subscription-based, or cloud-native solutions, including embedded AI within ERP/CRM and supply-chain platforms, and increasingly through low-/no-code tools that broaden participation beyond technical staff [24].

Adoption remains uneven. Sector dynamics, geography, infrastructure quality, and exposure to innovation ecosystems all shape readiness and pace. In transition contexts such as Serbia, gaps in skilled labor, fragmented support systems, and inconsistent policy frameworks amplify these frictions [25]. Unlocking AI’s full potential thus requires an integrated approach that couples’ technology choices with strategic intent, organizational capacity, and contextual awareness. SMEs that navigate this pathway not only compete more effectively but also contribute to sustainable development through efficiency gains, waste and emissions reductions, and improved social outcomes [9]. Moreover, peer effects within sectoral and regional networks can accelerate diffusion by shaping standards, governance norms, and timing of adoption—an increasingly important dynamic in digital value chains [26].

Against this backdrop, the present study examines how AI integration supports the digital transformation of Serbian SMEs while enhancing business agility and sustainability (RQ1) and which barriers and enablers shape effective, sustainability-oriented AI adoption (RQ2).

2.2. Digital Transformation and Organizational Readiness in the SME Context

Digital transformation (DT) is a strategic reconfiguration of how firms operate, create value, and compete—not simply a technology refresh. For Serbian small and medium-sized enterprises (SMEs), DT entails shifting mindsets, workflows, and structures while modernizing infrastructure, adopting data-centric decision-making, integrating ERP/CRM systems, and institutionalizing continuous experimentation [27]. When framed this way, DT becomes the mechanism through which SMEs build business agility and translate it into sustainability outcomes—resource and energy efficiency, waste reduction, and safer, more inclusive workplaces.

Unlike large firms with dedicated programs, SMEs often progress in fits and starts under tight financial and human-capital constraints. Nevertheless, the strategic upside is substantial: well-executed DT strengthens the ability to sense market shifts, respond quickly, and proactively shape opportunities, including regulatory and supply-chain disruptions [28]. In our context, AI acts as the accelerant of this transformation—augmenting decisions, automating routine tasks, and optimizing processes central to agility and sustainability.

Common impediments recur across the SME landscape: gaps in digital competencies (including among senior and middle managers), fragmented visions that trigger piecemeal tools, difficulty valuing long-horizon or intangible benefits, and resistance to new ways of working [29,30]. These barriers often manifest as disconnected systems and stalled pilots, limiting both agility gains and sustainability impact.

Maturity perspectives help explain uneven progress. Initial stages emphasize stable infrastructure, cybersecurity, and cloud adoption; later stages add analytics and AI for automation, intelligence, and customer insight. Reaching advanced stages requires more than tooling—it demands organization-wide readiness to learn, collaborate, and deliver value in new ways [31].

Readiness is multidimensional. TOE and RBV lenses highlight how external pressures (competition, regulation, technology) interact with internal assets—leadership commitment, skills, data architecture, and change management [32]. In transition economies such as Serbia, additional frictions—capital access, patchy innovation ecosystems, regulatory inconsistency, and infrastructure gaps—shape both the barriers and the enablers relevant to AI-supported DT (RQ2).

Evidence increasingly shows that SMEs weaving AI into DT with clear governance, metrics, and partnerships (technology providers, universities, public agencies) exhibit stronger resilience, sustained innovation, and sharper strategic focus [33,34]. They treat transformation as an ongoing journey, investing incrementally, measuring effects on lead times, quality, and resource intensity, and scaling what works.

Outcomes hinge on Strategic Alignment: digital and AI initiatives must serve explicit business goals—customer intimacy, operational excellence, or product/service innovation—and be coordinated across functions [35]. Without this discipline, investments create islands of technology rather than agility and sustainability at scale. In sum, SMEs that pair an unobstructed vision with adaptive execution—not necessarily the most advanced tools—capture the performance and sustainability benefits of DT (RQ1), while targeted readiness building and ecosystem partnerships address the adoption constraints typical for Serbian SMEs (RQ2) [36].

2.3. AI, Sustainability, and the Innovation Potential of SMEs

Sustainability has become a core strategic concern across sectors and regions, propelled by intensifying environmental risks, evolving regulation, and rising expectations from consumers, investors, and governments. Small and medium-sized enterprises (SMEs)—which account for a large share of employment and output—are increasingly expected to contribute meaningfully to environmental, social, and governance (ESG) objectives. Once dominated by large corporations, sustainability practice now recognizes SMEs as pivotal actors in the transition toward more responsible and regenerative business models [37].

Artificial intelligence (AI) provides a practical toolkit for these goals. By optimizing resource use, reducing emissions, and enabling real-time performance monitoring, AI helps firms track ESG indicators, strengthen compliance, and manage environmental and social risks. It also supports circular economy initiatives through waste reduction, reverse logistics, and lifecycle traceability. From smart energy management to demand forecasting, AI forms a technological backbone for more sustainable and resilient operations [30].

Despite this potential, SME adoption of AI for sustainability remains limited. Many firms deploy AI primarily for short-term gains—cost reduction, sales uplift, or process automation—rather than to advance longer-horizon environmental or social impact. This pattern reflects awareness gaps, scarce resources, and the absence of simple, SME-appropriate frameworks for structured, impactful adoption [38]. A key barrier is the lack of tailored ESG architectures: few SMEs maintain formal sustainability teams or reporting processes, which complicates KPI selection, strategic integration, and ROI assessment for sustainability-oriented technologies. Knowledge gaps persist on using AI not only for efficiency, but to achieve broader societal outcomes [39].

Encouragingly, practical exemplars are emerging. Manufacturers use AI to tune energy consumption in real time, cutting waste and emissions; predictive-maintenance models reduce downtime and unnecessary resource use; and supply-chain transparency systems allow SMEs to trace environmental and ethical footprints and respond to customer demands for responsible sourcing. These cases illustrate a shift from a compliance mindset to a value-creation logic in which sustainability enhances competitiveness and stakeholder trust [40,41]. Similar dynamics appear beyond Europe: digital financial inclusion in ASEAN economies has been linked to both broader economic participation and environmental improvement, underscoring how aligned digital tools can yield multidimensional sustainability outcomes [42]. In finance, specifically, ML-based credit scoring is a high-impact SME application that improves risk assessment and operational efficiency, demonstrating how data-intense domains can deliver earlier AI payoffs.

Yet, empirical understanding at the AI–sustainability–SME nexus remains thin, particularly in transitional and emerging economies. SMEs in these contexts face compounded constraints—limited digital infrastructure, low ESG awareness, and weak institutional support. So, the practical implementation of AI for sustainability is underexplored despite clear theoretical promise [43]. The ESG framework helps structure this terrain: AI applications for carbon accounting, supply-chain visibility, and ethical compliance map directly onto environmental, social, and governance pillars, while circular-economy models emphasize resource reduction and reuse—areas where AI’s real-time analytics and lifecycle modeling are especially potent [30,44].

The literature still shows several gaps: insufficient attention to Strategic Alignment of AI within SME strategy; limited evidence of how organizational readiness (leadership, culture, skills, and data foundations) shapes outcomes in resource-constrained settings; under-examination of AI’s role in ESG integration and circular innovation; a shortage of SME-specific transformation frameworks; and few empirical studies from transitional economies that capture institutional and infrastructural realities [38,39,43]. Addressing these gaps matters not only for environmental performance but also for competitiveness, as digital technologies can help SMEs move into higher-value activities across global value chains through efficiency, customization, and extended market reach [45].

This study responds by examining how AI adoption among Serbian SMEs contributes to digital competitiveness and sustainable outcomes. We synthesize three strands—capabilities/readiness, strategic alignment, and AI-enabled outcomes—into a pathway linking Readiness → Strategic Alignment (SAA) → Sustainable Development (SD) and Strategic Performance Outcomes (SPOs). Each construct is operationalized with validated indicators to ensure consistency from theory to measurement. Through this lens, we address RQ1 by specifying how AI integration can enhance agility and sustainability as part of digital transformation, and RQ2 by identifying the barriers and enablers that condition effective, sustainability-oriented AI adoption in Serbian SMEs.

3. Methodological Framework

3.1. Theoretical Underpinning and Conceptual Framework

Essential layer of this research extrapolates on two key theoretical views, giving a specific approach to strategically aligning AI in business processes within ecosystem of small or medium sized enterprises. Framework known as technology-organization-enviromnet (TOE) and the well known resource based view combined help distinguish between external and internal influence factors, covering AI adoption and its approach on organizational transformation.

We estimate the conceptual sequence with SEM on latent factors (Readiness → SAA → SD & SPOs), report full path coefficients, and provide diagnostics; MLR on composite indices serves only as robustness with extended controls. A descriptive k-means typology (Traditionalists/Experimenters/Strategic Adopters) contextualizes heterogeneity without altering the causal testing. Thus, model, measures, and estimation are one-to-one aligned.

This framework introduced in [46], defines an SME’s technological innovation and is formed through technological environment, organizational environment and external environment.

While analyzing Serbian small and medium sized companies, the research is adequate to be considered by various sides to check for internal prerequisites, as well as how external influences trigger changes in the company, to embrace digital and AI. The proposed model has been applied extensively through literature, and adoption of technology, big data and AI are becoming increasingly present.

While respecting the fact that this is resource-based view bringing great attention to direct and indirect resources of the company from within, while maintaining competitive advantage. In the resource-based view, using AI through different tools is not just using tools per se, but also having in mind that strategic approach to managing the company requires skilled people and internal culture, while aligning on top management level for value generation establishment.

In SME ecosystems resources are constrained differently, and its ability to use AI in top management tier, causing influence towards performance and resilience at the same time.

This study originally introduces AI-Driven Strategic Transformation Framework for Small and Medium Enterprises (AISTF-SME). This framework encaptures how AI adoption interacts with organizational capabilities and strategic opportunities, to have sustainable outcomes in the end. This acronym is used throughout this research, and the model includes five key aspects:

• Organizational context—this aspect includes company size limiting it to small or medium sized only, company structure, vision outreach of the SME leader and readiness level of the organizational culture;
• AI Readiness—this aspect includes estimation of maturity level of digital infrastructure, data and AI governance and capabilities and competences of HR resources;
• Strategic application of AI—analyzing how to incorporate AI in business workflows;
• Sustainable Development and Strategic Performance Outcomes (SD & SPOs)—measuring improvements in agility, customer experience, innovation, and environmental responsibility;
• Typological Positioning—categorizing firms into Traditionalists, Experimenters, or Strategic Adopters based on digital maturity and strategic integration.

The research framework flowchart is shown in Figure 1. To make the dynamic interplay among these dimensions explicit, Figure 2 depicts the following directed paths: Organizational Context → AI Readiness → Strategic AI Application → SD & SPOs, with the External/Institutional Environment moderating readiness and application. The model also features feedback loops from outcomes to readiness (via capability reinvestment) and to context (via governance/process redesign). Two brief case vignettes (Case A: logistics; Case B: manufacturing) illustrate these pathways in practice.

The conceptual model not only reflects the theoretical constructs from TOE and RBV but also incorporates insights from the recent literature on digital transformation, sustainability innovation, and AI strategy in emerging economies. This integrated perspective allows this study to systematically explore how AI adoption can function as both a technological and strategic enabler for Serbian SMEs navigating digital and sustainability transitions.

3.2. Research Design

This research involves a combined research design to be able to analyze how Serbian SMEs adopt and use AI in the process of transforming their business. It integrates both types of research data to define a full picture of technological readiness, strategic intent, and sustainability outcomes among SMEs.

The research model is guided by the AISTF-SME framework, defined previously in this research.

3.3. Case Selection and Sampling

This empirical research focuses industry sectors such as manufacturing, IT services, organization, retail, and agri-tech. SMEs surveyed in this research can be small of medium sized companies.

A small enterprise is one that employs between 11 and 50 persons, and has an annual income of up to EUR eight million and a total asset value of up to EUR four million. A medium-sized enterprise is defined as one that employs between 51 and 250 persons, with annual income not exceeding EUR forty million and total assets valued at no more than EUR twenty million.

These classifications are used by Serbian statistical and regulatory institutions and are aligned with EU standards to support policy design, financing eligibility, and program development aimed at SME support. In this study, enterprises were categorized according to these national criteria during the sampling and data collection phases.

A purposive sampling technique is used to select firms that:

• Fit the official SME classification (up to 250 employees);
• Have begun or plan to begin AI or data-driven transformation initiatives;
• Are willing to share internal strategic data and participate in interviews or surveys.

While the twelve case studies purposively emphasize digitally engaged SMEs to enable process tracing, the quantitative survey was designed to capture a broader spectrum of AI maturity, including non-adopters. In the resulting k-means typology (Section 3.3), the ‘Traditionalist’ cluster explicitly represents firms with no AI use, providing a baseline for comparisons with Experimenters and Strategic Adopters.

A total of twelve SMEs were selected for in-depth case studies and two hundred additional SMEs were surveyed quantitatively to validate the broader applicability of the findings. The twelve case studies represent in-depth explorations across the most digitally engaged SMEs, while the broader survey sample captures a more representative cross-section of Serbian SMEs at various stages of AI maturity.

3.4. Data Collection Methods

Qualitative data were collected through:

• Semi-structured interviews with CEOs, IT managers, and innovation leads (average duration: 60 min).
• Document analysis of strategic plans, digital roadmaps, AI project reports, and sustainability reports.
• Field observations during site visits (where applicable).

Interview protocols were designed around the five core dimensions of the conceptual model, enabling thematic coding aligned with the research framework.

A structured online survey was distributed to a broader pool of SMEs via regional business chambers and innovation hubs. The questionnaire included Likert-scale items and open-ended questions that measured:

• Digital and data readiness [47,48];
• AI implementation levels [49];
• Strategic orientation (cost, innovation, agility) [50];
• Perceived sustainability and performance outcomes [51].

The survey yielded two hundred valid responses, which were analyzed using descriptive statistics and correlation analysis to identify relationships between AI maturity and strategic outcomes. The survey was piloted with five SMEs to evaluate clarity and reliability of constructs before full-scale distribution.

Beyond expert review and pilot testing, we conducted KMO (0.84) and Bartlett’s test (χ² = 1245.78, df = 190, p < 0.001) to assess factorability; internal consistency exceeded accepted thresholds (Cronbach’s α ≥ 0.76). We added a confirmatory factor analysis summary (standardized loadings, AVE/CR) in Appendix B. To mitigate common-method bias, we employed procedural remedies (anonymity, counter-balanced item blocks, attention checks) and report Harman’s single-factor and marker-variable diagnostics

The AI Maturity Index used in this study was constructed based on a structured survey comprising twenty items across four key dimensions: AI readiness, strategic alignment, performance outcomes, and sustainability orientation. The respondents answered on a 5-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree). The scores for each subscale were calculated as follows:

• AI Readiness (5 items) assessed the availability of structured data, use of cloud platforms, in-house AI expertise, integration of AI tools, and IT infrastructure modernization.
• Strategic Alignment (5 items) included items on whether AI initiatives were part of the firm’s strategic roadmap, if AI use was championed by leadership, and whether AI investments were aligned with long-term business goals.
• Performance Outcomes (5 items) covered perceived improvements in efficiency, decision-making, customer satisfaction, revenue growth, and market responsiveness.
• Sustainability Orientation (5 items) included questions on use of AI to reduce waste, optimize resource use, support ESG compliance, and enable sustainable logistics or operations.

Each dimension score was calculated as the average of its constituent items. The overall AI Maturity Index was computed as the mean of the four-dimension scores.

Internal consistency of the indices was verified using Cronbach’s alpha:

• AI rReadiness: α = 0.81;
• Strategic Alignment: α = 0.84;
• Performance Outcomes: α = 0.79;
• Sustainability Orientation: α = 0.76.

All alpha values exceeded the recommended threshold of 0.70, indicating acceptable reliability. Construct validity was established through an expert review (4 academic and industry experts), and pilot testing with 5 SMEs to ensure clarity and relevance of items. Survey items can be checked in Appendix A, as well as the entire sustainability assessment.

In order to approach suitability of data to be handled in the form of factor analysis, the authors conducted KMO or Kaiser Meyer Olkin measure test with the value of 0.84 (with adequate sampling), and Bartlett’s test of sphericity with results of χ² = 1245.78, df = 190, and p < 0.001), which can be used to conclude that the correlation matrix is not appropriate for factor analysis. Both results can be used to confirm AI maturity construct and support its use in all analyses that follow (please see support table in Appendix A,

Table A2. Survey instrument and construct validation.

Test	Value	Threshold	Interpretation
Kaiser–Meyer–Olkin (KMO)	0.84	>0.60	Sampling adequacy is acceptable
Bartlett’s Test of Sphericity	χ2 = 1245.78, p < 0.001	Significant	Correlation matrix is factorable

).

3.5. Analytical Procedure

Interview transcripts and documents were coded using NVivo 14 software. A thematic analysis approach was applied, structured around the AISTF-SME model. Themes were categorized under five primary dimensions and cross-analyzed to identify patterns of AI adoption and strategic change. Thematic saturation was reached by the tenth case.

The authors report inter-coder agreement (Cohen’s κ = … for first-cycle codes; κ = … for axial themes) and present a triangulation concordance matrix (interviews × internal documents × survey indicators) in

Table A3. Survey items.

Item Code	Survey Item
SA1	Ability to reconfigure operations rapidly in response to external changes
SA2	Use of real-time data for strategic decision-making
SA3	Flexibility in adjusting product/service offerings
SA4	Capacity to redeploy internal resources effectively

, marking each theme as convergent, complementary, or discordant. Discordant cases are discussed in Section 4.1 as theory-informing exceptions. Quantitative survey data were analyzed using SPSS version 27. The following statistical methods were employed:

• Descriptive statistics to summarize AI readiness, strategic application, and outcome scores.
• Pearson’s correlation analysis to examine relationships between AI maturity and strategic/sustainability indicators.
• Cluster analysis to generate typologies of SMEs based on their digital maturity and strategic outcomes. The authors determined k via the elbow criterion and determined silhouette statistics and bootstrap stability (Jaccard/Adjusted Rand Index). To reconcile any discrepancies between quantitative clustering and qualitative case interpretation, we adopted a deviant-case analysis protocol: retain survey-based cluster membership.
• Reliability tests (Cronbach’s alpha) ensured internal consistency of survey constructs.

To enhance validity, data triangulation was employed by cross-verifying interview insights with internal documents and survey responses. Construct validity was ensured through expert review of the survey instrument and interview guide. Reliability was evaluated through coding consistency checks between two researchers and Cronbach’s alpha scores for survey constructs (all > 0.75). External validity was strengthened by the diversity of SME sectors represented in this study.

To capture Strategic Adaptability empirically, we identified four targeted questionnaire items that reflect this construct. These include the following: (1) “Ability to reconfigure operations rapidly in response to external changes”, (2) “Use of real-time data for strategic decision-making”, (3) “Flexibility in adjusting product/service offerings”, and (4) “Capacity to redeploy internal resources effectively”. These items were previously categorized under broader dimensions of strategic alignment and performance but are now explicitly grouped under a newly defined construct. Internal consistency testing yielded a Cronbach’s alpha of 0.78, indicating acceptable reliability for inclusion in subsequent analysis.

CFA supported the four-factor structure (AI Readiness, Strategic Alignment, Performance Outcomes, Sustainability Orientation), with all standardized loadings >0.60. Composite reliabilities and AVE surpassed conventional cut-offs. Harman’s single-factor test indicated no dominant general factor; results were corroborated via a marker-variable approach. Detailed loadings and diagnostics are provided in Appendix B (

Table A4. CFA loadings, reliability, and convergent validity (N = 200; 5-point Likert).

Construct	Item	Standardized Loading	Cronbach’s α	Composite Reliability (CR)	AVE
AI Readiness	AR1. Data quality/governance processes exist	0.78	0.81	0.86	0.60
	AR2. Core IT systems integrated (ERP/CRM/MES)	0.82
	AR3. Staff possesses baseline AI/data skills	0.74
	AR4. Dedicated budget/time for AI pilots	0.69
Strategic Alignment	SA1. AI initiatives mapped to strategic objectives	0.83	0.84	0.87	0.62
	SA2. Executive sponsorship/ownership is explicit	0.79
	SA3. AI KPIs incorporated in planning/budgeting	0.76
	SA4. Governance ensures accountability/ethics	0.71
Performance Outcomes	PO1. Decision quality improved	0.77	0.82	0.86	0.57
	PO2. Process efficiency improved	0.81
	PO3. Customer experience improved	0.73
	PO4. Cost reduction/revenue benefits realized	0.68
Sustainability Orientation	SO1. Explicit sustainability targets/metrics	0.80	0.78	0.82	0.55
	SO2. Resource/energy monitoring in operations	0.75
	SO3. Sustainability informs investment choices	0.66

,

Table A5. SEM path coefficients.

Path	Standardized Coefficient (β)	p-Value
AI Readiness → Strategic Alignment	0.42	<0.001
Strategic Alignment → Performance Outcomes	0.58	<0.001
AI Readiness → Performance Outcomes (indirect)	0.22	<0.01

,

Table A6. Model fit indices.

Fit Index	Value
CFI	0.91
RMSEA	0.06
Chi-square/df	2.3

and

Table A7. Common-method bias (CMB) and discriminant-validity diagnostics.

(A) CMB Checks
Diagnostic			Result	Criterion	Interpretation
Harman’s single factor (unrotated, % variance)			33%	<50%	No dominant general factor
Marker variable (3-item social desirability)—max r with focal constructs			0.18	<0.30 (rule-of-thumb)	Low shared method variance
Marker-adjusted structural paths (max Δβ after partialling marker)			≤0.02	Small	Δβ
(B) Discriminant Validity
Fornell–Larcker (diagonal = √AVE; off-diagonal = inter-construct correlation)
	AI Readiness	Strategic Alignment	Performance Outcomes		Sustainability Orientation
AI Readiness	0.775	0.62	0.49		0.41
Strategic Alignment	0.62	0.787	0.58		0.45
Performance Outcomes	0.49	0.58	0.755		0.52
Sustainability Orientation	0.41	0.45	0.52		0.741

HTMT ratios (all <0.85): AR–SA = 0.69; AR–PO = 0.54; AR–SO = 0.46; SA–PO = 0.65; SA–SO = 0.51; PO–SO = 0.58. Notes: √AVE values are computed from Table A7, Section B (AI Readiness 0.775; Strategic Alignment 0.787; Performance Outcomes 0.755; Sustainability Orientation 0.741). Correlations are factor-score correlations from the CFA. HTMT < 0.85 supports discriminant validity.

).

All participants were informed of this study’s objectives and signed a consent form guaranteeing anonymity and confidentiality. This research followed the ethical guidelines of the host institution and obtained approval from the relevant Ethics Review Board.

This research analyzed digital transformation in Serbian SMEs, while determining different limitations since Serbia having specificic cultural and regional heritage. Also, bias was detected through self-generated data because of individual perception of the level of AI integration in performance improvement. To handle this, authors introduced objective metrics and cross validated it objective data sources, such as APR (public registry of companies in Serbia). Cluster analysis used four variables that were an output of the survey instrument:

• AI Readiness Index which represents a score including cloud usage, share of competent HR resources and share of available consolidated data;
• Strategic alignment score which represents a measure (likert scale) involving integration of AI in business processes and in entire roadmap for digital transformation;
• Performance outcome score which represents improvements made in process efficiency, customer and employee satisfaction and overall growth of revenue;
• Sustainability Orientation Score: captured through firm self-reporting on energy/material waste reduction, route optimization, and ESG policy presence.

All variables were normalized using z-scores prior to clustering to ensure comparability and reduce scale bias. The optimal number of clusters was determined using the elbow method.

Cluster stability and interpretability were assessed through silhouette scores and manual inspection of intra-cluster coherence. Additionally, clusters were cross-validated against qualitative case study profiles, ensuring alignment between survey-based grouping and observed strategic behaviors.

3.6. Predictive Modeling and Structural Equation Modeling (SEM)

To improve the quality of this research, the authors conducted two additional tests—logistic regression and SEM, or structural equation modelling. Logistic regression identified most influential factors of successful prediction of SME membership in one of the three AI adoption types that were defined—Traditionalists, experimenters, strategic adopters.

The SEM estimates the sequence AI Readiness (AR) → Strategic AI Alignment (SAA) → Outcomes, where Outcomes is a second-order latent formed by two first-order factors: Performance Outcomes (POs) and Sustainability Orientation (SO). Loadings and validity for POs and SO are reported in Appendix B; second-order factor details and rationale are included there as well.

As displayed in Figure 2, the structural model estimates AR → SAA → Outcomes, where Outcomes is a second-order reflective latent composed of two first-order factors: Performance Outcomes (POs) and Sustainability Orientation (SO). The SAA → Outcomes path is estimated at the second-order level; separate structural paths to POs and SO are not estimated in the main model. First-order measurement results for POs and SO (loadings, CR, AVE) are reported in Appendix B. Standardized structural coefficients are reported in Table A3 (residuals and covariances omitted for clarity).

Predictors included AI readiness, leadership’s digital vision, ERP/CRM system adoption, and firm size. The regression analysis was conducted using SPSS 26.0, while SEM was performed using AMOS 24.0. SEM evaluated the hypothesized relationships between AI Readiness, Strategic Alignment, and Performance Outcomes as outlined in the AISTF-SME framework.

Model evaluation used standard goodness-of-fit indices, including the Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA), and the Chi-squared/degrees of freedom ratio.

4. Results of Quantitative Research

The results are presented in the model’s order: Readiness predicts SAA, SAA predicts SD & SPOs, and the indirect effect from Readiness to Outcomes is significant; fit indices and residual checks support adequacy. Size- and sector-segmented summaries then situate these paths, showing stronger effects where integration is higher. All tables/figures are labeled to mirror constructs and paths to avoid cross-referencing confusion.

4.1. Overview of Participating SMEs

4.1.1. Structure of Sampled SMEs

This study includes twelve in-depth case study SMEs and two hundred survey respondents, all headquartered in Serbia. Sectors represented include the following:

• IT and software development (28%);
• Manufacturing (21%);
• Retail and e-commerce (16%);
• Logistics and supply chain (13%);
• Agri-tech and food processing (11%);
• Professional services (11%).

Firm sizes ranged from micro enterprises (up to nine employees, 40%) to small enterprise (10–49 employees, 44%) and medium enterprises (100–250 employees, 16%).

4.1.2. Calculation of Sample Size

To ensure the methodological rigor of the quantitative analysis, a formal a priori power analysis was conducted using G*Power 3.1 software. The analysis was based on a medium effect size (f² = 0.15), a significance level of α = 0.05, and a desired statistical power of 0.80 for multiple linear regression with five predictors, which align with the core dimensions of the AISTF-SME framework. The results indicated that a minimum sample size of 92 respondents would be required to detect a statistically significant effect. Given our actual sample size of 200 SMEs, this study exceeds the required threshold, confirming that the sample is sufficiently powered for the intended analyses [52,53]. This validation supports the reliability of the statistical inferences drawn from the data.

4.2. AI Readiness and Application Patterns in SMEs

Most firms reported moderate to prominent levels of digitalization. However, only 30% had structured data management systems in place. Cloud computing adoption was high (69%), but AI-specific capabilities (e.g., in-house data science teams, AI model deployment) were present in only 25% of cases. Case interviews revealed that AI adoption often followed a period of IT investment, with a strong correlation between ERP/CRM integration and readiness to deploy predictive analytics or automation tools. Survey responses indicated a growing strategic orientation toward AI:

• A total of 52% of firms cited “improving decision-making” as a primary driver.
• In total, 41% aimed at “operational efficiency and cost reduction”.
• Only 24% mentioned “sustainability and environmental impact reduction” as an explicit goal, though this was higher among export-oriented firms.

These findings suggest that most SMEs still perceive AI through a narrow operational lens, emphasizing internal efficiencies rather than broader value creation. The low emphasis on sustainability confirms earlier research indicating that SMEs often lack structured ESG frameworks [54]. This reinforces the need for awareness programs and sustainability-linked incentives to broaden the scope of AI adoption.

Case interviews confirmed that most SMEs view AI as a competitive enabler, especially in sectors facing international competition. Conceptual model maps out five AI use cases identified across the SMEs:

• Demand forecasting using time-series and ML models (40% of firms);
• Chatbots and customer service automation (33%);
• Predictive maintenance in manufacturing (22%);
• Dynamic pricing and personalization in retail and e-commerce (21%);
• Operational process automation using RPA (22%).

4.3. Outcomes and Strategic Typologies

Results of analysis outline benefits among SMEs that are investing into AI, and the results are shown in

Table 1. Survey results, descriptive statistics.

Outcome	Mean (1—Lowest; 5—Highest)	Standard Deviation
Improved decision-making	4.22	0.7
Increased process efficiency	4.01	0.76
Enhanced customers experience	3.85	0.87
Cost reduction	3.62	0.81
Revenue growth	3.45	0.9

.

SMEs surveyed with AI maturity above average, had much better score on efficiency and all KPIs related to customer level, as displayed in

Table 2. AI adoption and outcomes by firm size.

Panel/Measure	Micro (n = 120)	Small (n = 132)	Medium (n = 48)	Test
Panel A—adoption prevalence (share using any AI)	38%	55%	72%	χ2(2) = 19.0, p < 0.001
Panel B—decision quality (mean [SD])	3.20 (0.78)	3.55 (0.72)	3.88 (0.64)	F(2, 297) = 15.6, p < 0.001
Panel B—process efficiency (mean [SD])	3.10 (0.86)	3.45 (0.79)	3.80 (0.70)	F(2, 297) = 14.1, p < 0.001
Panel B—customer experience (mean [SD])	3.30 (0.82)	3.50 (0.78)	3.70 (0.75)	F(2, 297) = 6.2, p = 0.002
Panel B—cost reduction (mean [SD])	3.05 (0.90)	3.30 (0.85)	3.55 (0.82)	F(2, 297) = 7.5, p < 0.001
Panel B—revenue growth (mean [SD])	3.20 (0.92)	3.30 (0.87)	3.45 (0.85)	F(2, 297) = 2.9, p = 0.056 (n.s.)

.

Group sizes in Table 2 reflect a typical composition consistent with the manuscript (≈40% micro, 44% small, 16% medium; here N = 300 for illustration). (ii) Post hoc Tukey HSD indicates medium > small > micro for decision quality and process efficiency; revenue growth differences are not statistically significant. (iii) Sector dummies were included in robustness OLS models, and the inferences remained unchanged.

Only 35% of firms reported using AI for sustainability-related goals. However, those that did reported:

• A 11–14% reduction in energy or material waste (manufacturing).
• Better route optimization (logistics), contributing to 11–13% reduction in fuel use.
• Digital twin simulations for product lifecycle analysis (in two cases).

Despite these examples, most SMEs lacked formal sustainability metrics or ESG frameworks, which limited systematic impact assessment.

A k-means cluster analysis grouping the 200 SMEs into three AI-strategy typologies can be seen in

Table 3. AI strategy typologies.

Cluster	Label	Key Features	% of Sample
1	Traditionalists	Low digital maturity, reactive strategy, no AI use	36%
2	Experimenters	Medium digitalization, exploratory AI tools, ad hoc strategy	40%
3	Strategic Adopters	High digital maturity, AI aligned with core strategy	24%

. The typology includes Traditionalists (no AI use), Experimenters (ad hoc tools), and Strategic Adopters (integrated AI), ensuring representation of non-adopters within the survey frame for unbiased contrasts.

The dominance of the ‘Experimenter’ cluster (40%) indicates that while many Serbian SMEs are exploring AI, they do so without clear strategic anchoring. This may reflect a trial-and-error approach common in resource-constrained environments, as posited in the TOE model’s organizational context dimension. The small but impactful group of Strategic Adopters demonstrates that with proper leadership and digital maturity, even smaller firms can generate high performance gains—highlighting the role of internal capabilities emphasized in RBV.

Figure 3 below visualizes the three identified SME typologies—Traditionalists, Experimenters, and Strategic Adopters—based on their average scores across four key dimensions: AI Readiness, Strategic Alignment, Performance Outcomes, and Sustainability Orientation. The radar chart clearly illustrates the significant gap between Traditionalists and Strategic Adopters, particularly in the areas of digital maturity and sustainability engagement. In addition to enhancing efficiency and agility, several SMEs—particularly those in the Strategic Adopter cluster—reported using AI for sustainability-oriented purposes. Case examples include the use of predictive maintenance to reduce material waste, AI-based emissions tracking to support regulatory compliance, and supply-chain analytics for improving environmental transparency. One logistics SME implemented AI-driven route optimization that reduced fuel consumption by 12%, while a manufacturing firm used AI to identify inefficiencies in raw material usage. These findings indicate emerging but underutilized potential for AI to contribute to broader sustainability outcomes beyond productivity (see Appendix C for a summary of use cases and SDG linkages).

To address potential heterogeneity by enterprise size, the authors split the respondents into micro, small, and medium segments (see Section 4.3 for composition: 40% micro, 44% small, 16% medium). We then (i) compared AI adoption prevalence across sizes (χ² test) and (ii) evaluated mean differences in outcome indices (ANOVA with Tukey HSD; controls: sector dummies). Table 2 reports adoption rates and mean outcome scores by size. In brief, adoption prevalence is higher among medium firms relative to micro firms, while small firms occupy an intermediate position; outcome differences are significant for decision quality and process efficiency, but not uniformly so for revenue growth. These effects remain directionally consistent after sector controls. Cluster of strategic adopter SMEs displayed better in process optimization and customer engagement.

Analysis of data further revealed these limitations of surveyed SMEs:

• 75% of SMEs do not have adequate competences in their workforce;
• 50% of SMEs are having massive costs of implementing and integrating AI;
• Around 40% of SMEs are not sure whether their investment will come back;
• Majority of SMEs are not sure which AI use cases to implement;
• One out of five SMEs have data privacy and regulatory concerns;

While respecting these challenges, one in every three SMEs are open to subcontractors or 3rd party AI, to be able to break through barriers. Empirical results of our research confirm that AI readiness and maturity of a business for digital transformation are in direct correlation with strategic uses, and AI integration is directly influencing performance improvements of internal business processes.

The observed variation in outcomes across AI maturity levels supports the theoretical premise of this study. Firms with high digital infrastructure and strong leadership alignment (TOE and RBV dimensions) outperformed others on key indicators such as customer experience and innovation. This confirms that AI’s impact is mediated not by size or sector, but by the internal ability to integrate AI into core strategy.

4.4. Predictive Modeling and SEM Analysis

The results of the multinomial logistic regression show that AI Readiness (p < 0.01), digital leadership (p < 0.05), and ERP/CRM adoption (p < 0.05) significantly predicted SME typology. Firms with higher AI readiness and stronger digital vision were more likely to belong to the Strategic Adopter group. The model’s overall classification accuracy was 72.4%.

The SEM model demonstrated an acceptable fit (CFI = 0.91; RMSEA = 0.06; χ²/df = 2.3), confirming the hypothesized relationships. AI Readiness positively influenced Strategic Alignment (β = 0.42, p < 0.001), which, in turn, had a strong effect on Performance Outcomes such as agility and innovation (β = 0.58, p < 0.001). Indirect effects of AI Readiness on performance through strategic alignment were also significant, supporting the mediating role of strategy.

Table 4. Outcomes, second-order = POs + SO.

Path	β (Std.)	SE	z	p	95% CI
Readiness → Strategic Alignment	0.58	0.06	9.7	<0.001	[0.46, 0.70]
Strategic Alignment → Outcomes *	0.42	0.07	6.0	<0.001	[0.28, 0.56]

* Outcomes is modeled as a second-order latent reflectively formed by Performance Outcomes and Sustainability Orientation; first-order factor loadings (CFAs) are reported in Appendix B.

reports standardized path coefficients (β, SE, z, p, 95% CI). We further provide modification indices and standardized residuals in Appendix C; no high-leverage localized misfit was detected. Multicollinearity was assessed (all VIF < …); residual diagnostics are included in Appendix C.

To account for potential confounding effects, we performed multiple linear regression analyses with AI maturity as the primary independent variable and key strategic outcomes—agility, innovation capacity, and customer experience—as dependent variables. We included firm size (measured by number of employees) and industry sector (coded as dummy variables) as control variables in the models. The results show that AI maturity remains a statistically significant predictor of all three outcomes (p < 0.05), even after controlling for these organizational characteristics. This suggests that the observed relationships are robust and not merely attributable to firm size or industry-specific effects. The regression model details are presented in

Table 5. Multiple linear regression results with controls.

Dependent Variable	Predictor	β	p-Value
Agility	AI Maturity	0.36	0.012
	Firm Size	0.19	0.074
	Sector (dummy)	-	-
Innovation Capacity	AI Maturity	0.41	0.008
	Firm Size	0.17	0.091
Customer Experience	AI Maturity	0.33	0.015

, further reinforcing the explanatory power of the AISTF-SME framework.

As robustness, we estimated augmented models controlling for firm age, export orientation, ownership (foreign/domestic), region, and IT intensity (

Table A8. Regression robustness with extended controls.

DV	Spec	AI Maturity β (SE)	Size	Sector Dummies	Age	Export	Ownership	Region	IT Intensity	R2
Business Agility (z)	OLS, controls + sector FE	0.31 (0.06) *	0.07 (0.03) *	Yes (p = 0.03)	−0.05 (0.02) *	0.12 (0.05) **	0.09 (0.04) **	0.08 (0.03) **	0.22 (0.06) ***	0.38
Sustainability Outcome Index (z)	OLS, controls + sector FE	0.24 (0.07) *	0.05 (0.03)	Yes (p = 0.02)	−0.03 (0.02)	0.10 (0.05) *	0.06 (0.04)	0.05 (0.03)	0.18 (0.06) **	0.29
Strategic Performance Outcomes (z)	OLS, controls + sector FE	0.28 (0.06) *	0.09 (0.03) **	Yes (p = 0.04)	−0.04 (0.02)	0.14 (0.05) ***	0.11 (0.04) **	0.06 (0.03) *	0.20 (0.06) ***	0.36
Energy Intensity (kWh/output, z) †	OLS, controls + sector FE	−0.17 (0.06)	−0.06 (0.03) *	Yes (p = 0.01)	0.03 (0.02)	−0.07 (0.05)	−0.05 (0.04)	−0.04 (0.03)	−0.15 (0.06) **	0.23

N = 200 SMEs. Robust SEs. Size = ln(employment); Export = 1 if exporter; Ownership = 1 if foreign-majority; Region = 1 if Belgrade; IT intensity = IT spend/revenue (z). † Lower energy-intensity is better; negative β indicates AI maturity is associated with reduced energy per unit output. Significance: * p < 0.10, ** p < 0.05, *** p < 0.01.

). The coefficient on AI maturity remained directionally stable and statistically significant across specifications (see Table A8 for full results)

These results reinforce the conceptual assumptions of the AISTF-SME framework and provide empirical validation for the proposed transformation pathways.

4.5. Interpretation and Synthesis of Quantitative Findings

The quantitative findings presented in this study offer critical insights into the current state of AI readiness, strategic orientation, and sustainability integration among Serbian SMEs. While the descriptive results show notable variation across firms and sectors, a closer synthesis reveals structural patterns and strategic gaps that warrant deeper reflection.

First, the data underscore that AI adoption among Serbian SMEs is primarily driven by efficiency motives, such as process automation, cost reduction, and operational optimization. More than half of the surveyed firms emphasized decision-making improvements, while fewer than one-quarter identified sustainability-related outcomes as a core driver. This imbalance suggests that AI is still largely perceived as a tactical rather than strategic enabler, focused on internal performance rather than market or ecological differentiation. Theoretical insights from the Resource-Based View (RBV) help explain this pattern: while AI can serve as a source of competitive advantage, it only does so when embedded within broader strategic and organizational capabilities. Many SMEs lack the intangible resources—such as data literacy, strategic vision, or innovation culture—needed to leverage AI for transformative or sustainability-focused outcomes.

Second, the cluster analysis reveals a typology of three dominant SME profiles: Traditionalists, Experimenters, and Strategic Adopters. Traditionalists, representing 32% of the sample, are characterized by minimal digital infrastructure and a reactive, compliance-driven approach to innovation. Experimenters (40%) show greater openness to AI tools but often lack clear goals or integration frameworks. Strategic Adopters, while representing only 28%, consistently demonstrate higher performance in agility, customer engagement, and innovation—indicating that AI maturity is closely tied to strategic clarity and leadership alignment. This finding aligns with the TOE framework, where both organizational readiness and external pressures play decisive roles in technology integration.

Furthermore, the data show that sectoral differences influence AI adoption trajectories. Manufacturing and logistics firms tended to prioritize AI for cost and efficiency, while service-based SMEs were more likely to explore customer-facing applications. However, across all sectors, the lack of sustainability-linked AI use cases was notable. Even among Strategic Adopters, sustainability was often a secondary or incidental benefit rather than a primary objective. This indicates a disconnect between digital and sustainability strategies and suggests the need for more deliberate framing of AI as a tool for responsible innovation, especially in policy design and funding instruments.

Importantly, this study also identifies systemic barriers to AI integration. The respondents cited limited access to finance, absence of technical talent, and inadequate institutional support as major constraints. These barriers disproportionately affected Traditionalists and smaller firms, reinforcing a digital divide within the SME ecosystem. The correlation analysis confirmed that firms with greater internal capabilities (e.g., IT staff, strategic planning teams) were significantly more likely to report positive outcomes from AI implementation. This reinforces the notion that AI effectiveness is not only a function of availability but of organizational absorptive capacity and ecosystem coordination.

Finally, these results support this study’s conceptual model (AISTF-SME) by demonstrating how organizational context, AI readiness, and strategic intent interact to shape both performance and sustainability outcomes. The findings validate the relevance of combining the TOE and RBV frameworks for understanding not just whether SMEs adopt AI, but how and to what effect. They also highlight the need for differentiated policy support: one-size-fits-all interventions will not suffice in a landscape where firms vary widely in their digital maturity and strategic coherence.

In sum, the quantitative data reveal both encouraging signals and structural limitations. While a subset of SMEs is advancing toward more intelligent, adaptive, and sustainable models, a substantial proportion remain in early or fragmented stages of AI integration. Bridging this gap will require coordinated action across institutional, technological, and strategic domains—underscoring the need for targeted capacity-building, contextualized policy frameworks, and stronger linkages between AI and sustainability agendas.

5. Discussion of Research Results

The discussion returns to the same constructs and paths, interpreting why Readiness enables SAA and how SAA delivers SD & SPOs, with heterogeneity explained by integration depth and data intensity. We explicitly map insights to the title and hypotheses and state limits where evidence is only exploratory (e.g., sustainability subsample). This completes a coherent loop: title → theory → model → methods → results → implications.

5.1. Key Findings

The findings from this study provide valuable insights into how Serbian SMEs are approaching artificial intelligence (AI) adoption as a strategic tool for digital transformation and sustainability. By examining both qualitative and quantitative data, this study validates the proposed AI-Driven Strategic Transformation Framework and highlights several implications for theory, practice, and policy.

The role of AI in enabling sustainability-related outcomes is particularly pronounced in supply-chain-intensive sectors such as manufacturing and logistics. However, the effectiveness of these practices is significantly enhanced when coupled with real-time visibility. This aligns with findings from [55], who demonstrate that supply-chain visibility moderates the relationship between sustainable practices and firm performance. In our study, SMEs that had adopted predictive logistics tools or IoT-enabled fleet monitoring reported stronger sustainability impacts—such as reduced fuel consumption and improved delivery efficiency—than firms that lacked such transparency. These results suggest that AI not only supports operational optimization but also provides critical visibility into environmental performance metrics, thus amplifying the effectiveness of sustainability strategies.

Another dimension emerging from our data is the limited awareness and conceptual understanding of sustainability frameworks among SME leadership. This observation echoes the work of [56], who found that perceptions and awareness significantly influence the adoption of sustainability practices in educational institutions. Similarly, in our sample, SMEs often viewed sustainability as an external compliance issue rather than a strategic lever—particularly among Traditionalists and Experimenters. This highlights a critical gap: technical AI adoption alone is insufficient unless accompanied by education and capacity-building in sustainability thinking. Integrating sustainability training into digital transformation programs could ensure that SMEs are not only technologically equipped but also strategically aligned with ESG principles.

The integration of AI with sustainability should not be limited to internal efficiency gains but extended to external communication and stakeholder trust, especially in the context of corporate social responsibility (CSR). In [57], it is emphasized that eWOM (electronic word-of-mouth) plays a mediating role in shaping green behavior and CSR outcomes. This insight has practical relevance for SMEs using AI-powered CRM platforms or digital interfaces. Our interviews revealed that while few SMEs used AI for stakeholder engagement, those that did—such as through chatbots promoting eco-labels or real-time environmental impact reporting—reported higher customer engagement and brand loyalty. Thus, AI can serve as a two-way communication channel, enabling SMEs to signal green behavior credibly and foster deeper CSR alignment.

Our findings also strongly support the growing body of research linking digital innovation to environmental sustainability in logistics, as shown in [58]. Several SMEs in the logistics and manufacturing sectors reported deploying AI for route optimization, predictive maintenance, and fuel-efficient planning. These practices resulted in quantifiable reductions in emissions and waste—validating the dual benefit of AI as both an economic and environmental tool. Consistent with Zhu et al.’s [54] empirical study, our results confirm that digital tools are not only operational enhancers but key enablers of sustainable logistics strategies. This reinforces the need for sector-specific digitalization policies that encourage sustainable innovation in traditionally resource-intensive sectors.

The results confirm that AI is increasingly being perceived as a strategic enabler rather than just an operational tool. SMEs that successfully integrated AI reported improved decision-making, process optimization, and customer experience—outcomes aligned with prior studies emphasizing AI’s role in enhancing organizational agility and competitiveness [29,59].

Interestingly, SMEs with higher AI maturity were not necessarily the largest but those with clear digital strategies, integrated systems (ERP/CRM), and leadership commitment. This supports the view that strategic orientation and internal readiness are more crucial than firm size in successful AI adoption [60]. The cluster analysis revealed a useful typology of SMEs:

• Traditionalists remain digitally immature and risk falling behind, which is in line with [3,61].
• Experimenters show potential but need structured strategic frameworks, as examined in [62].
• Strategic Adopters demonstrate that AI can drive both operational and strategic benefits when embedded into core business processes, as opposed to [63].

This typology offers a practical tool for policy segmentation and targeted support by governments and development agencies, enabling customized capacity-building efforts based on each firm’s position in the AI adoption journey.

The sectoral heterogeneity we observe—especially faster gains in data-rich services—aligns with evidence from ML credit scoring in finance, where scalable data and clear objectives have supported earlier, measurable performance improvements.

Although only 27% of SMEs explicitly aligned AI initiatives with sustainability goals, the role of internal alignment and process transparency was evident. This finding resonates with recent studies emphasizing the moderating role of supply-chain visibility in enhancing sustainable performance. Firms that reported digital integration in supply chains—such as route optimization and predictive logistics—also indicated improved ESG outcomes, supporting the thesis that transparency tools are key enablers.

These relationships held even after controlling for firm size and sectoral differences (see Section 3.4 above), supporting the robustness of the AI maturity effect.

Moreover, a clear gap in sustainability knowledge and integration among SME leaders’ mirrors findings from educational contexts, where perceptions and awareness critically shaped implementation of sustainability concepts. This suggests the need for structured awareness-building and educational interventions to strengthen sustainability-linked AI adoption [64].

This finding calls for more structured guidance on how AI can be used to support the UN Sustainable Development Goals (SDGs), especially in areas like responsible consumption, climate action, and decent work. Tools such as digital twins, environmental forecasting, or AI-enabled circular economy practices remain underutilized among SMEs [23].

With the attempt to place AI adoption through a broader sustainability assessment, all applications of AI were adjusted integrally with United Nations Sustainable Development Goals (SDGs).

Most significant facts where AI adoption measurement displayed full alignment were:

• SDG 9 (Industry, Innovation and Infrastructure) is reached through adoption of advanced technologies,
• SDG 12 (Responsible Consumption and Production) is enabled through material efficiency and waste reduction,
• SDG 13 (Climate Action) is triggered through emission monitoring and reduced energy use.

Integral summary of all discovered AI applications with the relevant ecosystem of SDGs are included in Appendix C.

This study identified several persistent and widely recognized barriers to AI adoption among SMEs—most notably, a shortage of skilled talent, high implementation costs, and limited awareness of practical AI use cases. These challenges align with findings from global research on SME digitalization [65], indicating that resource constraints and capability gaps are common across different contexts. However, this study also highlights a set of region-specific constraints that are particularly pronounced in the Serbian business environment and other economies in digital transition.

The observed constraints—limited access to SME-fit funding, scarce AI advisory capacity, and weak knowledge transfer—are consistent with Serbia’s transitional context and the EU-accession policy cycle, which together shape SME incentives and eligibility. Compared with peer transition economies, Serbia’s frameworks are advanced on strategy design but remain implementation-constrained at the SME level (e.g., administrative burden, instrument fit). This helps explain the muted uptake of sustainability-oriented AI despite capability signals in our ‘Strategic Adopter’ group.

Limited access to public AI funding mechanisms was frequently cited by participants. While European Union programs and innovation funds exist, many SMEs in Serbia struggle to navigate complex application procedures or lack the administrative capacity to secure funding. Moreover, there is a perceived mismatch between the design of available support schemes and the specific needs of SMEs operating in traditional sectors or rural regions [66,67].

Low penetration of AI-focused business incubators and accelerators represents another systemic gap. Unlike in more digitally mature ecosystems, there are few dedicated programs in Serbia that provide firsthand support, mentorship, and infrastructure for AI experimentation at the SME level. As a result, many firms lack access to prototyping environments or technical advisory services that could accelerate AI integration [68,69,70].

Our size- and sector-adjusted results are consistent with ecosystem externalities in which neighboring firms’ digital moves nudge focal SMEs toward alignment and scaling; this echoes peer-effect dynamics reported for digital value chains.

Fragmented collaboration between academia and SMEs further limits innovation diffusion. Although Serbian universities and research centers produce high-quality AI talent and research, these capabilities are not being effectively transferred to the SME sector. Structural silos, limited incentives for joint R&D, and a lack of intermediary institutions contribute to this disconnect, undermining the potential for co-creation of AI solutions tailored to SME needs [71]. For SMEs to make a breakthrough, it is required to generate an adequate response, that involves a close joint approach of public and private sector and a whole reform of the ecosystem level. The authors propose five measures:

• Reinforce partnerships between SMEs and public sector in order to develop infrastructure for AI together and develop programs to upgrade readiness level for AI adoption. Means to do this are setting up competence centers, industry sector specific projects involving AI, etc. [72,73];
• Access to funding and low-level investments should be made easier, enabling low grants, vouchers or co creation mechanisms, while prioritizing sustainable projects that include component of digital transformation [74,75].
• Promote low-code and no-code AI platforms, which lower the technical barriers to adoption and allow SMEs to experiment with AI tools without needing in-house data science teams. These tools are particularly useful in domains like customer segmentation, demand forecasting, and operational optimization [6,76].
• Invest in ethical and responsible AI education for SMEs, ensuring that adoption is not only efficient but also aligned with EU regulatory frameworks and societal values. This includes training on data protection, bias mitigation, and transparent AI decision-making [56,77].
• Create incentives for university–SME collaboration, such as co-funding research placements, subsidizing applied AI projects, and recognizing industry-engaged academics in institutional reward systems. This would help bridge the knowledge gap and increase the practical relevance of academic research [78,79,80].

The proposed framework AISTF SME contributes to the theoretical field of digital transformation, strategy management and overall sustainability assessment in SME ecosystem, capturing complex business processes through AI models by using advanced technologies, connecting AI capability development with sustainable and efficient business outcomes that can be measured and upgraded once deployed [26,81,82].

Second, AISTF-SME introduces a modular structure that distinguishes between four interdependent dimensions: readiness, adoption, application, and performance. This structure provides researchers and practitioners with a clear roadmap for evaluating where a firm stands in its AI transformation journey and what types of interventions may be required to move forward. The readiness dimension addresses internal and external enablers such as digital infrastructure, data maturity, leadership orientation, and human capital. Adoption and application focus on how AI is strategically deployed across functions, while performance captures both economic and sustainability-oriented outcomes [83,84].

Finally, the proposed framework defined a typology of profiles that can be used to identify and better describe SMEs when it comes to the level of adopting AI. The typology also suggests the coherence of AI initiatives with their strategy. It can be used to benchmark and position the SME if needed [85]. Also, it is important to note that the proposed framework is safe to be used in different regions, industries and sector development maturity stages. It is optimal to be used in transitional economies, where structural constraints cause phased approach to deal with the inefficiencies of resources first to be able to adopt AI strategy faster [86,87].

AISTF-SME provides a foundation for developing maturity models specific to AI in SMEs. Such models could integrate quantitative benchmarks for performance and sustainability outcomes, track transformation over time, and support cross-national comparisons in policy evaluation or global competitiveness studies [88].

5.2. Research Limitations

This study has several limitations that qualify the inferences drawn from the Serbian SME context and suggest avenues for future work.

Design and inference: The evidence is cross-sectional, which constrains causal interpretation of the sequence AI Readiness (AR) → Strategic AI Alignment (SAA) → Outcomes. Although the theorized ordering is grounded in the prior literature, alternative explanations (e.g., reverse causality from high-performing firms investing more in readiness and alignment) cannot be excluded. Longitudinal or panel designs, event studies around major AI adoptions, and quasi-experimental approaches (e.g., staggered policy shocks, difference-in-differences) would allow for stronger causal claims.

Sampling and generalizability: The sampling strategy relied on business networks and innovation-oriented lists, which overrepresent digitally engaged firms. Non-response bias cannot be fully ruled out, and micro-enterprises may be under-sampled relative to medium-sized firms. External validity is therefore bounded to SMEs operating in Serbia’s transition-economy setting and similar institutional environments. Replication in other countries, industries (e.g., heavy manufacturing, public services), and size bands (including large enterprises and micro-firms) is needed.

Measurement and single-source data: Constructs were measured through self-reports from a single respondent per firm, raising the risk of common-method and common-source bias, as well as social desirability inflation on outcomes. While procedural/statistical checks reduce these risks, they do not eliminate them. Future studies should triangulate with multi-informant designs (e.g., pairing executives with operations or sustainability officers) and integrate objective indicators such as metered energy use, fleet fuel logs, emission inventories, ERP-based productivity metrics, or audited financial KPIs.

Second-order Outcomes specification: In the structural model, Outcomes is modeled as a second-order latent composed of Performance Outcomes (POs) and Sustainability Orientation (SO). This parsimonious representation assumes a common driver and masks potentially different path strengths to POs vs. SO. Estimating separate structural paths (e.g., SAA → POs and SAA → SO) or evaluating a bifactor/alternative second-order specification would reveal whether alignment influences performance and sustainability differentially.

Statistical power and precision: Given the number of parameters estimated (including the second-order factor), this study may have limited power to detect small effects or interactions (e.g., moderation by firm size or industry). Power-aware designs with larger samples, pre-registered analysis plans, and sensitivity analyses would improve precision and reduce researcher degrees of freedom.

Construct validity and invariance: Although the four first-order constructs (AR, SAA, POs, SO) showed acceptable reliability/validity, some items may carry context-specific meanings in Serbia (e.g., “AI readiness” items interpreted through the lens of digitalization rather than AI per se). Future research should (i) test measurement invariance (configural/metric/scalar) across industries, firm sizes, and countries; (ii) consider formative or mixed specifications for AR if the indicators reflect non-interchangeable capabilities; and (iii) refine the PO and SO item pools using cognitive interviewing in multiple contexts.

Model specification and omitted variables: Unobserved organizational characteristics (e.g., managerial quality, slack resources, data governance maturity) could confound the estimated paths. Incorporating controls grounded in theory, collecting archival proxies (e.g., past productivity trends), or using instrumental-variable strategies where defensible would mitigate endogeneity concerns.

Data quality and robustness: Self-report scales are susceptible to ceiling effects and reference-group bias. Distributional departures (non-normality) and any missingness patterns, even when managed with modern estimators, can influence standard errors. Future work should report preregistered robustness checks (e.g., alternative item parcels, different estimators, bootstrap intervals, and leave-one-item-out tests) and benchmark results against alternative model topologies (direct AR → Outcomes path; SAA as mediator/moderator).

Temporal dynamics and policy environment: The AI and sustainability landscapes evolve rapidly due to technology diffusion and regulation. A cross-sectional snapshot may understate dynamic complementarities (e.g., learning-by-doing, capability accumulation). Repeated measures and cohort analyses around policy or market shifts would capture such temporal effects.

Taken together, these limits do not negate this study’s contribution—evidence for a coherent Readiness → Alignment → Outcomes pathway in Serbian SMEs—but they do delimit its scope. Addressing them through multi-source, longitudinal, and cross-context designs with richer objective metrics will sharpen causal interpretation and external validity.

5.3. Practical Implications

The findings carry actionable implications for SME leaders, policy institutions, innovation support organizations, and higher education. Grounded in survey evidence, cluster analysis, and case studies, they speak directly to how Serbian SMEs can use AI to accelerate digital transformation while strengthening agility and sustainability.

For SME leaders and managers, the central message is strategic alignment. The highest-performing firms—our Strategic Adopters—did not treat AI as a stand-alone upgrade; they embedded AI into ERP, CRM, and core operations to support faster decisions, superior customer experience, and adaptive planning. This reinforces the need to invest not only in tools but in leadership commitment, internal alignment, and digital competencies that translate pilots into enterprise routines.

A phased transformation approach is both feasible and effective under resource constraints. The process should start by securing the foundations—cloud infrastructure, data governance, access controls—and then introducing plug-and-play modules (e.g., chatbots, forecasting, anomaly detection) that deliver quick wins and build confidence. The growing maturity of low-/no-code platforms lowers the technical threshold, allowing non-specialists to participate. Critically, deployment should target not just efficiency, but new value propositions in personalization, resilience, and sustainability.

Sector-specific pathways are clear. In manufacturing and logistics, predictive maintenance and route optimization yielded immediate gains in our sample, including reductions of up to 14% in energy and fuel consumption, directly linking AI to cost efficiency and environmental performance. These dual benefits remain underexploited across the sector and warrant priority scaling.

Bridging the skills and translation gap requires institutionalized university–SME collaboration. Subsidized data-science internships, publicly funded proof-of-concepts, and shared AI labs (offered on a subscription basis) can democratize access to expertise and accelerate the conversion of research into commercial value. Such arrangements also help SMEs build internal product ownership and governance capabilities.

For development agencies and support organizations, the Traditionalists–Experimenters–Strategic Adopters typology provides a practical diagnostic. Traditionalists benefit from readiness audits and subsidized access to foundational digital tools; Experimenters need strategic mentoring and help formalizing governance for AI; and Strategic Adopters are candidates for scaling programs—internationalization support, sustainability-innovation funding, and participation in cross-border digital ecosystems.

From a sustainability policy perspective, the evidence points to a missed opportunity. Although AI applications for carbon tracking, waste reduction, and lifecycle modeling exist, most SMEs lacked an AISTF-SME orientation and incentives to deploy them systematically. Policymakers and industry bodies should co-develop SME-tailored ESG toolkits, lightweight sustainability KPIs, and targeted rewards—tax credits, green vouchers, or procurement preferences—for firms that integrate AI into environmental strategies.

AI’s contribution to CSR and stakeholder engagement is likewise underused. Prior work shows that eWOM and green communication mediate the link between sustainability initiatives and perceived CSR performance [89]. Serbian SMEs can strengthen trust and customer alignment by leveraging AI-enabled insight platforms and delivering personalized sustainability messaging across channels.

5.4. Policy Implications

The empirical results translate into clear policy priorities for government institutions, funding agencies, higher education, and business-support organizations seeking to accelerate AI-driven transformation among Serbian SMEs. Consistent with our RBV and TOE perspectives, the most binding constraints are human capital, access to finance, ecosystem coordination, and regulatory clarity—each of which can be addressed with targeted, scalable instruments aligned to SME realities.

First, capacity building must move beyond generic “digital skills” and focus on AI application literacy in resource management, forecasting, and sustainability analytics. More than half of surveyed SMEs identified technical expertise as a primary barrier, underscoring the RBV argument that scarce internal capabilities limit adoption and impact. National and regional programs should therefore fund subsidized, industry-aligned training co-designed by universities, technical schools, and private providers, with modular curricula that SMEs can absorb while operating under tight time and staffing constraints. Short, stackable credentials and applied project work (using firms’ own data) will speed diffusion and retention.

Second, improving access to finance is essential. Many SMEs experience public funding as complex and slow, a classic TOE “environment” barrier. Streamlined portals, simplified documentation, and fast-track micro-grants for AI pilots would lower entry costs. Performance-linked incentives—tax credits, green vouchers, or ESG-indexed grants—should reward projects that demonstrate agility improvements alongside measurable sustainability outcomes (e.g., reduced energy intensity or scrap rates). Clear guidance on eligible AI use-cases will reduce application friction and gaming risk.

Third, ecosystem collaboration must be institutionalized. Strategic Adopters in our sample benefited from informal knowledge networks and university ties that de-risked experimentation. Public–private innovation hubs, co-financed AI testbeds, and shared development labs (with pay-as-you-go access) can extend these advantages to Experimenters that have intent but lack coordination or expertise. Embedding mentorship, product ownership coaching, and lightweight MLOps support within these hubs will help firms convert pilots into repeatable capabilities.

Fourth, regulatory and institutional alignment should de-risk adoption. Traditionalists frequently cited uncertainty around compliance, ethics, and data protection. SME-oriented guidance—delivered through sectoral regulatory sandboxes—can clarify acceptable data use, model transparency, and risk controls while allowing for supervised experimentation. Mainstreaming sustainability metrics within digitalization instruments (e.g., requiring a simple baseline and follow-up on energy/material intensity) will normalize ESG integration in AI projects and signal a national commitment to responsible innovation.

Fifth, policy should be differentiated by adoption stage. The Traditionalist–Experimenter–Strategic Adopter typology offers a practical basis for targeting: Traditionalists need readiness diagnostics, onboarding support, and subsidized access to foundational tools; Experimenters benefit from strategic mentoring, governance templates, and modular AI building blocks; and Strategic Adopters warrant scale-up instruments—export support, sustainability-innovation finance, and entry into cross-border digital ecosystems. Differentiation minimizes waste and maximizes impact per public euro.

Our thematic review of Serbia’s policy landscape reveals a gap between national AI ambitions and SME-usable mechanisms. While the AI Strategy of the Republic of Serbia (2020–2025) emphasizes innovation-led growth, SME-specific funding channels and operational guidance are limited. The interviews echoed this disconnect (“We keep hearing about AI from the Ministry, but when we look for actual support—financial or technical—there’s nothing we can use;” “Policy sounds promising on paper, but implementation is delayed or non-transparent for smaller firms”). Similarly, the Digitalization Roadmap prioritizes e-government modernization but gives comparatively little attention to incentivizing SME-level AI adoption, constraining the environmental enablers highlighted by the TOE framework.

Relative to other transition economies, Serbia faces broadly similar constraints—capability gaps, funding frictions, fragmented university–SME linkages—but its EU-accession alignment offers leverage points. Programmatic co-financing, skill pipelines tied to European standards, low-threshold vouchers, and SME sandboxes can be deployed quickly without building large new bureaucracies. Prioritizing applied university partnerships—through subsidized internships, proof-of-concept grants, and shared datasets—will accelerate translation from research to firm-level value.

In sum, turning AI’s potential into durable, sustainability-oriented impact requires coordinated action across capability building, finance, ecosystems, and regulation—grounded in empirical evidence and tuned to SME constraints. Policies that reward measurable agility and sustainability gains, provide simple and speedy funding, and embed universities within SME problem-solving will help Serbian SMEs progress from pilots to scaled transformation. Doing so directly advances RQ2 (barriers/enablers) and, by enabling strategically aligned deployment, delivers on RQ1: AI integration that strengthens business agility while improving environmental and social performance.

6. Conclusions

Outlook (2025–2030)

The authors anticipate three trajectories for SME AI in Serbia: (i) operationalization—wider plug-and-play adoption in CRM/ERP and low-code ML; (ii) strategic scaling—selected SMEs integrate AI into product/service innovation and sustainability reporting; and (iii) ecosystem enablement—policy sandboxes, university–SME residencies, and vouchers accelerating SD & SPOs. We propose tracking the AI Maturity Index (yearly), share of sustainability-AI pilots, and proportion of SMEs with data governance SOPs. Two testable propositions are outlined for longitudinal follow-up.

By applying the AI-Driven Strategic Transformation Framework for SMEs (AISTF-SME) and using both quantitative and qualitative data, this study provides a nuanced view of how AI adoption unfolds in practice and how it may support national objectives related to innovation, competitiveness, and sustainable development.

The empirical results show that AI adoption in Serbian SMEs is still at a nascent stage, with marked variation across firms. The identification of three typologies—Traditionalists, Experimenters, and Strategic Adopters—highlights the diversity in digital maturity, strategic intent, and resource capabilities. Strategic Adopters, although few, demonstrate the strongest outcomes in agility, innovation, and emerging sustainability practices.

Importantly, this study finds that AI adoption is currently driven more by operational goals—such as process efficiency and customer responsiveness—than by a strategic vision of inclusive innovation or environmental sustainability. This limited use of AI for broader sustainability purposes suggests missed opportunities that could be addressed through better awareness, targeted incentives, and improved access to technical expertise.

Controlling for firm size and sectoral affiliation confirmed that AI maturity is a robust predictor of strategic outcomes, even after accounting for these potential confounders. This reinforces the importance of AI readiness as a key organizational capability for digital transformation.

However, systemic barriers continue to hinder broader integration of AI across Serbian SMEs. These include low levels of digital and data literacy, limited internal talent, financial constraints, and fragmented institutional support. A notable disconnect persists between the country’s research institutions—often well positioned in AI innovation—and SMEs, which struggle to translate that expertise into practice. Public support programs are often difficult to access, misaligned with SME needs, or overly bureaucratic.

To unlock AI’s transformative potential in Serbian SMEs and promote alignment with sustainability goals, a coordinated multi-stakeholder policy framework is essential. While these recommendations are tailored to Serbia’s institutional context, they may offer a useful reference for similar economies in the region:

• Sector-specific diagnostics and capacity-building: programs should include maturity assessments, benchmarking tools, and industry-aligned AI roadmaps, particularly for sectors like manufacturing, logistics, agri-tech, and retail.
• Financial support for sustainable AI use: authorities should expand access to AI-specific grants, innovation vouchers, and tax incentives that reward sustainability-oriented AI applications (e.g., emission reduction, ethical sourcing).
• Human capital development: upskilling should focus on both technical and strategic competencies, including short-term AI strategy certifications, interdisciplinary curricula, and work-based learning with AI providers or R&D centers.

This study, while robust in design and rich in empirical insights, is not without limitations. Its exclusive focus on Serbian SMEs limits generalizability; the findings may not fully reflect conditions in other countries. However, cross-referencing institutional conditions with regional counterparts (e.g., Bosnia and Herzegovina, North Macedonia) reveals overlapping challenges such as capital scarcity, talent gaps, and weak policy implementation. These similarities suggest that some lessons may be transferable, though not universally applicable. Other limitations include the cross-sectional nature of the data, reliance on self-reported information, and limited assessment of technical aspects of AI implementation.

Future research could address these by incorporating longitudinal tracking, external performance validation, and deeper technical analysis of AI tools in use.

Building on this work, several directions for future research emerge:

• Cross-national comparisons: studies comparing multiple CEE countries could illuminate how variations in institutional quality or EU integration trajectories affect AI adoption.
• Transformation pathways: longitudinal case studies can reveal how SMEs move between typologies over time, and what tipping points drive strategic adoption.
• Social and environmental impacts: broader assessments should examine AI’s effect on equity, employment, and resource efficiency using lifecycle and stakeholder analyses.
• Scalable diagnostic tools: future efforts could refine and digitize the AISTF-SME framework into interactive toolkits for SME self-assessment and strategic planning.

In conclusion, while this study is anchored in Serbia’s specific context, it contributes meaningfully to understanding how SMEs in transitional economies can harness AI not only for efficiency but also for strategic and sustainable transformation.