Small businesses in underserved US communities that report higher AI adoption also report substantially better financial literacy, decision-making and higher profit growth (9.5% vs 5.8%), but gains are curtailed by low training uptake, uneven education and demographic barriers.

Main Finding

AI adoption among small businesses in underserved U.S. communities is positively associated with higher financial literacy, improved decision-making quality, and faster profit growth (average 9.5% for high adopters vs. 5.8% for low adopters). However, these gains are conditional on complementary capabilities (digital literacy, contextualized tools, training) and are constrained by demographic and educational disparities. AI’s transformative potential is therefore not automatic but dependent on ecosystem supports.

Key Points

• Theoretical framing: AI treated as a strategic intangible resource under the Resource-Based View (RBV); value arises when AI is integrated with complementary capabilities.
• Positive associations:
  • Higher AI adoption correlates with higher financial literacy scores.
  • Firms reporting high AI adoption cite better decision-making quality.
  • Observed mean profit growth: 9.5% (high adopters) vs. 5.8% (low adopters).
• Barriers / mediators limiting benefits:
  • Low participation in AI training and capacity-building programs.
  • Uneven educational backgrounds and digital literacy across owners.
  • Demographic disparities (gender and age) associated with lower adoption and weaker outcomes.
• Policy- and practice-oriented recommendations highlighted:
  • Multi-stakeholder collaborations (policymakers, financial institutions, entrepreneurs).
  • Design inclusive, contextualized AI tools and training.
  • Targeted efforts to bridge digital skills gaps and support marginalized groups.

Data & Methods

• Design: Cross-sectional quantitative survey grounded in RBV theory.
• Sample: 400 small business owners operating in underserved U.S. communities.
• Instrument: Structured questionnaire measuring:
  • Self-reported AI adoption level.
  • Financial literacy (score-based measure).
  • Decision-making quality (self-assessed / Likert-type items).
  • Firm-level outcomes (reported profit growth rates).
  • Demographic and background covariates (education, gender, age, training participation).
• Analysis: Correlational/statistical comparisons between high- and low-AI adopters; mediation/association tests to identify barriers (details of specific statistical models not provided in the summary).
• Limitations to note:
  • Cross-sectional design limits causal inference (reverse causality or omitted variables possible).
  • Self-reported measures (AI use, profit growth, literacy) risk reporting bias.
  • Sample representativeness and external generalizability beyond surveyed communities not established.
  • Granularity of “AI adoption” (type, sophistication, intensity) appears limited.

Implications for AI Economics

• Complementarities drive returns: The study reinforces theory that returns to AI investment depend on complementary assets (human capital, managerial practices, contextual tools). Policymakers and economists should model AI as productive only when complements are present.
• Distributional effects and inequality: Because benefits depend on access to training and education, AI adoption can widen within-sector inequality unless interventions target underserved demographic groups (gender, age). Economic models of automation/adoption should incorporate heterogeneity in complements and access.
• Policy levers:
  • Subsidies or vouchers for AI/digital skills training may raise marginal returns to AI investments among small firms.
  • Public–private partnerships to develop contextualized, low-cost AI tools tailored to small-business needs can increase effective adoption.
  • Targeted interventions (e.g., for women entrepreneurs, older owners) are likely more efficient than one-size-fits-all programs.
• Market and finance implications:
  • Financial institutions can use AI to improve credit access for small firms, but must pair tools with capacity-building to avoid adverse selection or unequal uptake.
  • Measured productivity gains (profit growth differential) suggest potential aggregate gains if scaled, but scaling requires addressing training, infrastructure, and design constraints.
• Research and evaluation priorities:
  • Need for causal evidence (RCTs, phased rollouts) to quantify causal effects of AI tools and training on firm performance.
  • More precise measurement of AI sophistication, intensity, and use-cases to identify which applications generate the largest economic returns.
  • Longitudinal studies to observe persistence of gains and dynamic complementarities (learning-by-doing, network spillovers).
• Practical takeaway for economists: interventions that reduce frictions in complementary assets (skills, relevant software, ecosystems) may yield higher social returns than subsidizing AI tools alone.