AI Works Differently Inside SMEs: Strategic Use, Not Adoption, Tracks Agility

Artificial intelligence (AI) is spreading across small businesses, but adoption alone may tell policymakers surprisingly little about whether firms are becoming more adaptable. A study of Indian enterprises finds that AI becomes far more closely associated with operational agility when it is embedded in strategic decision-making and governance rather than merely present somewhere inside the business.

Published in Administrative Sciences, the study, "Governance-Specific Use of AI, Ownership Structure, Financing, and Operational Agility: Evidence from Indian SMEs," is authored by Samreen Akhtar of Saudi Electronic University. Using linked World Bank Enterprise Survey data, it examines how ownership structures, formal financing and different forms of AI use relate to firms' ability to respond to changing demand.

The findings complicate a common assumption behind digital-transformation strategies: getting more businesses to adopt AI does not necessarily mean those businesses have changed how they operate. Among the firms studied, general AI adoption had no statistically significant relationship with operational agility, while the use of AI in strategic and compliance-related functions showed strong positive associations.

Adoption figures can conceal shallow digital transformation

The study combines the World Bank Enterprise Surveys India 2025 round with the 2026 AI Follow-up Survey. After removing two regions where responses showed unusually uniform patterns, the primary analysis covered 1,215 formal enterprises, while several robustness checks tested whether the exclusions materially affected the results.

Around 35.5% of the firms had adopted AI, yet general adopter status was not associated with the study's measure of operational agility. Firms using AI were statistically no more likely than non-users to report that they could absorb a 20% increase in demand without recruiting additional workers.

Among AI adopters, however, the organizational location of the technology produced a very different result. Firms using AI for strategic decision-making had odds of reporting operational agility almost four times as high as other adopters, while those using AI for compliance and governance functions showed a similarly strong association.

The study does not establish that AI caused greater agility. More capable firms may be simultaneously better positioned to integrate AI and respond to demand. Even with that limitation, the divergence between adoption and functional use is difficult to dismiss: a binary measure of AI uptake captures technological presence, but says little about whether the technology has entered the firm's operating architecture.

Nearly two-thirds of adopters reported neither strategic nor compliance-related AI use. For governments tracking digitalization through adoption rates, the implication is significant. Rising uptake can coexist with limited organizational change if firms use AI only for peripheral activities rather than incorporating it into decisions that shape resource allocation, compliance or strategy.

Ownership shapes where AI enters the business

The research also brings corporate governance into the AI discussion. In 46.3% of the sample, the largest owner was also the firm's top manager, allowing the study to compare enterprises where ownership and executive authority were concentrated with those where the two roles were separated.

Among AI adopters, owner-managed firms had 70% lower odds of using AI in strategic decision-making than firms with separate ownership and management, after accounting for incorporation and firm-size categories. The result challenges the assumption that concentrated authority automatically accelerates deeper technology deployment by reducing internal approval barriers.

One interpretation advanced by the study is that owner-managers may depend more heavily on personal knowledge, direct control and accumulated experience when making strategic choices. Such firms can still make decisions quickly, but speed based on centralized judgment is not the same as formalizing decisions around technology-supported analysis.

Legal form offered a weaker explanation. Incorporation was not significantly associated with the use of AI for compliance and governance functions in the primary analysis. Formal reporting obligations therefore did not translate automatically into greater use of AI for governance-related activities within the firms observed.

This creates an important distinction for SME digital policy. Technology readiness cannot be understood entirely through infrastructure, software access or finance. Decision rights, managerial habits and willingness to formalize strategic processes may influence whether AI progresses from an available tool into an organizational capability.

Finance supports agility, but not necessarily AI integration

A second pathway emerged around formal finance. 32.8% of firms held a bank loan or credit line, and owner-managed businesses were more likely to possess such financing. Their odds of holding a loan or credit facility were more than twice those of firms where ownership and management were separated.

Formal financing was also strongly associated with operational agility. Firms holding a loan or credit line had 4.20 times the odds of reporting that they could absorb a 20% increase in demand without adding employees, compared with firms without those forms of finance.

The financing variable requires careful interpretation. It measures possession of a bank loan or credit line rather than affordability, borrowing conditions, rejected applications, informal funding or unmet financial need. Firms with credit may have greater scope to adjust inventories, working capital or productive capacity, but the research cannot determine precisely how those funds were used.

Perhaps more revealingly, formal financing showed no statistically detectable association with general AI adoption, strategic AI use or compliance-related AI use. Access to credit and AI integration therefore appear as separate dimensions of organizational capability in this dataset rather than one straightforward investment chain running from finance to technology.

The study's decomposition reinforces this contrast. Owner-manager overlap had a negative indirect association with agility through strategic AI use but a positive indirect association through formal financing. Both patterns survived additional controls, although they remain statistical pathways rather than evidence of causal mediation.

For policymakers, the result cautions against assuming that credit programmes will automatically accelerate meaningful AI adoption, or that digital-support initiatives can compensate for financing constraints. Firms may require both financial flexibility and organizational capabilities, but the study does not show that one reliably generates the other.

SME policy needs to look inside the firm, not just at technology uptake

The strongest policy lesson concerns measurement. Digital-transformation programmes frequently treat adoption as the key milestone: whether businesses have acquired a technology, begun using it or crossed a basic threshold of digital participation. The evidence here suggests that such indicators can stop well before the organizational changes associated with stronger operational performance become visible.

Support programmes may therefore need to distinguish between introductory adoption and deeper functional integration. Training focused solely on operating AI tools could miss equally important capabilities involving interpretation, accountability, decision authority, data access and the incorporation of AI outputs into managerial routines.

The study also reports strong optimism among adopters: firms using AI were substantially more likely to expect future productivity gains as the technology advances. Expectations, however, should not be confused with realized performance. The absence of a general adoption-agility relationship indicates that enthusiasm for AI can precede meaningful organizational integration.

Evidence on demand for government-supported AI workforce training was less stable. Ownership characteristics changed in significance when additional controls were introduced, preventing a firm conclusion about which types of enterprises should be targeted. An exploratory finding nevertheless showed a pronounced association between operational agility and demand for AI training support, a relationship the author identifies as a subject for further investigation rather than established evidence.

Several limitations restrict how widely the results should be applied. The follow-up sample overrepresented digitally intensive sectors, excluded informal firms and businesses with fewer than five employees, and was analyzed without follow-up-adjusted sampling weights. The findings therefore describe the observed formal-enterprise sample rather than the Indian SME sector as a whole.

Operational agility is also defined narrowly. A firm's ability to absorb additional demand without hiring may reflect flexibility, but it could also capture spare capacity, outsourcing, automation or differences in labour intensity. Governance-specific AI use was measured only among adopters, creating further risks of selection bias and limiting generalization beyond that subgroup.

Nevertheless, the study reframes an increasingly important policy problem. AI diffusion can be counted quickly; organizational transformation cannot. Firms may acquire the same technology while deriving very different value depending on who makes decisions, how formalized those decisions are, where AI is deployed and whether businesses have the resources to adjust when conditions shift.