AI capability alone does not guarantee startup success in emerging economies: without reliable digital infrastructure, founder AI literacy and supportive ecosystems, AI investments often fail to convert into competitive advantage, the paper argues. The authors formalize these three gating conditions into a testable framework repositioning conversion into advantage as the key theoretical problem.

Main Finding

AI capability in startups does not automatically translate into competitive advantage or improved performance in resource-constrained (emerging-economy) settings. Instead, the conversion is mediated by competitive advantage and that mediation is gated — i.e., the AI capability → competitive-advantage link holds only when specific complementary resources are present. The paper formalizes three measurable gating conditions (digital infrastructure, founder AI literacy, ecosystem support) and reframes the core theoretical problem from “build capability and advantage follows” to “under what complementarity conditions does capability become advantaging?”

Key Points

• Core claim: AI capability → competitive advantage → startup performance, with competitive advantage mediating the capability–performance relation, but the capability→advantage path is contingent on complements.
• Complementarity matters on the conversion path (capability → advantage), not just as antecedents of capability or general context. This relocation is the paper’s unique contribution.
• AI capability is multidimensional for startups:
  • Human dimension (AI skills, founder/managerial AI literacy)
  • Venture-level infrastructure (data, cloud/subscription AI tools, computing access)
  • Strategic dimension (alignment of AI with core value proposition)
  • Innovation dimension (embedding AI into products/processes)
• Distinguishes internal venture infrastructure from external digital infrastructure (the latter is a moderator): a venture can have internal AI resources but still be constrained by unreliable electricity, connectivity, or affordability.
• Three concrete gating conditions specified:
• Digital infrastructure (reliable electricity, connectivity, affordable computing)
• Founder AI literacy (prior related knowledge / absorptive capacity)
• Ecosystem support (finance, talent markets, institutional/regulatory support)
• Startups in emerging economies face systematic scarcity of these complements (liability of newness, acute resource constraints), making the gating effect especially salient.
• Literature review finds convergence on conceptualizing AI capability as a firm-level capability but divergence in measurement approaches (dimensions, reflective vs. formative modeling, order), which complicates cross-study comparisons.
• Empirical evidence is mixed: some studies show positive capability–performance links, others find partial or conditional effects — consistent with the proposed gated mediation thesis.
• The paper is conceptual/theoretical: it derives testable propositions and calls for empirical testing focused on startups in emerging economies (notably Tanzania / East Africa).

Data & Methods

• Method: Structured integrative literature review (interpretive synthesis aimed at framework building rather than meta-analysis).
• Search scope/time: Targeted searches across major scholarly databases and publisher platforms (ScienceDirect, Emerald, Springer, Wiley, Taylor & Francis, MDPI) plus Google Scholar and institutional sources; focused on 2019–2026 for AI empirical work but included older foundational theory.
• Search terms combined AI capability/adoption with competitive advantage, startup/SME performance, mediation/moderation methods, and emerging-economy country terms (e.g., Tanzania, Kenya). Searches conducted April–June 2026.
• Retrieval and screening (approximate counts): ~550–600 initial records → after duplicates and title/abstract screening ~230 full-text reviews → ~34 most directly relevant citations used to anchor the framework. (Authors note these are good-faith approximations, not a formal PRISMA log.)
• Evidence types: peer-reviewed empirical studies (largest share), foundational theoretical works (RBV, dynamic capabilities, complementary assets, absorptive capacity), and contextual/institutional policy sources documenting African/infrastructure conditions.
• Theoretical grounding: Resource-Based View (VRIO), Dynamic Capabilities (sensing–seizing–reconfiguring), Complementary Assets theory, Absorptive Capacity.
• Limitations of method: non–fully-reproducible single-protocol review, heterogeneity in measures across studies, and the paper is conceptual—propositions are not yet empirically tested within this manuscript.

Implications for AI Economics

• For theory:
  • Reframes RBV-style claims about technology as conditional — economists and strategy scholars should model AI as a resource whose returns are non-linear and contingent on complements and absorptive capacity.
  • Encourages explicit modeling of mediated and gated pathways (moderated mediation) rather than assuming direct effects from capability to performance.
  • Suggests that firm-level heterogeneity in returns to AI is structural (due to complement scarcity), not merely stochastic.
• For empirical work:
  • Test the proposed moderated-mediation model in startups and SMEs, especially in emerging economies. Key estimands: (a) mediation of capability → performance by competitive advantage, and (b) moderation of capability → competitive advantage by the three gating variables.
  • Measurement recommendations: use multidimensional AI-capability constructs (human, venture infrastructure, strategic, innovation), and separately measure external digital infrastructure and absorptive indicators (founder/manager literacy).
  • Use designs that permit heterogeneity analysis (interaction terms, subgroup estimations, instrumental variables if endogeneity concerns).
  • Collect context-rich data (electricity reliability, connectivity prices, venture access to cloud services, founder prior AI experience, local funding/talent availability).
• For policy and development economics:
  • Investments that enable AI returns should prioritize complements (beyond promoting AI adoption): digital infrastructure, skill-building targeted at founders/managers (absorptive capacity), and ecosystem formation (finance, training, regulatory clarity).
  • Cost–benefit analyses of AI-promoting interventions must include complementarities: subsidies for AI tools alone may yield limited returns without parallel investment in infrastructure and skills.
  • Recognize distributional implications: without complement investments, AI adoption may widen performance gaps between well-connected firms/founders and others.
• For managers and entrepreneurs in emerging economies:
  • Focus scarce resources on building complements that enable AI to matter strategically: improve founder literacy, align AI with core value proposition, secure stable infrastructure or architect around its absence (edge/cloud hybrid strategies, offline-capable models, partner with local data centers).
  • Leverage ecosystem partners (incubators, universities, cloud vendor programs) to bootstrap missing complements.
• For macro / modeling of AI-driven growth:
  • Aggregate growth models should incorporate complement constraints and heterogeneity in firm readiness; ignoring gates will overstate the diffusion-driven productivity gains in constrained settings.
  • Policy counterfactuals (e.g., infrastructure upgrades vs direct firm subsidies) should be evaluated for multiplier effects on AI returns.

(Concise note: the paper is conceptual and argues for empirical validation of its moderated-mediation propositions; its evidence base is an interpretive structured review rather than new primary data.)