Economic Turbulence Is Raising AI's Appeal for Small Businesses, but Risks Are Holding Back Wider Use

Economic instability is reshaping how smaller businesses evaluate artificial intelligence, creating stronger incentives to explore its advantages without necessarily encouraging greater technological use. As firms navigate fluctuating demand, supply-chain uncertainty and shifting trade conditions, AI increasingly enters strategic calculations. Yet the decision to deepen its integration depends on more than the pressures businesses face. Perceived benefits, implementation risks and technological familiarity play distinct roles in determining how extensively organizations use AI.

A study titled Economic Turbulence and Organizational AI Use in MSMEs: The Competing Roles of Perceived Benefits and Risks, published in the journal Systems, examines these relationships among 537 Chinese micro-, small- and medium-sized enterprises (MSMEs). Authored by Rob Kim Marjerison, Hee Kyung Jeun, George Kuan and Jong Min Kim, the research reveals that economic turbulence is positively associated with perceived AI benefits, while concerns about technological risks remain negatively associated with organizational use.

The research also challenges a common assumption about digital transformation: greater familiarity with technology does not necessarily reduce apprehension. Businesses reporting stronger AI familiarity also reported greater awareness of associated risks. Together, the findings offer a more nuanced account of why some smaller enterprises integrate AI more extensively than others, even when facing similar economic uncertainty.

Economic uncertainty is changing the business case for AI

Economic turbulence places considerable pressure on smaller businesses because their financial, technical and managerial resources are generally more constrained than those of larger enterprises. Unpredictable customer orders, tariff-policy changes, demand fluctuations and supply-chain disruptions can undermine established operating arrangements, forcing managers to reassess how they allocate resources and maintain business responsiveness.

AI presents potential opportunities in this environment. Applications supporting information processing, transaction efficiency, operational problem-solving and business decisions may become more attractive when conventional processes struggle to accommodate rapidly changing conditions. Nevertheless, identifying possible advantages is different from demonstrating that technology investments will deliver measurable organizational improvements.

The researchers examined these competing considerations through an online survey of 537 Chinese enterprises already using AI. The sample comprised 105 microenterprises, 247 small businesses and 185 medium-sized firms across technology, manufacturing, services and agriculture. Using statistical modelling, the researchers assessed how economic turbulence, perceived benefits, perceived risks and familiarity related to reported organizational AI use.

The results revealed a statistically significant positive association between economic turbulence and perceived AI benefits, with a standardized coefficient of 0.125. Perceived benefits were also positively associated with organizational AI use, suggesting that businesses recognizing greater technological value tended to report more extensive integration of AI into their operations.

More revealing was the distinction between direct and indirect relationships. Economic turbulence had a significant positive indirect association with AI use through perceived benefits, with an estimated coefficient of 0.058. However, its direct association with organizational use was not statistically significant. The overall turbulence–use association remained positive and significant, highlighting the importance of distinguishing the different relationships identified by the model.

These findings are consistent with the possibility that uncertain commercial conditions make AI's operational advantages more prominent in managerial evaluations. They do not establish that economic instability causes firms to expand AI use, since the survey measured organizational conditions and technology assessments at a single point in time.

The AI investment dilemma extends beyond technological benefits

The research identifies perceived risks as a significant counterweight to the advantages associated with AI. While perceived benefits were positively associated with organizational use, with a standardized coefficient of 0.250, perceived risks demonstrated a negative association of −0.382. Both relationships were statistically significant, indicating that favorable evaluations and implementation concerns were separately associated with the extent of AI use.

The risk assessments covered several challenges relevant to smaller enterprises, including data storage, security and confidentiality, insufficient technological knowledge, application complexity and concerns about AI regulation. Such issues can become particularly consequential when firms lack specialized personnel or sufficient financial capacity to manage unsuccessful technology investments.

However, the relationship between economic turbulence and perceived risks did not follow the pattern initially anticipated by the researchers. Economic turbulence was not significantly associated with greater AI risk perceptions, and the proposed indirect relationship between turbulence and organizational use through risks was also statistically unsupported.

The distinction prevents a simplistic interpretation in which worsening economic conditions simultaneously intensify technological opportunities and implementation concerns. The findings instead suggest that perceived risks may remain important regardless of whether businesses experience greater commercial instability. The researchers identify organizational readiness and institutional conditions as possible explanations, although these mechanisms were not directly tested.

This distinction carries practical relevance for technology providers and business-support organizations. Demonstrating AI's efficiency advantages may not adequately address concerns about confidentiality, regulatory responsibilities or implementation complexity. The study therefore points toward the potential value of transparent technical requirements, accessible support and clearly defined safeguards alongside explanations of anticipated business benefits.

For smaller enterprises, a staged approach to implementation may offer a way to evaluate potential returns while containing uncertainty. Limited pilots, explicit performance indicators and systematic assessments before wider deployment are among the measures discussed by the researchers. Their effectiveness, however, was not directly evaluated.

Greater AI familiarity brings more use, but also sharper risk awareness

Greater familiarity with AI was positively associated with reported organizational use, producing a standardized coefficient of 0.262. The relationship suggests that enterprises with stronger exposure to AI also tend to incorporate the technology more extensively into business activities.

Yet familiarity did not produce uniformly favorable evaluations. Its association with perceived benefits was statistically insignificant, while its relationship with perceived risks was positive and significant, with a standardized coefficient of 0.273. Businesses reporting greater familiarity therefore also reported heightened awareness of potential technological difficulties.

The finding complicates the assumption that technological education necessarily reduces resistance. Greater exposure may allow managers to recognize implementation requirements, technical limitations and governance responsibilities that are less apparent during initial consideration. Familiarity can consequently coexist with caution rather than eliminating it.

For policymakers developing MSME digitalization initiatives, the relationship suggests that awareness campaigns and introductory training may be insufficient when delivered without practical implementation assistance. Experiential learning, sector-specific demonstrations and guidance on data protection could help businesses evaluate both technological possibilities and organizational responsibilities.

The researchers nevertheless caution against treating familiarity as an independent measure of technological knowledge. Several questionnaire items captured routine AI use, creating potential overlap with the organizational-use measure. Accordingly, part of the observed relationship may reflect similarities in what respondents were asked to assess.

The statistical model explained 70.6% of the variation in reported organizational AI use, alongside 38.7% of perceived-benefit variation and 31.2% of perceived-risk variation. These results indicate substantial explanatory power within the surveyed sample, although they do not establish causation or independently verify organizational productivity improvements.

Small-business AI policy needs capability building alongside risk management

The research offers a framework for reconsidering how public institutions encourage AI use among resource-constrained enterprises. Policies emphasizing technological promotion or financial incentives may have limited effectiveness when firms continue to face uncertainty over data security, implementation requirements, technical skills and organizational disruption.

The authors identify several potential approaches, including demonstration projects, advisory services, workforce training, technical assistance and opportunities for low-risk experimentation. Shared AI service platforms could also allow smaller enterprises to explore applications without immediately bearing the full cost of developing specialized technological capabilities.

Partnerships involving governments, universities, industry associations and technology providers may help extend experiential learning opportunities to enterprises with limited internal expertise. Locally tailored support could also address differences in access to digital infrastructure, skilled personnel and implementation services, although the study did not directly evaluate region-specific policy effectiveness.

Important limitations constrain how widely these conclusions can be applied. The research relied on voluntary, non-probability recruitment through WeChat and included only Chinese firms already using AI. Individual respondents reported organizational conditions, while their specific roles were not separately recorded. Consequently, the findings cannot explain initial AI adoption among nonusers or establish population-wide patterns.

The cross-sectional design also leaves open whether perceived benefits encourage greater use or whether extensive use leads businesses to recognize more benefits. Future investigations involving longitudinal data, independently verified technology use, multiple organizational informants and cross-country comparisons could help clarify these relationships.

For developing economies, the study raises questions about how smaller firms can build technological capabilities without assuming disproportionate financial and operational risks. Differences in institutional support, governance arrangements and digital infrastructure warrant further investigation before extending the Chinese findings to other economic settings.