Canaries in the Gold Mine: Early Productivity Gains from Artificial Intelligence Creating Organization Capital -- by Tania Babina, Alex X. He, Renhao Jiang

TEXT START: Using a new firm-level measure of AI investment based on AI-skilled employment—spanning machine learning through generative and agentic AI—we show that AI investments are associated with productivity growth in recent years, but not over the previous decade.

THE DISSECTION

The paper identifies a real transition mechanism: firms are converting AI-assisted work into organization capital—durable, firm-specific production knowledge. Its central signal is not permanent worker empowerment. It is that firms are learning how to store AI-enabled capabilities inside processes, systems, and institutional memory. The human jobs may be the construction crew for a capital layer that eventually reduces dependence on human labor.

THE CORE FALLACY

The abstract treats firm-level productivity growth as if it implied durable human economic necessity. It does not. Under the Discontinuity Thesis, AI can raise productivity precisely by transforming human learning and AI-assisted work into owned capital. Even if the relationship is causal, the finding demonstrates capital formation—not preserved employment, wage growth, broad consumption power, or survival of the wage circuit.

“AI-skilled jobs build organization capital” may describe the lag phase, not the end state. Once the knowledge is codified and reproducible, the marginal need for the people who created it can decline. The paper’s word “early” is therefore decisive: this is evidence of accumulation before displacement, not evidence against displacement.

HIDDEN ASSUMPTIONS

• AI-skilled employment is a clean measure of AI investment rather than a proxy contaminated by firm selection, managerial sophistication, or hiring patterns.
• Job descriptions accurately measure organization capital and the work that produces it.
• Firm-specific knowledge will remain embodied in workers rather than being progressively encoded in AI systems and agents.
• Recent productivity gains will persist and scale across firms and the wider economy.
• Productivity gains will be distributed to workers rather than captured by owners of the AI-enabled capital.
• The workers who build organization capital will remain necessary after that capital matures.
• Firm productivity can rise without weakening the mass employment-to-consumption circuit.

The supplied abstract provides no evidence on ownership, displacement, wage shares, concentration, or whether later production requires fewer workers. Those omissions are not peripheral. They are the variables that determine whether the post-WWII order survives.

SOCIAL FUNCTION

Partial truth functioning as transition management. The paper correctly detects early productivity gains and the time required for organizational learning. But its framing can anesthetize the structural consequence: firms may be building a capital stock that outlives the human jobs used to create it. It supplies institutions with a respectable narrative—AI creates knowledge and productivity—while leaving the ownership and productive-participation question outside the frame. That is not necessarily the authors’ intent; it is the discourse’s function.

THE VERDICT

This paper is compatible with the Discontinuity Thesis and may be an early indicator of it. The canary is not announcing worker liberation. It is detecting the first oxygen loss in the wage system. Recent AI-skilled employment is the scaffolding phase in which firms accumulate AI-enabled organization capital. If that capital becomes reproducible through increasingly capable AI and agents, the same productivity gains now celebrated can sever output from mass human employment. The abstract cannot prove P1–P3 or establish a collapse date, but it identifies the transition mechanism: AI first creates organization capital through labor, then threatens to make that labor unnecessary.