Page 105 - ISC PROCEEDINGS 21.4

spillovers that raise TFP systematically. In absolute terms, labor productivity stood at USD
9,182 per worker in 2024, substantially below Thailand (approximately USD 17,000),
Malaysia (approximately USD 37,000), and South Korea (approximately USD 90,000)
(World Bank, 2024b). Although the 5.88% annual growth rate exceeds the National
Assembly's target and places Vietnam among the world's three fastest-growing
productivity nations alongside China and Ethiopia (WIPO, 2025), the absolute level gap
reflects persistent structural dependence on labor-intensive, low-value-added production
that AI has not yet fundamentally disrupted.
Second, AI deployment remains shallow and organizationally fragmented. While
88% of Vietnamese knowledge workers use generative AI tools, above the 75% global
average (Microsoft, 2024), only 27% of organizations are fully prepared for systematic AI
deployment (Cisco, 2024). Most implementations remain point solutions and isolated
pilots rather than integrated, organization-wide transformations capable of generating
TFP gains at the firm level. This precisely mirrors Brynjolfsson et al. 's (2018) productivity
paradox: impressive individual AI tools fail to aggregate into macroeconomic TFP gains
because complementary organizational changes, redesigned workflows, new business
models, restructured incentive systems, have not yet materialized. Sequoia's (2025)
characterization of the current global AI moment as high adoption, low transformation
applies with particular acuity to Vietnam.
Third, the R&D investment base for endogenous AI innovation remains
underdeveloped. Vietnam's gross R&D expenditure stood at only 0.42-0.43% of GDP in
2021 (World Bank, 2024a), far below the OECD average of 2-4% and the 2% threshold for
knowledge-economy transition (Vietnam Net, 2025). Without a domestic research base
capable of generating AI tools adapted to Vietnam's industrial, agricultural, and linguistic
contexts, Vietnam remains primarily an AI consumer rather than developer, a position
that limits both TFP depth and the long-run economic returns from AI investment.
4.3. Structural constraints: skills deficits, data fragmentation, and institutional
gaps
Three mutually reinforcing structural constraints explain why AI has not yet scaled
from experimentation to economy-wide TFP improvement, and must be addressed in
concert rather than sequentially.
Human capital and skills gaps represent the most acute bottleneck. Public First
(2024) found that over 70% of Vietnam's workforce, approximately 38 million workers,
lacked formal qualifications, fundamentally limiting capacity to deploy or benefit from AI-
integrated processes. Over 60% of businesses in 2023 reported inability to find
adequately skilled workers; only 24% feel their workforce is currently AI-ready despite the
Ministry of Information and Communications estimating AI literacy will be required in
50% of jobs (Vietnam News, 2025). The scarcity premium is clear: firms would raise salary
offers by 40% to attract demonstrably AI-skilled candidates. This skills deficit is a primary
barrier to AI-driven TFP gains, firms cannot redesign workflows or realize AI productivity
multipliers without a workforce capable of human-AI task complementarity. At the
specialized level, the semiconductor sector faces a critical engineer shortage against a
target of 50,000 qualified engineers by 2030.
Data infrastructure fragmentation constitutes the second major constraint. TFP
gains from AI depend critically on access to interoperable, standardized, high-quality
datasets that enable AI models to generate cross-sector knowledge spillovers. Yet
Vietnam's data remains siloed across ministries, enterprises, and administrative tiers with

104