AI promises productivity and new business models for Vietnamese firms, but benefits are concentrated among well-resourced organizations; fragmented data, scarce skilled workers and unclear regulation mean SMEs risk being left behind.

Main Finding

The paper proposes a four-layer enterprise AI capability framework (data, algorithms, infrastructure, applications) integrated with the Resource‑Based View (RBV) to explain how AI creates firm value in emerging economies. AI generates value through three mechanisms—cognitive automation, decision intelligence, and business‑model innovation—but realising that value in Vietnam is constrained by fragmented data governance, talent shortages, limited SME investment capacity, weak digital leadership, and evolving regulatory frameworks. The study is a conceptual synthesis based on secondary sources, offering a diagnostic and policy-oriented roadmap rather than empirical validation.

Key Points

• Theoretical framing

  • Uses RBV and dynamic‑capability logic: proprietary data, algorithms, and organizational capabilities can be VRIN resources when integrated via AI.
  • Treats AI capability as a multi‑layer capability rather than a standalone technology.

• Four‑layer AI framework

  • Data layer: collection, integration, governance—foundational for downstream value.
  • Algorithm layer: ML/DL models turning data into predictive/prescriptive insights.
  • Infrastructure layer: cloud, storage, compute for scalability and performance.
  • Application layer: deployment that generates measurable business value (automation, decision support, customer engagement).

• Value creation mechanisms

  • Cognitive automation: replaces routine cognitive tasks, cuts cost, boosts efficiency.
  • Decision intelligence: improves decision speed/quality via predictive analytics.
  • Business‑model innovation: creates new data‑driven products, services, platforms.

• Vietnam / emerging‑economy context

  • Concentrated adoption among large firms, banks, and tech companies; SMEs lag.
  • Macroeconomic projections suggest large potential (paper cites ~USD 79.3bn by 2030 for Vietnam), but firm‑level capture is uneven.
  • Progress uneven across layers: application and infrastructure advancing faster than data governance and human capital.

• Structural barriers identified

  • Fragmented data systems and weak governance (interoperability, legacy silos).
  • Shortage of skilled AI professionals and managerial digital leadership.
  • High initial costs and unclear ROI measurement for AI projects, especially for SMEs.
  • Incomplete regulatory frameworks around AI transparency, ethics, and governance.

Data & Methods

• Research design: qualitative conceptual synthesis (non‑empirical).
• Sources: secondary data collected late 2025–early 2026, including global consulting reports (McKinsey, PwC, Deloitte), Vietnamese policy documents (e.g., Decision No.127/QĐ‑TTg), Google Asia Pacific materials, and industry publications.
• Analytical approach: comparative analysis (global best practice vs Vietnam) and thematic analysis to identify patterns of AI value creation and structural constraints.
• Theoretical integration: RBV + four‑layer AI stack to map technological components to organizational resources and value mechanisms.

Implications for AI Economics

• Modelling and measurement

  • AI should be modelled as a layered, complementary bundle of assets (data, algorithms, infrastructure, applications) and organizational capabilities—returns depend on complementarities and bottlenecks across layers.
  • Standard productivity and ROI estimates risk over‑statement if they ignore institutional and organizational frictions (data governance, skill gaps, SME finance constraints).
  • Empirical work should quantify both tangible (cost savings, revenue) and intangible (decision quality, innovation capacity) returns and develop metrics for cross‑layer complementarities.

• Policy and market design

  • Data governance and interoperability are public‑good enablers: investments and regulation here can have high social returns by unlocking private AI value creation.
  • Human capital and managerial training matter as much as hardware: subsidies, public‑private training, and university‑industry collaboration should be central to AI industrial policy.
  • Regulatory sandboxes and clearer AI governance can reduce uncertainty and accelerate adoption while managing externalities (privacy, fairness, transparency).

• Distributional and structural effects

  • Heterogeneous adoption (large firms vs SMEs) implies potential widening of firm‑level productivity gaps and market concentration; competition and SME finance policies matter for inclusive diffusion.
  • Path dependence: early investments in data architecture and governance can create persistent advantages (VRIN resources), affecting firm dynamics and long‑run market structure.

• Research agenda for AI economics

  • Empirically validate the four‑layer framework: firm‑level panel studies, difference‑in‑differences around policy changes (e.g., data‑sharing laws), and RCTs for managerial interventions.
  • Identify causal channels: how much of AI value is due to data quality vs algorithmic sophistication vs organizational adoption?
  • Cross‑country comparisons: assess how institutional variation (data law, labor markets, finance) shapes the speed and gains from AI diffusion.
  • Develop better ROI measurement tools that capture long‑horizon and intangible value components of AI investments.

Takeaway: The paper frames AI as a layered, capability‑dependent production factor whose economic impact in emerging markets hinges on complementary investments in data governance, human capital, finance, and institutions. For AI economics, this implies moving beyond technology‑centric forecasts to models and empirical work that account for cross‑layer complementarities, institutional constraints, and heterogeneous firm responses.