AI productivity gains are real but so is bad management - Fortune

TEXT START: Earlier this year, Apollo chief economist Torsten Slok made an observation about AI's "productivity paradox" as data emerged that employees could potentially save an entire workday a week by deploying AI, but economic data showed a steep drop in productivity.

THE DISSECTION

This article is a corporate sector autopsy dressed as management advice. It documents the early-stage chaos of AI deployment in enterprise environments: tokenmaxxing, cost overruns, abandoned mandates, and workers gaming metrics. The article frames everything through a leadership failure lens—bad communication, no vision, fear-driven secrecy, insufficient upskilling.

But beneath the operational diagnostics, this piece is functioning as transition management copium for executives. It tells them the problem is fixable. It implies that if they just communicate better, deploy access more thoughtfully, and build "sharing cultures," the productivity gains will materialize.

THE CORE FALLACY

The article's foundational error is treating the productivity paradox as a management execution problem.

It assumes the gains are real and waiting to be unlocked through better leadership. This is wrong in two directions:

First: The macro productivity paradox isn't about bad management. It's about what you're measuring. AI productivity gains are accruing to capital owners as margin expansion, not flowing through labor markets as measurable aggregate worker output. You're measuring the wrong variable. Workers saving eight hours a week doesn't show up as productivity when those workers are being moved toward termination, not redeployment. The gains are being captured by the owners, not distributed through the employment relationship.

Second: The fear workers exhibit isn't irrational. It's correct. BCG notes that when AI agents are treated as "digital employees," fear spikes. The article treats this as a behavioral problem to be solved with upskilling and communication. But the workers are reading the situation accurately: they are being systematically positioned for obsolescence. Their fear is a rational response to structural displacement, not an emotional malfunction requiring management intervention.

HIDDEN ASSUMPTIONS

The article smuggles in three assumptions that are each individually false under DT mechanics:

1. The goal is to make workers more productive. Wrong. The goal from capital's perspective is to eliminate the need for workers entirely. "Productivity" is the interim metric being used to justify deployment, but the actual objective is labor substitution.

2. Productivity gains flowing to corporations translate to system-wide prosperity. Wrong. When gains accrue as capital returns rather than wage distribution, you don't get consumption-driven growth. You get concentration.

3. Fear is a problem to be solved so workers share knowledge. Wrong. Fear is the correct signal that workers should hoard advantages and avoid cooperation. BCG's "sharing culture" recommendation is asking humans to act against their rational self-interest in a structurally hostile environment.

SOCIAL FUNCTION

This is transition management propaganda with consulting firm branding.

It's telling executives three things they want to hear:
- The AI transformation is still on track
- The problem is fixable with better leadership
- Your role remains central to the outcome

This serves the function of maintaining executive engagement with the transition while the actual mechanics work against their long-term position. The article gives them something to "do" about AI—better communication, thoughtful access policies, upskilling programs—that doesn't threaten their own role while the structural displacement accelerates.

BCG gets to sell the solutions (strategy articulation, organizational redesign, upskilling programs) to the same executives being told they've failed at the basics.

THE VERDICT

The productivity paradox exists because you're looking for labor productivity gains in a system where the gains are flowing to capital. The tokenmaxxing chaos exists because corporations deployed tools without strategic clarity because there is no strategic clarity—the goal keeps shifting between "make workers more productive" and "replace workers entirely," and leadership hasn't been told which one they actually serve.

The workers are not wrong to be afraid. The executives are not wrong to be confused. The chaos is not a management failure that better leadership can fix.

It's the system working exactly as designed. And the design is: extract value from AI deployment, reduce labor dependency, concentrate returns. Management can optimize the extraction, but they cannot change the direction.

The article's recommendations will be implemented. They will produce marginal improvements in adoption rates and cost control. They will not change the fundamental mechanics. The productivity gains will continue to accrue to capital owners while the workforce continues to be structurally decoupled from economic necessity.

This is the corporate sector's first look at the transition. It will not get cleaner from here.