Burning the Bridge. Imported Intelligence, Reversibility Loss, and External Liquidity

Foreign-provided agentification can be understood as black-box offshoring: a domestic task is delegated to a technological system supplied from abroad rather than to a foreign workforce. As adoption spreads, firms may stop performing the task locally and gradually thin the shared pool of workers, training routes, routines, and complementary suppliers needed to bring it back. No individual buyer owns that pool or has much reason to preserve it for the sector. Under concentrated foreign provision, the weaker domestic fallback can raise future imported-agent payments. This is a dynamic terms-of-trade externality whose source is endogenous: today’s imports make tomorrow’s outside option worse. The broader macroeconomic implication is that agentification can raise productivity while tightening the external constraint. It replaces part of a domestic production bill with recurring foreign-currency payments, and capability erosion can make that bill more persistent or more expensive over time. This liquidity channel is not unique to AI, and it can reverse sign when agentification generates enough tradable earnings. Ex ante competition can redistribute lifecycle rents through introductory concessions, but early inflows and later foreign-currency outflows need not be socially equivalent when external liquidity is valued differently across states or dates. The policy implication is to preserve valuable reversibility, contestability, contractual protection, and external-liquidity buffers rather than to impose a general tax on AI.