In 1869, Massachusetts reformers persuaded the state to try a modest experiment: counting. The Second Industrial Revolution was tearing through New England’s mills and factories, and the gains in efficiency were being extracted from human bodies that were never designed to keep pace with machines. Owners understood the limits—both biological and political. There is only so much misery people will tolerate before unrest follows.
Massachusetts responded by creating the nation’s first Bureau of Statistics of Labor. The premise was simple: if injustice could not be prevented by conscience, perhaps it could be constrained by measurement. By tracking wages, hours, and injuries—what we now call externalities—policy makers hoped to make exploitation legible, and therefore governable. A few years later, amid violent labor unrest, Congress scaled the idea nationally, creating the Bureau of Labor Statistics.
Counting does not eliminate injustice. It rarely settles arguments. But it signals a commitment to shared facts—and over time, that commitment matters. The BLS remains one of the quiet achievements of American democracy, documenting a workforce that has grown richer, more complex, and remarkably adaptive. It is how we know that food trucks, pet grooming, and massage therapy have exploded into six-figure employment categories. It is the scoreboard of a country that has avoided violent class conflict for a century.
But statistics have limits. The BLS excels at telling us what has already happened. It is far less useful at telling us what comes next—especially when the future arrives faster than the past can explain it.
Enter artificial intelligence.
After early warnings framed like cosmic horror—“summoning the demon”—the AI industry shifted to corporate anesthesia: reimagining workflows, driving efficiency, unlocking value. Yet beneath the fleece vests is a technology capable of absorbing, analyzing, and producing skilled cognitive labor at a scale and speed no prior tool has matched. AI already drafts legal briefs, writes software, produces music, and replaces tasks that once justified entire career ladders.
Economists, drawing on history, argue this will be fine. ATMs didn’t eliminate bank tellers; Excel didn’t kill accountants. Productivity rose, new roles emerged, wages followed. The BLS projects continued job growth. But history also teaches that speed matters. Gradual change can be absorbed. Rapid reorganization cannot.
And here, the warnings are not coming from critics—they’re coming from CEOs. Leaders at Anthropic, Ford, OpenAI, and elsewhere have openly predicted that AI could eliminate 10–50 percent of white-collar jobs within a decade. Then, just as abruptly, they stopped talking. The silence is telling.
So far, the data remains ambiguous. Federal Reserve officials see strong productivity growth and low unemployment—signals that contradict mass displacement. But economists acknowledge that labor hoarding, delayed layoffs, or internal restructuring may be masking early damage. By the time the numbers are unmistakable, options may be gone.
The deeper risk is not just job loss, but institutional lag. AI spreads faster than electricity ever did. It doesn’t require rebuilding factories—only plugging into systems that already exist. Firms that delay adoption risk being undercut by competitors who don’t. Once one company automates, the rest must follow.
Policy, meanwhile, is stalled. Programs that once helped workers adapt to trade shocks have expired. Congress avoids the issue. States scramble independently. The AI industry spends heavily to ensure regulation doesn’t slow deployment. Everyone is betting that the transition will be survivable.
That bet may be wrong.
AI is still young. It may ultimately make societies healthier, wealthier, and more productive. But it also tests the foundations of a wage-based economy and a political system already stretched thin. Democracies adapt slowly. Markets do not.
The United States created the Bureau of Labor Statistics because it believed the first duty of a republic was to know what was happening to its people. Today, the BLS is underfunded, understaffed, and trying to measure a transformation unfolding faster than its tools were designed to capture.
If we cannot even commit to counting clearly—expanding surveys, tracking AI use, understanding where work is disappearing—then we are choosing blindness.
And if a democracy cannot bring itself to measure reality, it will not be prepared to govern it.
Good luck with the machines.