Companies Are Realizing They're Spending a Fortune on AI Just So Their Employees Can ...

TEXT START: Loads of stories have emerged over the past few months of corporate executives realizing their all-out drive to use AI was getting them absolutely nowhere.

THE DISSECTION

The text converts an early deployment failure into a story about AI disillusionment. It bundles workslop, wasted token budgets, executive retrenchment, and catastrophic incidents into one conclusion: the AI revolution is failing. Its real maneuver is to reduce a structural labor threat to a management problem—bad rollout, bad governance, bad employee behavior, bad outputs.

THE CORE FALLACY

It confuses adoption friction with terminal capability limits. AI-generated workslop proves that companies deployed cheap generation without adequate evaluation, workflow redesign, accountability, or verification. It does not prove that systems with better models, proprietary data, tool access, and controlled workflows cannot outperform human labor.

Retrenchment is selection pressure, not reversal. Firms do not need AI to be universally competent. They need it to be reliably cheaper or faster in enough cognitive domains to justify substitution. Failed experiments are discarded; successful systems are integrated and scaled. The article treats the pruning of weak deployments as proof that the tree cannot grow.

HIDDEN ASSUMPTIONS

• “AI spending” is treated as one homogeneous investment rather than a mixture of experiments, theater, infrastructure, and genuine automation.
• Verification and integration costs are assumed to remain permanently high.
• AI incidents are treated as evidence against AI rather than evidence of immature governance and reckless deployment.
• Executive cutbacks are assumed to mean reversal rather than reallocating capital from theatrical pilots to controlled systems.
• Workslop is treated as uniquely an AI pathology rather than an accelerated, cheaper form of existing bureaucratic output.
• Because replacement is currently unreliable, it is assumed it cannot become economically dominant.
• Slowing adoption is mistaken for defeating automation. Under the Discontinuity Thesis, institutional lag delays the break; it does not cancel it.

SOCIAL FUNCTION

Partial truth serving as transition management and elite self-exoneration, with a layer of copium. The article gives executives a respectable explanation for wasted budgets and failed rollouts while making the labor threat appear less structural. It shifts blame toward employees “frittering away” resources and toward immature tools, allowing management to avoid admitting that it bought capabilities before building the systems needed to control them.

THE VERDICT

This is a report on the ugly first stage of automation: firms buy generative capability before they build reliable control, and workers use it to manufacture the appearance of productivity. That waste is real. It is not a refutation of the Discontinuity Thesis.

The article documents lag defenses—organizational incompetence, governance failure, and workflow friction—not a reversal of the underlying mechanism. Weak AI deployments will be cut. The systems that survive will be narrower, more integrated, more heavily verified, and more effective at replacing labor. The article mistakes smoke from failed experiments for evidence that the fire has gone out.