Position: Adopting AI in Practice Does Not Guarantee the Productivity Boost

TEXT ANALYSIS PROTOCOL

THE DISSECTION

This is a position paper whose surface argument—that AI adoption doesn't automatically yield productivity gains due to human and organizational moderating factors—is factually accurate but structurally irrelevant to the actual question of whether post-WWII capitalism survives AI. The paper performs a localized friction analysis while ignoring the system-level math.

The authors identify five moderating factors:
1. Human resource composition
2. Baseline capability of individuals
3. Learning curve of practitioners
4. Incentives for fair use
5. Flexibility of objectives

These are real. They matter at the firm level. They are friction, not resistance. The paper confuses micro-level implementation lag with macro-level structural viability.

THE CORE FALLACY

The paper assumes the relevant question is "Will this firm get the productivity gains from AI?" when the actual question under DT logic is "Does the distribution of those gains preserve the employment-wage-consumption circuit?"

The five moderating factors are lag-level phenomena. They describe the speed and distribution of AI value capture within firms. They do not address whether the value is captured by:
- The sovereign entity (the firm/owner) extracting efficiency gains while displacing labor
- The workers whose skills are moderated into obsolescence
- The consumer class whose purchasing power depends on wages that are being eliminated

The Gries and Naudé (2022) framework the paper invokes is a partial equilibrium model. Partial equilibrium models cannot resolve questions that require general equilibrium analysis—including the collapse of labor as a factor of production.

HIDDEN ASSUMPTIONS

1. Implicitly assumes the productivity gains exist somewhere and that the problem is merely distributing them fairly. This is the assumption of a functioning economic system that can absorb and redistribute AI-generated surplus. The paper never interrogates whether that assumption holds.

2. Treats "productivity" as a unitary concept. In DT terms, there is a critical difference between firm productivity (output per worker) and systemic productivity (whether the output can be purchased by anyone other than capital). AI can raise firm productivity to extraordinary levels while simultaneously destroying mass purchasing power. The paper never distinguishes these.

3. Frames "human factors" as moderating variables rather than the primary variable. Under DT, the fact that human factors moderate AI's impact is temporary. As AI systems become more robust, less dependent on human calibration, and cheaper to deploy, the moderating factors erode. The paper treats them as durable features of organizational life.

4. Assumes the goal is productivity improvement within existing organizational forms. The paper does not ask whether the organizational form survives the transition.

SOCIAL FUNCTION

Partial truth dressed as comprehensive analysis. This paper is intellectually honest within its chosen frame but that frame is deliberately narrow. It serves the function of legitimizing continued AI investment by reassuring practitioners that the technology isn't magic—it requires human effort to implement. This is soothing copium for middle management who need to justify adoption budgets. It does not grapple with the structural displacement thesis at all.

The call-to-action for "stakeholders" is the tell: this is transition management propaganda. It implies that if we just get the human factors right, adjust incentives properly, manage the learning curves, we can capture the gains. This is the same failed logic that said "re-skilling" would solve deindustrialization.

THE VERDICT

The paper is a friction analysis of a process that will become frictionless. It identifies lag effects and presents them as durable moderating factors. The five factors it identifies will weaken over time as AI systems require less human intermediation, less fine-tuning, and less organizational context to operate. The learning curve flattens. The baseline capability requirement drops. The flexibility of objectives increases as AI generalizes.

What the paper describes is the transitional window—and within that window, yes, human factors matter. But this window is closing. The paper mistakes the decelerator for the brake.

Structural judgment: The paper is technically competent friction cataloging. It is systemically irrelevant. It is not wrong, but being not wrong about micro-dynamics while being silent on macro-trajectory is a form of epistemic failure with real consequences—it gives decision-makers false confidence that the moderating factors are structural features rather than transitional friction.

Survival relevance: Near-zero for sovereign entities (who don't need the gains to be fairly distributed, only to capture them). Moderate for servitors in the short term (human intermediation matters while AI remains brittle). Descending to zero for both within the timeline the paper itself implicitly accepts by citing 2022 models and 2026 submissions.