Build talent, import talent, or move the firm

Korean firms facing scarce AI capability can build workers, recruit internationally, place tasks remotely, or move activities closer to talent.
Each option has different costs, delays, knowledge-transfer effects, and implications for Korea’s industrial base.
National talent policy should recognize firm location and task location as alternatives to physical immigration.

The Korean Policy Question

The usual policy debate offers Korea two choices: train more domestic workers or attract more foreign ones. Firms face a wider menu. They can reorganize tasks across borders, acquire foreign companies, establish research units abroad, or relocate a larger part of the production system. The AI Trade Shock Will Reward Countries That Move Workers, Not Just Data frames AI-era adjustment around the movement of work as well as data.

The relevant comparison is not the wage of one worker in two countries. It is the total cost of producing and integrating capability, including training time, coordination, intellectual-property protection, management attention, and the risk that knowledge remains isolated from the rest of the firm.

A Firm-Level Choice

When a firm cannot find a required capability locally, public discussion often presents two options: train domestic workers or recruit migrants.

1. build capability internally;
2. hire a worker who already lives locally;
3. bring a worker to the firm's location;
4. employ or contract the worker remotely;
5. move a team or function toward the talent;
6. automate or redesign the task; or
7. abandon the activity.

These options differ in cost, speed, control, learning, and strategic value.

The Make-Buy-Move Framework

The framework distinguishes output, labor cost, coordination cost, institutional and regulatory cost, technological cost, and the fixed cost of establishing the strategy.

The lowest wage strategy need not maximize value. A remote specialist may have a lower wage but impose high coordination cost. Internal training may be expensive initially but create cumulative organizational knowledge.

Comparing the Options

Table 1. Strategic options for accessing scarce capability

Strategy	Main advantage	Main limitation	Best fit
Train internally	Builds firm-specific capability	Slow and uncertain	Recurring needs with teachable foundations
Hire locally	Low relocation and integration cost	Limited candidate pool	Common or adjacent capability
Recruit internationally	Places talent near the core team	Migration and retention friction	High-value, interdependent work
Contract remotely	Fast access and flexible scale	Control and knowledge-retention risk	Modular, observable tasks
Build a distributed team	Access to several labor markets	Requires strong documentation and management	Long-lived, separable functions
Move a function	Places work inside a stronger ecosystem	High fixed and governance cost	Capability tied to a location or cluster
Automate or redesign	Reduces labor requirement	May shift rather than remove the constraint	Standardizable tasks

The table is a starting point. The same role may move across categories as the firm learns.

Build: The Economics of Training

Training is attractive when the skill is cumulative, the firm can provide good feedback, and the capability will be used repeatedly. It is less attractive when the need is immediate, the prerequisite gap is large, or the firm cannot offer a developmental environment.

Training should not be used as a slogan for postponing a difficult hiring decision.

Import: Bringing Talent to the Organization

International recruitment can improve coordination for tasks requiring close interaction with the core team. It also transfers local institutional knowledge to the worker over time.

This strategy works when the organization offers a productive platform that the worker values. If the destination weakens the worker's research access, professional options, or household welfare, compensation must offset the loss and retention remains uncertain.

The firm cannot assume that a vacancy is an attraction.

Remote: Moving the Task Instead of the Person

Remote work allows the task to cross the border.

The framework distinguishes modularity, observability of output, documentation quality, and dependence on tacit knowledge.

Read as a diagnostic rather than a second scorecard, the framework becomes clear: Clear interface and acceptance test — Less remote-suitable: Ambiguous problem definition. Reproducible data and environment — Less remote-suitable: Access depends on informal local knowledge. Output can be reviewed asynchronously — Less remote-suitable: Constant real-time negotiation. Limited security or licensing friction — Less remote-suitable: Sensitive data and local regulation. Stable task ownership — Less remote-suitable: Rapidly changing authority.

Remote work fails when firms transfer ambiguity rather than a well-designed task.

Move: Locating the Firm Near Talent

Sometimes the knowledge base, professional network, customers, and suppliers are concentrated elsewhere. Moving a team can be more rational than trying to recreate the ecosystem.

Moving a function is most credible when the location supplies a capability the firm cannot efficiently reproduce.

Organizational Knowledge and the Boundary of the Firm

Contracting can provide immediate expertise, but the firm may fail to retain what it learns.

The sourcing decision should therefore include a knowledge-retention objective.

Labor Cost Is Not Unit Cost

Comparing salaries across locations is insufficient.

A lower wage can coincide with a higher unit cost when rework, delay, management attention, or quality failure is substantial. A higher-wage specialist can be cheaper when the person selects a simpler model, prevents an invalid deployment, or builds reusable infrastructure.

This is especially important in AI, where an error can be fluent, scalable, and difficult to detect.

A Portfolio Strategy

Firms rarely need one global policy for all work. They can allocate tasks across strategies.

• core model governance may remain close to decision makers;
• specialized research may connect to an external cluster;
• modular engineering may be globally distributed;
• local staff may be trained for recurring domain work; and
• routine tasks may be automated.

The boundaries should follow task economics, not symbolism.

The portfolio should also preserve option value. A pilot with a remote specialist can reveal the true task before the firm pays the fixed cost of a new office. A temporary internal rotation can test whether local staff possess the prerequisites for deeper training. A small co-located team can discover which interfaces must remain close to decision makers before the firm distributes the rest.

Implications for Education and Policy

Education should prepare students for a labor market in which firms can move tasks as well as people. Graduates compete and collaborate across locations.

• internationally legible technical communication;
• reproducible work;
• clear interfaces;
• independent problem solving;
• domain knowledge;
• remote collaboration; and
• the ability to learn inside unfamiliar institutional systems.

For policy, restricting labor mobility does not necessarily preserve local work. Firms may contract remotely, automate, or move functions. A durable local strategy must raise the productivity of the ecosystem, not rely only on barriers.

The four routes should be treated as a portfolio of real options. Internal training is slow but preserves organizational knowledge. Migration brings scarce capability into the existing team but creates integration and policy risk. Remote contracting offers speed and geographic reach but may weaken control over tacit knowledge. Relocation places the firm inside a deeper labor market but exposes it to high fixed costs and possible loss of existing relationships. Uncertainty makes a staged strategy valuable: a firm can begin with a remote team, use the project to learn which capability is truly scarce, then decide whether to train internally, sponsor migration, or establish a permanent location. The best choice is often a sequence rather than a single irreversible answer. Governance should assign an owner to the learning generated by each stage. If a remote pilot succeeds, the firm must record which expertise mattered, which interfaces failed, and whether the capability should remain external. Otherwise, the pilot reduces an immediate workload without improving the later location or hiring decision. Option value exists only when experimentation produces information that changes the next commitment. The staged approach also reduces false precision. Cost estimates made before the firm has observed the work are often dominated by assumptions about coordination, review, and tacit knowledge. A bounded pilot replaces part of that uncertainty with evidence. Management can then compare the options on observed unit cost, learning, retention of knowledge, and control rather than salary or tax rates alone. The strategic comparison should therefore value learning as well as the immediate delivery of scarce work.

Technological change makes the location decision more dynamic. The AI Trade Shock emphasizes worker reallocation across sectors and borders. Jevons Paradox and AI Adoption shows that firms compare total task economics rather than technical exposure alone, while The AI Data Center Jobs Debate demonstrates that the employment footprint of infrastructure extends beyond the facility itself. Building, importing, contracting remotely, and relocating are therefore portfolio choices whose costs depend on scale, demand, integration, and the surrounding production network.

The Industrial-Location Consequence

If Korea makes high-skill immigration difficult while global collaboration becomes easier, firms do not necessarily wait for domestic supply. They can move the task. That response preserves corporate competitiveness but may reduce local spillovers, supplier learning, and the development of a domestic professional community.

Conversely, importing workers without changing the organization can leave expertise underused. Building domestic talent is slower but can strengthen firm-specific knowledge. Remote work expands the search set but increases coordination demands. Relocation places the activity inside a thicker capability market while weakening the Korean cluster.

Policy should evaluate these as a portfolio of choices. The objective is not to keep every task physically inside Korea; it is to preserve enough high-value activity, knowledge transfer, and institutional learning that Korean participation in the production system remains cumulative.

From Mechanism to Evidence

For Build talent, import talent, or move the firm, the policy mechanism should be read as a chain rather than as a headline target. A government changes funding, rules, information, or institutional incentives; organizations respond; workers and students adjust; and only then do productivity or social outcomes change. Each link can weaken the intended effect. A credible evaluation therefore identifies the behavioral response at every stage instead of treating announced spending or participation as proof of success.

Conclusion

A skill shortage does not imply one remedy. Firms can build, buy, import, distribute, move, redesign, or exit.

The decision depends on more than wages. It depends on the time required to develop capability, the transferability of knowledge, task modularity, coordination cost, cluster advantage, retention, and the firm's ability to absorb what external specialists produce.

The strategic objective is a productive system with cumulative learning. Sometimes that means bringing talent to the firm. Sometimes it means taking the firm to talent. Increasingly, it means designing an organization that can do both.

References

Ronald H. Coase, “The Nature of the Firm”, Economica, 1937.
Frank Levy and Richard J. Murnane, The New Division of Labor, Princeton University Press, 2004.
Richard Baldwin, The Globotics Upheaval, Oxford University Press, 2019.
The Economy Editorial Board (2026) ‘The AI Trade Shock Will Reward Countries That Move Workers, Not Just Data’, The Economy Review, 2 June.
The Economy Editorial Board (2026) ‘Jevons Paradox and AI Adoption: Why Comparative Advantage Matters More Than Exposure’, The Economy Review, 1 September.
The Economy Editorial Board (2026) ‘The AI Data Center Jobs Debate Is Counting the Wrong Workers’, The Economy Review, 17 August.