AI and Employment: So Far, So Good - Marginal REVOLUTION

TEXT START: In September 2023, the Census Bureau added questions about AI to its Business Trends and Outlook Survey.

THE DISSECTION

The post performs a narrow empirical maneuver: it substitutes short-term firm headcount for system-level labor demand. Its headline emphasizes that 95.7% of firms report no employment change, while its own data show expanding adoption and rising task substitution. The evidence describes an early diffusion phase, not economic safety. The author’s use of AI tools is rhetorically suggestive but economically irrelevant.

THE CORE FALLACY

The central error is treating “no change in total employment” as “no material displacement.” Firms can automate tasks without immediate layoffs through attrition, hiring suppression, reassignment, reduced hours, wage pressure, or increased output. A six-month, self-reported headcount measure cannot capture cumulative displacement or economy-wide competitive effects.

The data do not establish Cognitive Automation Dominance or Productive Participation Collapse. They also do not refute them. Rising substitution intensity among adopters is fully compatible with a lagged transition toward those outcomes.

HIDDEN ASSUMPTIONS

• Headcount is an adequate proxy for labor’s economic necessity.
• Firms accurately identify AI’s employment effects.
• A six-month window captures structural change.
• Current, mostly assistive use resembles the endpoint of adoption.
• Firm-weighted and employment-weighted averages reveal distributional effects.
• Unchanged employment means unchanged bargaining power, wages, workload, and future hiring.
• Productivity gains will not later produce competitive labor substitution.

SOCIAL FUNCTION

Partial truth functioning as ideological anesthetic and transition management. The snapshot is real: aggregate headcount collapse has not yet appeared in these firms. But the framing converts “not yet” into “so far, so good,” granting institutions permission to treat a lag indicator as a verdict on the future.

THE VERDICT

This is not a rebuttal to the Discontinuity Thesis. It is a lag report from the waiting room. The post shows delayed visible layoffs while AI adoption expands and substitution deepens among users. “So far, so good” is accurate only at the surface level and only so far. The article mistakes the absence of a corpse for proof that the system is not being poisoned.