AI adoption in small and medium firms correlates with higher reported value creation—largely through task automation and better database use—while transaction digitization, though boosted by AI, does not independently raise value creation.

Main Finding

Artificial intelligence (AI) has a positive effect on value creation in MSMEs in Cameroon both directly and indirectly. The indirect effects operate through task automation and use of databases; AI also increases transaction digitization, but transaction digitization by itself did not have a positive effect on value creation in this sample.

Key Points

• Dataset and sample: 514 distinct MSMEs in Cameroon (700 questionnaires distributed; 566 returned; 403 usable paper responses + 111 online = 514 total). One respondent per company.
• Theoretical framing: resource-based view and core-competence literature (Penrose, Barney, Prahalad & Hamel) — AI as a firm resource that creates value when combined with organizational competencies (databases, automation, learning).
• Hypotheses tested:
  • H1: AI → value creation (positive) — supported.
  • H2: AI → database use → value creation — mediation supported.
  • H3: AI → task automation → value creation — mediation supported.
  • H4: AI → transaction digitization → value creation — AI→digitization supported, but digitization→value creation not supported (no mediation).
• Original contribution: simultaneous examination of three mediators (database use, task automation, transaction digitization) linking AI to value creation in MSMEs; finding that digitization alone does not translate to value unless paired with other capabilities.
• Contextual note: sample spans sectors (commerce, services, finance, manufacturing, agrifood, etc.) and company sizes per Cameroonian MSME definitions.

Data & Methods

• Survey instrument administered in urban areas (Douala, Yaoundé, Bafoussam) and online across regions where MSMEs operate.
• Sample size: 514 firms; respondents mainly managerial but included operational staff involved with digital processes.
• Analytical approach:
  • Exploratory Factor Analysis (EFA) to identify measurement structure,
  • Confirmatory Factor Analysis (CFA) to validate constructs,
  • Structural Equation Modeling (SEM) to estimate direct and mediated effects.
• Constructs measured: perceived AI use (and capabilities), value creation (economic and perceived/customer value dimensions), task automation, database use, transaction digitization.
• Limitations acknowledged by authors (implicit or explicit): cross-sectional survey, self-reported measures, single-respondent per firm, Cameroon-specific context (limits to generalizability), and potential omitted variables or common-method bias.

Implications for AI Economics

• Microeconomic / firm-level:
  • AI raises firm value directly but gains are amplified when firms invest in complementary capabilities — especially database creation/management and task automation. Policymakers and managers should treat AI investments as part of a bundle: algorithms + data infrastructure + process redesign + human skills.
  • Transaction digitization alone may not increase firm value; digitization appears necessary but not sufficient. Without analytics, quality data practices, automation, and organizational learning, digitized transactions may not generate measurable gains.
• Labor and productivity:
  • Results support complementarity between AI and human skills through automation that reallocates workers to higher-value tasks; however, substitution risks remain — policy should emphasize reskilling and upskilling to capture productivity benefits.
  • Measurement of AI’s macroeconomic impact must account for heterogeneity in complementary investments (data infrastructure, automation, skills). Cross-firm differences in these mediators help explain why aggregate AI impacts may be muted in some estimates.
• Policy & investment priorities:
  • Promote data governance, affordable data-management tools, and training for MSMEs to exploit databases and automation effectively.
  • Encourage bundled support (technical assistance + finance + training) rather than solely subsidizing digitization tools.
  • Regulation should address data security and continuity risks that can undermine value from digitization and AI.
• Research implications:
  • Macro estimates of AI’s contribution to GDP should explicitly model complementarities (data, automation, skills) rather than treating AI adoption as a homogeneous input.
  • Future empirical work should use longitudinal designs and objective performance metrics (productivity, profitability, employment composition) to track causal dynamics and heterogeneity across sectors and firm sizes.

Suggested next research steps (concise): longitudinal studies tracking post-adoption outcomes; objective performance data; cross-country comparisons to test whether the non-effect of transaction digitization generalizes or reflects capability gaps in this context.