Generative AI is remapping services trade: comparative advantage shifts from raw wage gaps to AI-adjusted task productivity determined by access to models, data, compute and organizational integration. Firms and countries will compete on AI-enabled task composition and complementary capabilities rather than on nominal labor costs alone.

Main Finding

Generative artificial intelligence (GenAI) is reshaping international trade in services by changing the task-level drivers of comparative advantage. Rather than eliminating comparative advantage, GenAI alters its basis: relative advantages increasingly depend on AI-adjusted task productivity and cross-border tradability, which are jointly determined by task composition, human-machine complementarity, access to models/compute, data resources, and organizational integration. The paper formalizes this as “AI-mediated comparative advantage” and shows that GenAI produces three coexisting effects—substitution, augmentation, and creation—that redraw the boundary of offshoring in a task-selective way.

Key Points

• Unit of analysis: Tasks (not industries or occupations). Service production is decomposed into identifiable tasks that differ in codifiability, routineness, contextual dependence, interaction intensity, and regulatory sensitivity.
• Three GenAI effects:
  • Substitution: Automation of standardized, codifiable tasks (e.g., basic translation, routine copywriting, entry-level coding, simple back-office processing), reducing the role of wage differentials in those tasks and shrinking some offshoring demand.
  • Augmentation: GenAI as a productivity tool for complex, judgment-intensive tasks (e.g., professional writing, customer service, high-end programming). AI increases effective output per worker and changes the wage–productivity tradeoffs.
  • Creation: Emergence of new tradable tasks (model customization, systems integration, data governance, model evaluation, safety testing) that create new areas of comparative advantage.
• Four channels by which GenAI affects tradability: lowers content-generation costs; increases task standardization; reduces language/communication frictions (e.g., machine translation); and improves cross-border quality-control efficiency.
• AI-mediated comparative advantage: defined as the relative advantage in cross-border-deliverable service tasks achieved by combining task composition, human capital, model & compute access, data, and organizational processes to deliver lower AI-adjusted unit task cost or higher quality-cost ratios.
• Heterogeneous outcomes: Substitution, augmentation, and creation can occur simultaneously within a single service process (e.g., software development). The net effect depends on task verifiability, liability, data sensitivity, client trust, and regulatory constraints.
• Role of human capital changes: Value of human skills depends on ability to complement AI (verification, judgment, client relations, domain expertise). GenAI can compress experience premiums on some tasks while raising demand for high-end, domain-specific talent.
• Infrastructure broadened: Beyond broadband and payments, competitiveness now requires model access, cloud compute, data availability & governance, and organizational capabilities to integrate AI into processes.

Data & Methods

• Approach: Conceptual/theoretical synthesis and framework-building. The paper integrates:
  • Comparative advantage theory and trade-in-tasks models (Grossman & Rossi-Hansberg; Baldwin & Robert-Nicoud).
  • Automation and task-based labor economics (Autor, Levy & Murnane; Acemoglu & Restrepo).
  • Recent empirical and experimental results on GenAI’s productivity effects (Felten et al.; Noy & Zhang; Brynjolfsson, Li & Raymond; Hui, Reshef & Zhou; Brynjolfsson, Hui & Liu).
• Methods used: Literature review, conceptual mapping of task characteristics to GenAI capabilities, and analytical argumentation to derive implications for tradability and comparative advantage. No primary empirical dataset or formal econometric estimation is presented.
• Evidence base: Cites empirical findings that show heterogeneous productivity and labor-market impacts from GenAI (both augmentation and displacement), and previous trade-in-tasks results on how tradability shapes international specialization.
• Limitations: Framework is largely theoretical and illustrative—empirical quantification of AI-mediated comparative advantage across countries and tasks is left for future work. The paper emphasizes mechanisms rather than presenting new causal estimates or cross-country task-level trade data.

Implications for AI Economics

• Rethinking comparative advantage: Research and policy should shift from wage-centric views to measures of AI-adjusted task productivity and tradability. Comparative advantage will be a multidimensional, capability-based concept (models/compute/data/organization + task mix).
• Measurement challenges and research priorities:
  • Need for task-level trade and productivity data to track which tasks are automated, augmented, or newly created.
  • Develop granular AI-exposure metrics that combine technical feasibility with access to models/compute/data and organizational adoption.
  • Empirical work to estimate how AI changes cross-border costs, verification costs, and the elasticity of offshoring demand by task.
• Labor-market impacts: Expect heterogeneous effects—declines in demand and wages for some standardized offshorable tasks, augmented productivity (and possible reshoring) in complex tasks, and rising demand for AI-integration, governance, and domain experts. Policies should focus on reskilling, certification, and incentives for complementary investment.
• Impacts on development and global inequality:
  • Low-wage countries' traditional offshoring advantages may erode for highly automatable tasks unless they scale AI adoption, data access, and organizational capabilities.
  • New opportunities exist for countries that can build AI-integration services, model customization, and domain-specific data assets.
• Firm strategy and competition:
  • Firms should invest in model access, compute resources, data governance, and organizational change (process redesign, quality control) to translate technical exposure into productivity.
  • Competition will increasingly be over AI-adjusted productivity and the ability to bundle domain knowledge, client relationships, and AI tools.
• Trade and regulatory policy:
  • Policies addressing data flows, model access, liability, certification, and cross-border governance will affect which tasks remain tradable.
  • Trade statistics and agreements may need updating to capture AI-enabled services and newly tradable activities (model services, AI integration).
• Broader economic dynamics:
  • GenAI can both concentrate and diffuse capabilities: general models lower entry barriers, but effective service delivery requires complementary assets that may be unevenly distributed, potentially reshaping specialization patterns across countries.
  • Overall services trade may not uniformly contract or expand; it will reallocate across task types—some cross-border human inputs decline, while complex and AI-related cross-border services may grow.

Suggested next empirical steps (based on the paper’s framework): - Build cross-country, task-level indicators of AI-adjusted productivity (combining exposure, model/compute access, data availability, and organizational adoption). - Estimate the impact of GenAI adoption on offshoring flows for specific service tasks (translation, customer service, coding, professional writing). - Map emergence and trade growth of AI-specific services (model integration, data labeling, governance) and their comparative advantage determinants.

(Adapted from: Yu, L. & Yang, S. (2026). Reconstructing Comparative Advantage in International Trade in Services in the Era of Generative Artificial Intelligence: A Trade-in-Tasks Perspective. Economics & Business Management, 6(3), 119–131.)