A Systematic Literature Review by Daniel da Costa, January 2026
About AGENS
AGENS is a strategic consultancy specialising in digital transformation and the implementation of artificial intelligence solutions for the Portuguese business market. With a focus on humanising technology and enhancing human capabilities through AI, AGENS positions itself as a strategic partner for companies seeking to grow in a sustainable and innovative way.
Report Summary
Artificial Intelligence is being adopted more and more by Small and Medium-sized Enterprises, yet many initiatives fail to generate a positive Return on Investment, especially in resource-constrained settings. This study carries out a systematic literature review to identify the factors that influence the achievement of ROI in AI implementations in SMEs. Using a PRISMA-based methodology, the review synthesises evidence on value dimensions, critical success factors, barriers and enablers, as well as cross-cutting organisational and sustainability impacts.
The results indicate that a positive ROI emerges from a multidimensional interplay between operational efficiency gains, human capital development, innovation capabilities and sustainability-oriented outcomes, rather than from technology adoption alone. Existing approaches to ROI assessment are found to be fragmented and poorly suited to the reality of SMEs. To close this gap, the study proposes an integrated, process-oriented framework that structures AI decision-making and implementation into sequential phases designed to increase the likelihood of achieving a positive ROI. The results further suggest that alignment with the Sustainable Development Goals emerges as a secondary outcome of well-designed AI implementations. The study contributes to theory and practice by reframing how AI-driven ROI is understood in SMEs and by providing actionable guidance for managers and policymakers.
Keywords Artificial Intelligence, Small and Medium-sized Enterprises, Return on Investment, Technology Adoption, Systematic Literature Review, Sustainable Development Goals
Introduction
Artificial Intelligence has emerged as a critical determinant of competitive advantage across global industries, and organisations increasingly use it to optimise operations and create new value propositions. Behind this narrative of technological promise, however, lies a contrasting reality: despite the acceleration of AI adoption, success rates remain surprisingly low (Ransbotham et al., 2020) [1]. The studies by Ransbotham et al. (2020) indicate that seven in ten companies report minimal or no impact from their AI initiatives, with many projects failing to deliver the expected business value.
This disparity becomes even more evident in the context of Small and Medium-sized Enterprises. Although they account for 99% of all enterprises in the European Union and contribute approximately 56% of total economic value added (European Commission, 2023) [2], SMEs face disproportionate barriers to AI adoption. Only 8% of European SMEs have adopted AI technologies, compared with significantly higher rates among large enterprises (European Investment Bank, 2021) [3].
The implications go beyond the performance of individual companies. As large corporations use AI to achieve substantial gains in operational efficiency and revenue growth (McKinsey Global Institute, 2021) [4], SMEs that do not adopt these technologies risk progressive marginalisation in an increasingly AI-driven market. This dynamic reinforces a cycle in which large companies accumulate competitive advantages through AI, while resource-constrained SMEs struggle to justify the initial investment (Malik et al., 2021) [5], contributing to growing asymmetries in market power.
At the heart of this challenge lies the measurement and achievement of Return on Investment. While large companies have dedicated resources to manage the complexity of AI implementation, SMEs must approach these investments with greater caution because of tighter financial constraints and higher opportunity costs. Even so, the literature on how SMEs can achieve a positive ROI through AI implementation remains fragmented, and existing frameworks mostly reflect the realities of large-scale implementation.
This gap has significant consequences. Without evidence-based guidance for assessing, implementing and measuring AI initiatives, SME managers face a binary choice: forgo AI adoption and accept competitive decline, or press ahead with implementation without adequate support, increasing the risk of costly failure.
The urgency of this challenge is further amplified by rapid advances in AI capabilities. The emergence of Generative AI has lowered the technical barriers to entry, opening a window of opportunity for SMEs to overcome traditional implementation challenges (Brynjolfsson et al., 2023) [6]. That window is narrowing, however, as large companies rapidly integrate these technologies and establish new competitive advantages, raising the cost of late adoption for SMEs.
Research Gap and Contribution of the Study
Despite growing recognition of the strategic importance of AI, a systematic synthesis of the factors associated with achieving a positive ROI in SME contexts remains largely absent from the literature. Existing studies focus predominantly on implementations in large companies, offering limited transferability to resource-constrained environments. This systematic literature review addresses that gap by consolidating fragmented evidence into an evidence-based framework tailored to the decision-making realities of SMEs.
Accordingly, the study aims to synthesise the critical factors that influence Return on Investment in Artificial Intelligence implementations in Small and Medium-sized Enterprises. Four research questions guide the analysis. The first, “What are the critical factors that influence ROI in AI implementations in SMEs?”, examines the determinants that distinguish successful initiatives from failed ones. The second, “What frameworks and methodologies exist to measure and assess AI ROI in resource-constrained organisational contexts?”, reviews existing approaches and their applicability beyond large companies. The third, “What SME-specific barriers prevent the achievement of a positive ROI, and what enablers foster successful outcomes?”, investigates the constraints and enabling conditions specific to SMEs. Finally, the fourth, “What evidence-based practices maximise the likelihood of successful AI implementation and a positive ROI in SMEs?”, translates academic insights into actionable guidance for management decision-making.
Structure of the Article
The remainder of the article is organised as follows. Section 3 reviews the literature on AI adoption in SMEs, approaches to ROI measurement and key success factors. Section 4 describes the PRISMA-based systematic literature review methodology. Section 5 presents the results of the review, including the value dimensions identified, success factors, and barriers and enablers. Section 6 discusses the results in light of the research questions and introduces an integrated framework for SME contexts. Section 7 concludes with the main contributions, implications, limitations and directions for future research.
Literature Review
Artificial Intelligence
Artificial Intelligence (AI) can be defined as the study of computational agents that receive percepts from their environment and perform actions (Russell & Norvig, 2020) [7]. This broad definition encompasses both systems that replicate human thinking and those that operate according to principles of ideal rationality. Russell and Norvig (2020) [7] propose a taxonomy that organises AI research along two dimensions: (1) thinking versus acting and (2) human versus rational, giving rise to four distinct paradigms that characterise the historical evolution of the field.
Within AI, Machine Learning (ML) is a subfield focused on algorithms that improve automatically through experience, without explicit programming for each task (Banh & Strobel, 2023) [8]. Deep Learning (DL), in turn, is a specialisation of ML based on artificial neural networks with multiple layers of abstraction (Goodfellow et al., 2016) [9]. This conceptual hierarchy, in which AI encompasses ML, which in turn encompasses DL, is fundamental to understanding the different technological approaches available for business applications.
The recent emergence of Generative Artificial Intelligence (GenAI) marks a paradigm shift from discriminative to generative tasks. Banh and Strobel (2023) [8] define GenAI as systems capable of producing new, realistic content, such as text, images, code and music, among others, which reflects the characteristics of the training data without replicating it directly. The underlying architectures include Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs) and, more recently, transformer-based Large Language Models (LLMs) (Duan et al., 2019) [10].
In business contexts, GenAI offers applications that differ from those of previous generations of AI. While traditional ML structures focus on optimising existing processes through prediction, GenAI enables the automation of creative tasks, the generation of insights from unstructured data and the personalisation of content at scale (Duan et al., 2019) [10]. This ability to “create”, rather than merely “analyse”, represents a qualitative change in AI’s value potential for organisations.
Return on Investment: Conceptual Framework
Return on Investment (ROI) is a financial metric that assesses the performance of an investment by measuring net gain relative to cost. The basic formula, shown below, appears simple, but applying it to technology investments reveals significant complexities (Botchkarev, 2011) [11].
ROI = (Gain from Investment − Cost of Investment) / Cost of Investment
For investments in information systems and technology, ROI measurement faces three main conceptual challenges. First, timing: technology benefits often materialise over different time horizons from the costs, creating asymmetries between the initial investment and the accumulated returns. Second, attribution: the difficulty of isolating the specific impact of a technology when several organisational variables change at the same time. Third, and most critical, the quantification of intangibles: benefits such as improved customer satisfaction, faster decision-making or innovation capability are real, but resist direct monetisation (Botchkarev, 2011) [11].
The distinction between tangible and intangible benefits is not merely academic; it has substantial practical implications. Botchkarev (2011) [11] shows, through real cases, that an ROI calculated using tangible benefits alone may stand at 108%, while including estimated intangibles raises the same project to 563%. This order-of-magnitude difference suggests that purely financial ROI analyses systematically underestimate the value of technology investments. This point is particularly relevant to the way SMEs assess investments in AI.
The accuracy of ROI estimates in technology projects is itself a critical variable. Botchkarev (2015) [12] shows, through Monte Carlo simulations, that relatively modest errors in estimating costs and benefits (+10%) can produce significant inaccuracies in the calculated ROI, calling into question excessive reliance on single ROI values without uncertainty ranges. This is especially relevant for emerging technologies such as AI, where both costs and benefits are less predictable than in traditional IT investments.
Characterisation of Micro, Small and Medium-sized Enterprises
The European Union defines Small and Medium-sized Enterprises (SMEs) through Recommendation 2003/361/EC on the basis of three cumulative criteria: number of employees, annual turnover and annual balance sheet total (European Commission, 2003) [13]. Specifically, a company is considered an SME if it employs fewer than 250 people and has an annual turnover not exceeding €50 million or a balance sheet total not exceeding €43 million. Within this category, a distinction is made between micro-enterprises (fewer than 10 employees, ≤€2 million), small enterprises (fewer than 50 employees, ≤€10 million) and medium-sized enterprises (fewer than 250 employees, ≤€50 million).
Beyond the quantitative thresholds, the EU definition incorporates the criterion of independence: linked or partner enterprises must aggregate their figures to determine their category. This distinction is essential, as it affects access to specific SME funding and support programmes.
SMEs account for 99% of all enterprises in the EU and employ around two thirds of the private-sector workforce (European Commission, 2003) [13], making them the engine of the European economy. However, their organisational characteristics, such as limited resources, less formalised decision-making structures, and reduced access to capital and specialised talent, create both specific constraints and specific advantages for technology adoption. While large companies benefit from economies of scale and the capacity to absorb risk, SMEs show greater agility, faster decision cycles and greater proximity to the customer. This duality is particularly relevant in the context of AI, where the initial investment can be significant, but the benefits of personalisation and rapid adaptation are substantial.
Theoretical Integration and Research Gap
The intersection of AI, ROI and SMEs remains a fragmented research space. Although the literature on AI in large organisations is abundant, and generic ROI frameworks for technology exist, there is still a lack of systematic syntheses that specifically address how to measure the ROI of AI implementations in resource-constrained contexts, which factors determine the success or failure of such implementations in SMEs, and which evidence-based practices can guide SME decision-makers when faced with both the promise and the complexity of AI.
The PRISMA methodology (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) offers a rigorous protocol for carrying out this synthesis in a transparent and replicable way, making it possible to identify, assess and synthesise all the relevant evidence available. This systematic literature review seeks to close this gap through a comprehensive search covering ROI frameworks, critical success factors, SME-specific barriers and enablers, and empirically validated good practice.
Study Methodology
Research Design
This systematic literature review followed the PRISMA 2020 guidelines, Preferred Reporting Items for Systematic Reviews and Meta Analyses Page et al., 2021 [14], with the aim of ensuring transparency, replicability and methodological rigour. The PRISMA framework provides a standardised approach for identifying, selecting and synthesising scientific literature, reducing bias and ensuring comprehensive coverage of the relevant studies.
A single-database search strategy was adopted, using Scopus (Elsevier) as the main source of information. Several factors justified this choice. First, Scopus offers multidisciplinary coverage spanning areas essential to the scope of this study, such as management, business, computer science and engineering. Second, it offers robust indexing of recent literature, which is particularly relevant given the emerging nature of generative artificial intelligence technologies and the focus on contemporary implementations in the period between 2020 and 2024. Finally, practical constraints linked to the review timetable called for a more focused approach, favouring depth over breadth. Although relying on a single database is a potential limitation in terms of exhaustiveness, the thematic scope and the careful construction of the search string were designed to maximise the retrieval of relevant literature from within the vast Scopus repository.
Information Sources and Search Strategy
The systematic search was carried out on 2 December 2024 using the Scopus database. The search strategy used Boolean logic combining three conceptual groups: AI technologies, business value measurement and the SME context. The full string applied to titles, abstracts and keywords is shown in Figure 1.
The search was limited to publications from 2020 onwards, in order to capture recent developments in AI implementation, in particular the emergence of generative AI. The language was restricted to English to ensure rigorous interpretation and analysis. The document types included were peer-reviewed journal articles and conference proceedings. The initial search returned 244 records in Scopus.
Studies were included if they cumulatively met the following criteria:
- they addressed the implementation of AI or machine learning in organisational contexts, and not merely the theoretical development of algorithms;
- they explicitly discussed or measured business value, ROI, financial performance or operational outcomes arising from AI adoption;
- they focused on small and medium-sized enterprises as the study population, or presented results clearly transferable to SME contexts.
Study Selection Process
The selection process followed the PRISMA guidelines and took place in several phases. In the identification phase, the Scopus search returned 244 records. Before formal screening, eligibility filters were applied to ensure minimum requirements of quality and accessibility, resulting in 60 records going forward to detailed assessment. These filters excluded publications with no citations at all, articles whose full text was not available through institutional access, and a small number of records incorrectly indexed as predating 2020, despite the search parameters.
The screening phase involved a systematic analysis of the titles and abstracts of the 56 records against the eligibility criteria. Each record was assessed independently for its relevance to the research questions. Studies that did not meet one or more inclusion criteria were excluded, with the reason documented. This process led to the exclusion of 40 studies: 23 lacked sufficient focus on AI implementation, dealing with generic digitalisation, cloud computing or analytics without specific AI technologies; 7 did not adequately address the SME context or present transferable conclusions; 5 did not discuss business value, ROI or performance outcomes; and 5 were excluded for other reasons, namely for being overly focused on very specific knowledge domains of little relevance to the study. The remaining 16 studies met all the eligibility criteria and were included in the final synthesis. Figure 2 presents the complete PRISMA diagram documenting the selection process. The final corpus of 16 studies, although apparently small in absolute terms, reflects the highly specific intersection of three distinct domains.
All the included studies were published in indexed, peer-reviewed journals or conferences, and provided sufficient methodological detail to assess their credibility and relevance. This approach is in line with pragmatic systematic review methodologies, recognising that quality assessment in interdisciplinary business and technology research poses specific challenges compared with clinical or experimental domains.
Additional Identification of Articles
To ensure comprehensive coverage, a retrospective analysis of the references (backward citation searching, or snowballing) of the 16 articles initially included from the database search was carried out. The references were assessed against the same eligibility criteria defined for the systematic search. This complementary approach identified a further 7 relevant articles that met all the inclusion criteria, resulting in a final corpus of 23 articles for analysis.
Results
Descriptive Profile of the Selected Studies
The final corpus of studies included in this review consists of empirical investigations, systematic literature reviews and applied validation studies that examine the adoption of Artificial Intelligence in Small and Medium-sized Enterprises across different economic contexts. The literature analysed covers SMEs operating in both developed and developing economies, including the European Union, South Asia, the Middle East and Sub-Saharan Africa, reflecting the global relevance of AI adoption across highly heterogeneous institutional and resource contexts (Takawira & Pooe, 2025; Wilczynska et al., 2024; Soomro et al., 2024) [15][16][17].
Methodologically, the selected studies rely mostly on quantitative approaches, including Structural Equation Modelling and hybrid SEM–ANN techniques, complemented by qualitative analyses and systematic reviews. In technological terms, the corpus covers a broad range of AI applications, including generative AI systems, AI-assisted digital marketing, financial technology solutions, worker assistance systems and implementations associated with Industry 4.0 and 5.0 (Basri, 2020; Konur et al., 2023; Bahaw et al., 2025) [18][19][20].
This diversity provides a sufficiently robust and heterogeneous empirical basis for analysing how value creation, ROI achievement, success factors and the broader organisational impacts of AI play out in SME contexts.
Value and ROI Dimensions Identified in AI Implementations
Across the literature reviewed, ROI in AI implementations is rarely conceptualised as an exclusively financial construct. Although financial performance remains central, the evidence consistently points to SMEs extracting value from AI through interconnected dimensions of a financial, operational, customer-centred and strategic nature.
Financial and Operational Value
Direct financial and operational benefits are the most immediately observable dimension of AI-related ROI. Several empirical studies indicate that AI adoption increases productivity, reduces operating costs and improves overall efficiency. Quantitative evidence shows that access to AI tools can raise individual productivity by around 14%, with disproportionately higher gains among less experienced workers, highlighting AI’s role as a productivity equaliser in SMEs (Brynjolfsson et al., 2023) [7].
At the organisational level, AI-supported automation and analytics contribute to significant cost reductions in key business functions, including customer service (Basri, 2020) [18], supply chain management (Enshassi et al., 2024) [21], finance (Soomro et al., 2024) [16] and human resources (Soomro et al., 2025) [22]. These efficiency gains are often associated with improvements in profitability and business growth, particularly in AI-assisted digital marketing, where most SMEs report positive performance impacts after adoption (Basri, 2020) [18].
Customer-Centred Value
Beyond operational efficiency, AI creates value by transforming the way companies interact with customers. The literature highlights AI’s ability to enable personalisation at scale, improve responsiveness and support continuous customer engagement. AI-based tools, such as chatbots, recommendation systems and predictive analytics, increase customer satisfaction and retention by delivering faster, more relevant and more consistent interactions (Enshassi et al., 2024; Magableh et al., 2024) [21][23].
The empirical results consistently associate AI adoption with improvements in customer relationship management and more effective marketing strategies, reinforcing the role of customer-centred value as an indirect but critical contributor to ROI in SMEs.
Strategic and Innovation Value
A recurring theme in the literature is the strategic value of AI as an enabler of innovation and competitive differentiation. SMEs tend to use AI not only to optimise existing processes, but also to explore new products, services and business models (Costa-Climent et al., 2023) [24]. This exploratory use is made easier by SMEs’ greater organisational agility and lower levels of structural inertia compared with large companies (Grashof & Kopka, 2022; Wilczynska et al., 2024) [25][16].
AI is conceptualised both as a General Purpose Technology and as an Invention of a Method of Inventing (Grashof & Kopka, 2022) [25], enabling the recombination of knowledge and supporting radical innovation trajectories (Machucho & Ortiz, 2025) [26]. By strengthening sensing, seizing and transforming capabilities, AI adoption reinforces SMEs’ dynamic capabilities and resilience (Soomro et al., 2024) [17], especially in volatile and crisis-prone environments (Gouveia et al., 2024) [27].
Critical Success Factors Associated with Positive ROI Outcomes
The literature is clear that a positive ROI from AI implementation does not come from acquiring technology alone. Success depends on the interplay between technological, organisational and environmental factors that work as an integrated system.
Technological Factors
Data quality and data management consistently emerge as fundamental requirements for AI success. As noted by Duan et al. (2019) [10], poor-quality data significantly undermines the performance of AI systems and can lead to substantial revenue losses, making data governance a critical determinant of ROI. Compatibility between AI solutions and existing systems, processes and working practices proves equally decisive for effective adoption and use (Ahani et al., 2017) [28].
Perceived relative advantage, usefulness and ease of use strongly influence adoption intentions, in line with the Technology Acceptance Model and its extensions (Gupta, 2024) [29]. SMEs are more likely to achieve positive results when AI systems are clearly aligned with business tasks and impose little cognitive and operational effort on users (Venkatesh & Bala, 2008) [30].
Organisational and Strategic Factors
Organisational readiness plays a central role in determining AI-driven ROI (Müller et al., 2018) [31]. Top management support is repeatedly identified as essential for resource allocation, strategic alignment and change management (Enshassi et al., 2025) [32]. AI initiatives explicitly linked to organisational strategy and supported by clear performance metrics tend to generate value in a more sustained way.
Human–AI collaboration emerges as a critical success factor. The studies warn that implementations focused exclusively on replacing human labour tend to generate short-term gains (Duan et al., 2019) [10], whereas greater value is achieved when AI complements human capabilities and facilitates the diffusion of knowledge within the organisation (Brynjolfsson et al., 2023) [7].
Structured implementation processes also influence outcomes (Tornatzky & Fleischer, 1990) [33]. Innovation is understood as a continuous process rather than a one-off event, requiring iterative adaptation, learning and organisational integration (Konur et al., 2023) [19].
External and Ecosystem Factors
External conditions significantly affect SMEs’ ability to achieve ROI from AI. Support from government and business incubators (Takawira & Pooe, 2025) [15], targeted funding and innovation policies help mitigate financial constraints and adoption risks (Wilczynska et al., 2024) [16]. Participation in collaborative networks involving universities, technology providers and peer companies facilitates knowledge sharing and capability building, while social influence within professional networks reduces perceived risk and accelerates adoption (Gupta, 2024) [29].
SME-Specific Barriers and Enabling Conditions in Achieving ROI
Despite AI’s demonstrated potential, SMEs face a specific set of barriers that can undermine successful implementation and the achievement of ROI.
Financial constraints emerge as the most frequently cited barrier (Takawira & Pooe, 2025) [15], including high upfront investment costs, limited access to credit and low levels of financial literacy among entrepreneurs (Enshassi et al., 2025) [32]. Skills shortages compound these challenges, as SMEs struggle to attract and retain qualified AI talent while lacking the internal capacity to provide training (Bründl et al., 2025) [34].
Data-related challenges are another critical obstacle. Many SMEs lack the infrastructure needed to collect, manage and protect high-quality data, which limits AI effectiveness and increases implementation risk (Dwivedi et al., 2021; Machucho & Ortiz, 2025) [35][26]. Organisational and cultural barriers, such as resistance to change, weak leadership and risk aversion, often outweigh purely technical or financial constraints (Enshassi et al., 2025) [32].
On the other hand, several enabling conditions are identified. Cloud-based AI solutions and the AI as a Service model reduce upfront investment requirements and give access to advanced capabilities at lower cost (Costa-Climent et al., 2023) [24]. Strategic partnerships with universities and technology providers help close skills gaps (Konur et al., 2023) [19], while phased implementation approaches, focused on operational quick wins, generate early ROI and build internal momentum (Müller et al., 2018) [31].
Cross-Cutting Impacts Beyond Financial ROI
A comprehensive assessment of AI’s impacts on SMEs requires considering outcomes that go beyond financial ROI. The literature reviewed consistently identifies organisational, human and sustainability-oriented effects that contribute to long-term value creation and show clear alignment with specific Sustainable Development Goals.
At the organisational level, AI improves the quality of decision-making through data-driven insights and predictive analytics (Duan et al., 2019) [10], strengthening SMEs’ capacity to innovate and adapt (Gouveia et al., 2024) [27]. These outcomes are aligned with SDG 9 Industry, Innovation and Infrastructure, by supporting inclusive innovation and strengthening productive capacity.
From a human perspective, AI acts as a capability-amplifying tool. Generative AI disproportionately benefits less experienced workers by embedding good practice and accelerating skills development, contributing to workforce upskilling and more resilient employment structures (Bahaw et al., 2025) [20]. These effects align with SDG 4 Quality Education and SDG 8 Decent Work and Economic Growth. At the same time, the literature warns of risks associated with job displacement, bias and lack of transparency, reinforcing the need for responsible implementations that safeguard decent working conditions (Storey et al., 2025) [36].
Finally, AI adoption generates sustainability-oriented impacts by enabling resource optimisation, waste reduction and improved environmental monitoring, contributing to more responsible production patterns (Soomro et al., 2024) [17]. Stronger engagement with customers and communities further contributes to social value creation and to the legitimacy of SMEs (Magableh et al., 2024) [23]. These impacts align mainly with SDG 12 Responsible Consumption and Production and SDG 13 Climate Action.
To consolidate these results and provide a structured overview, Figure 3 summarises the impacts identified in the literature, explicitly mapping each AI-driven outcome in SMEs to the corresponding Sustainable Development Goals. This synthesis clarifies how the organisational, human and sustainability-oriented effects reported across the different studies translate into concrete alignments with the SDGs, while keeping an evidence-based rather than prescriptive approach.
Discussion
This study reframes the concept of Return on Investment in Artificial Intelligence implementations in SMEs, framing it as a multidimensional construct rather than an exclusively financial metric. The results show that, in SME contexts, a positive ROI results from the interplay between operational efficiency gains, human capital development, innovation capability and sustainability-oriented outcomes. This finding challenges strictly financial assessment approaches and reinforces the need for broader value assessment frameworks, suited to resource-constrained organisations.
In response to RQ1, “What are the critical factors that influence ROI in AI implementations in SMEs?”, the results indicate that a positive ROI is driven above all by a combination of technological readiness, organisational alignment and support from the external ecosystem. Data quality, compatibility, perceived usefulness and ease of use are necessary but, on their own, insufficient conditions; leadership commitment, employee involvement and structured implementation processes prove equally decisive.
With regard to RQ2, “What frameworks and methodologies exist to measure and assess AI ROI in resource-constrained organisational contexts?”, the analysis shows that existing approaches are fragmented and largely adapted from generic IT or large-company contexts. Models such as TAM and TOE effectively explain adoption dynamics, but fail to capture the sequential mechanisms of decision, implementation and reinforcement needed to achieve ROI in SMEs.
As for RQ3, “What SME-specific barriers prevent the achievement of a positive ROI, and what enablers foster successful outcomes?”, the results show that the barriers are predominantly structural and people-centred, rather than purely technical. Financial constraints, skills shortages, data limitations and resistance to change consistently outweigh technological challenges. By contrast, enabling conditions such as phased implementation, cloud-based AI solutions, strategic partnerships and public policy support significantly reduce adoption risk.
Finally, RQ4, “What evidence-based practices maximise the likelihood of successful AI implementation and a positive ROI in SMEs?”, highlights that prioritising operational use cases, adopting a logic of human–AI complementarity, making a minimum investment in data governance and building in continuous learning mechanisms are key practices for reducing uncertainty and increasing the likelihood of sustained value creation.
Integrated Framework and Contributions
Based on these results, the study proposes an integrated, process-oriented framework (Figure 4), designed to maximise the likelihood of achieving a positive ROI in AI implementations in SMEs. Unlike existing models that focus on adoption or performance in isolation, this framework links strategic readiness, use case selection, phased implementation and ROI measurement in a coherent decision logic adapted to the constraints of SMEs.
From a theoretical perspective, the work contributes to the literature on technology adoption and innovation by linking AI adoption to the achievement of ROI, and not only to adoption itself. From a practical perspective, it gives SME managers and policymakers a structured approach to reducing implementation risk while generating economic, human and sustainability-oriented value. Importantly, the results suggest that alignment with the Sustainable Development Goals emerges as a secondary effect of well-designed AI implementations, rather than as a primary adoption goal.
Conclusion
This study examined how Small and Medium-sized Enterprises can achieve a positive Return on Investment from implementing Artificial Intelligence in resource-constrained contexts. Through a systematic literature review, the results show that, in SMEs, ROI does not result from simply adopting AI, but from the coordinated interplay between organisational readiness, appropriate use case selection, phased implementation and continuous measurement and learning mechanisms. Based on this synthesis, the study proposes a process-oriented framework that structures AI decision-making and implementation so as to increase the likelihood of achieving a positive ROI. Alignment with the Sustainable Development Goals emerges as a secondary outcome of well-designed, people-centred AI implementations, rather than as a primary adoption goal.
Beyond consolidating existing knowledge, this work identifies several critical directions for future research. First, empirical validation of ROI-oriented AI frameworks in real SME contexts is essential, particularly through longitudinal studies able to capture deferred and cumulative value effects. Second, future research should explore how different types of AI use cases, operational, analytical and generative, produce distinct ROI trajectories over time. Third, greater attention is needed to measurement approaches that integrate financial and non-financial dimensions of ROI, including human capital development and sustainability-related outcomes. Finally, research should examine how public policy instruments, funding mechanisms and ecosystem-level support influence the effectiveness and scalability of ROI-oriented AI frameworks in SMEs.
By shifting the focus from whether SMEs should adopt AI to how AI implementations can be systematically designed to generate sustainable returns, this study lays a solid foundation for a more mature, impact-oriented research agenda in the field of AI and ROI in SME contexts.
References
[1] Ransbotham, S., Khodabandeh, S., Fehling, R., LaFountain, B., & Kiron, D. (2019). Winning With AI: Pioneers Combine Strategy, Organizational Behavior, and Technology. MIT Sloan Management Review and Boston Consulting Group.
[2] European Commission. (2023). Annual report on European SMEs 2022/2023. Brussels: European Commission.
[3] European Investment Bank. (2021). Digitalisation in Europe 2020-2021: Evidence from the EIB Investment Survey. European Investment Bank.
[4] McKinsey Global Institute. (2021). The state of AI in 2021. New York: McKinsey & Company.
[5] Malik, M. R., & Chakraborty, A. (2021). Gig Economy and Artificial Intelligence. Empirical Economics Letters, 20(Special Issue 2), 25–30.
[6] Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at Work (Working Paper 31161). National Bureau of Economic Research. http://www.nber.org/papers/w31161
[7] Russell, S. J., & Norvig, P. (2021). Artificial Intelligence: A Modern Approach (4th global ed.). Pearson.
[8] Banh, L., & Strobel, G. (2023). Generative artificial intelligence. Electronic Markets, 33, 63. https://doi.org/10.1007/s12525-023-00680-1
[9] Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
[10] Duan, Y., Edwards, J. S., & Dwivedi, Y. K. (2019). Artificial intelligence for decision making in the era of Big Data – evolution, challenges and research agenda. International Journal of Information Management, 48, 63–71. https://doi.org/10.1016/j.ijinfomgt.2019.01.021
[11] Botchkarev, A., & Andru, P. (2011). A Return on Investment as a Metric for Evaluating Information Systems: Taxonomy and Application. Interdisciplinary Journal of Information, Knowledge, and Management, 6, 245–269. http://www.ijikm.org/Volume6/IJIKMv6p245-269Botchkarev566.pdf
[12] Botchkarev, A. (2015). Estimating the Accuracy of the Return on Investment (ROI) Performance Evaluations. Interdisciplinary Journal of Information, Knowledge, and Management, 10, 217–233. https://doi.org/10.28945/2338
[13] European Commission. (2003). Commission Recommendation 2003/361/EC of 6 May 2003 concerning the definition of micro, small and medium-sized enterprises. Official Journal of the European Union, L 124, 36–41.
[14] Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., … Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71
[15] Takawira, B., & Pooe, D. (2025). SME readiness for Industry 5.0: A systematic literature review. Southern African Journal of Entrepreneurship and Small Business Management, 17(1), a946. https://doi.org/10.4102/sajesbm.v17i1.946
[16] Wilczynska, M., Walenia, A., Dabek, A., Salabura, D., & Kot, H. (2024). Potential Impact of Artificial Intelligence on Small and Medium Enterprises Innovation in the EU: A Perspective from Poland. Markets, Globalization & Development Review, 9(2), Article 2. https://doi.org/10.23860/MGDR-2024-09-02-02
[17] Soomro, R. B., Memon, S. G., Dahri, N. A., Al-Rahmi, W. M., Aldriwish, K., Salameh, A. A., Al-Adwan, A. S., & Saleem, A. (2024). The Adoption of Digital Technologies by Small and Medium-Sized Enterprises for Sustainability and Value Creation in Pakistan: The Application of a Two-Staged Hybrid SEM-ANN Approach. Sustainability, 16(17), 7351. https://doi.org/10.3390/su16177351
[18] Basri, W. (2020). Examining the Impact of Artificial Intelligence (AI)-Assisted Social Media Marketing on the Performance of Small and Medium Enterprises: Toward Effective Business Management in the Saudi Arabian Context. International Journal of Computational Intelligence Systems, 13(1), 142–152. https://doi.org/10.2991/ijcis.d.200127.002
[19] Konur, S., Lan, Y., Thakker, D., Morkyani, G., Polovina, N., & Sharp, J. (2023). Towards design and implementation of Industry 4.0 for food manufacturing. Neural Computing and Applications, 35, 23753–23765. https://doi.org/10.1007/s00521-021-05726-z
[20] Bahaw, P., Forgenie, D., Sadiq, G., & Sookhai, S. (2025). Generative AI for business sustainability: Examining usability, usefulness, and triple bottom line impacts in small and medium enterprises. Sustainable Futures, 10, 100815. https://doi.org/10.1016/j.sftr.2025.100815
[21] Enshassi, M., Nathan, R. J., Soekmawati, Al-Mulali, U., & Ismail, H. (2024). Potentials of artificial intelligence in digital marketing and financial technology for small and medium enterprises. IAES International Journal of Artificial Intelligence, 13(1), 639–647. https://doi.org/10.11591/ijai.v13.i1.pp639-647
[22] Soomro, R. B. et al. (2025). A SEM–ANN analysis to examine impact of artificial intelligence technologies on sustainable performance of SMEs. Scientific Reports (Online). https://doi.org/10.1038/s41598-025-86464-3
[23] Magableh, I. K., Mahrouq, M. H., Ta’Amnha, M. A., & Riyadh, H. A. (2024). The Role of Marketing Artificial Intelligence in Enhancing Sustainable Financial Performance of Medium-Sized Enterprises Through Customer Engagement and Data-Driven Decision-Making. Sustainability, 16(24), 11279. https://doi.org/10.3390/su162411279
[24] Costa-Climent, R., Haftor, D. M., & Staniewski, M. W. (2023). Using machine learning to create and capture value in the business models of small and medium-sized enterprises. International Journal of Information Management, 73, 102637. https://doi.org/10.1016/j.ijinfomgt.2023.102637
[25] Grashof, N., & Kopka, A. (2022). Artificial intelligence and radical innovation: an opportunity for all companies? Small Business Economics, 61, 771–797. https://doi.org/10.1007/s11187-022-00698-3
[26] Machucho, R., & Ortiz, D. (2025). The Impacts of Artificial Intelligence on Business Innovation: A Comprehensive Review of Applications, Organizational Challenges, and Ethical Considerations. Systems, 13(4), 264. https://doi.org/10.3390/systems13040264
[27] Gouveia, S., de la Iglesia, D. H., Abrantes, J. L., & López Rivero, A. J. (2024). Transforming Strategy and Value Creation Through Digitalization? Administrative Sciences, 14(11), 307. https://doi.org/10.3390/admsci14110307
[28] Ahani, A., Rahim, N. Z. A., & Nilashi, M. (2017). Forecasting social CRM adoption in SMEs: A combined SEM-neural network method. Computers in Human Behavior, 75, 560–578. https://doi.org/10.1016/j.chb.2017.05.032
[29] Gupta, V. (2024). An Empirical Evaluation of a Generative Artificial Intelligence Technology Adoption Model from Entrepreneurs’ Perspectives. Systems, 12(3), 103. https://doi.org/10.3390/systems12030103
[30] Venkatesh, V., & Bala, H. (2008). Technology Acceptance Model 3 and a Research Agenda on Interventions. Decision Sciences, 39(2), 273–315. https://doi.org/10.1111/j.1540-5915.2008.00192.x
[31] Müller, J. M., Kiel, D., & Voigt, K.-I. (2018). What Drives the Implementation of Industry 4.0? The Role of Opportunities and Challenges in the Context of Sustainability. Sustainability, 10(1), 247. https://doi.org/10.3390/su10010247
[32] Enshassi, M., Nathan, R. J., Soekmawati, & Ismail, H. (2025). Unveiling barriers and drivers of AI adoption for digital marketing in Malaysian SMEs. Journal of Open Innovation: Technology, Market, and Complexity, 11(1), 100519. https://doi.org/10.1016/j.joitmc.2025.100519
[33] Tornatzky, L. G., & Fleischer, M. (1990). The processes of technological innovation. Lexington Books.
[34] Bründl, P., Wegener, C., Stoidner, M., Bayer, J., Scheffler, B., Nguyen, H. G., & Franke, J. (2025). Designing worker assistance systems–Methodology development and industrial validation. Journal of Manufacturing Systems, 80, 272–293. https://doi.org/10.1016/j.jmsy.2025.02.022
[35] Dwivedi, Y. K., Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T., … & Williams, M. D. (2021). Artificial Intelligence (AI): Multidisciplinary perspectives on emerging challenges, opportunities, and agenda for research, practice and policy. International Journal of Information Management, 57, 101994. https://doi.org/10.1016/j.ijinfomgt.2019.08.002
[36] Storey, V. C., Yue, W. T., Zhao, J. L., & Lukyanenko, R. (2025). Generative Artificial Intelligence: Evolving Technology, Growing Societal Impact, and Opportunities for Information Systems Research. Information Systems Frontiers. https://doi.org/10.1007/s10796-025-10581-7
Next AGENS AI Barometer: April 2026
For more information on strategic AI consultancy and human-centred digital transformation, contact AGENS.
