What the 1920s Really Teach Business Leaders About AI

When George Anders published his Wall Street Journal piece drawing parallels between the 1920s technological boom and today’s AI revolution, he did something genuinely useful: he reminded anxious executives and policymakers that the sensation of unprecedented disruption has antecedents. The automobile, radio, electrification, and commercial aviation all arrived within the same decade, rewriting social norms, displacing entire occupational categories, and generating equal measures of utopian excitement and existential dread. Society absorbed those shocks and emerged more productive for it.

The impulse to find comfort in historical precedent is understandable, and in this case, not entirely misplaced. But the risk of the analogy is that it becomes a sedative when what business leaders actually need is a stimulant. The 1920s offer a framework, not a forecast. Understanding where the comparison holds, where it breaks down, and what it demands of today’s leaders is more valuable than the reassurance alone.

The Analogy’s Genuine Strength

Anders is on firm ground when he argues that the shock of living through transformational technological change is not a new sensation. The sociologists Helen and Robert Lynd, conducting their famous “Middletown” study in Muncie, Indiana, documented a community experiencing something that contemporary observers might easily mistake for a description of social media’s impact on teenagers: eroding traditional authority structures, reshaping courtship, accelerating consumerism, and generating widespread anxiety about what values the next generation would carry forward.

The parallel to AI’s current cultural impact is striking. Today’s debates about ChatGPT undermining student learning, large language models distorting truth, and autonomous systems replacing professional judgment echo, with uncanny precision, the early 1920s hand-wringing about automobiles enabling unsupervised courtship, radio delivering culturally inferior content, and assembly lines reducing skilled craftsmen to interchangeable parts.

What history confirms is that the anxiety is usually more acute than the eventual reality justifies. H.G. Wells dismissing radio as entertainment fit only “for the sick, the lonely and the suffering” looks embarrassingly wrong from a century’s distance. Historians of technology have long observed this pattern: new technologies are initially evaluated against the standards of the institutions they displace rather than the possibilities they create. The printing press was condemned for enabling heresy; the telephone for enabling improper conversations; television for rotting minds. Each eventually reshaped civilization in ways that required entirely new frameworks for evaluation.

The job-creation dimension of Anders’ argument also deserves serious credit. The automobile industry’s spillover employment effects were dramatic and empirically well-documented. By 1930, hundreds of thousands of Americans were employed in rubber manufacturing, road construction, and related trades that had barely existed a decade before. Economists use the term “general purpose technology” to capture this dynamic: technologies like electricity, the internal combustion engine, and now AI don’t merely replace existing functions but enable entirely new categories of economic activity. MIT economist Erik Brynjolfsson, who has spent decades studying technology and productivity, has noted that general purpose technologies typically create more jobs than they destroy, though the timing of that creation lags the destruction by years or even decades (Brynjolfsson and McAfee, The Second Machine Age, 2014).

Where the Analogy Strains Under Scrutiny

The most important limitation of the 1920s comparison is one that Anders acknowledges briefly but does not fully examine: pace. The electrification of American homes went from 35% penetration to 68% over an entire decade. ChatGPT reached 100 million users in two months. GitHub Copilot, Microsoft’s AI coding assistant, was adopted by more than one million developers within its first year of general availability. The speed differential between 1920s technology adoption and AI adoption is not incremental; it is categorical.

This matters enormously for both workers and organizations. The transition from horse-drawn transportation to automobile culture took long enough that an entire generation could observe the shift, recalibrate their skills, and make deliberate choices about where to invest their professional development. The adjustment was painful and uneven, but it operated on a human timescale. AI is operating on a computational timescale.

Research from the McKinsey Global Institute published in 2023 estimated that generative AI could automate tasks accounting for 60 to 70 percent of employees’ time, compared to less than 50 percent for previous automation technologies. More critically, this automation capacity is expanding rapidly, not gradually. For business leaders managing workforce strategy, the 1920s precedent suggests that adaptation is possible; it does not suggest that adaptation will be automatic at the speed AI is advancing.

There is also a qualitative difference that Anders’ analogy underweights. The 1920s technologies were predominantly physical: they moved bodies, transmitted sound waves, lit rooms, and powered machines. Their displacement of human labor was largely confined to physical and routine cognitive tasks. AI operates in the cognitive domain itself, including creative work, professional judgment, and complex analytical reasoning that previous generations of automation never reached.

MIT economist Daron Acemoglu has argued in peer-reviewed research that AI may be less of a job creator and more of a task automator than historical precedents suggest, precisely because it targets the high-skill cognitive functions that typically generate wage premiums (Acemoglu and Restrepo, “Tasks, Automation, and the Rise in US Wage Inequality,” Econometrica, 2022). Unlike the automobile, which created jobs partly because humans were still needed for cognitive work such as driving and navigation, AI targets the cognitive work directly. The comparison flatters AI’s job-creation potential.

Furthermore, the 1920s did not end well. Anders’ piece reasonably focuses on the decade’s adaptation mechanisms rather than its denouement. But business leaders should note that the Roaring Twenties ended in the Great Depression, a catastrophe partly enabled by the financial speculation and wealth concentration that accompanied rapid technological change. The benefits of the 1920s boom were distributed unevenly; wage growth lagged productivity growth significantly. Selective historical memory can be a dangerous guide.

The Regulatory Lesson: The 1920s’ Most Transferable Insight

Where Anders’ argument is strongest, and most urgently relevant for today’s leaders, is in its examination of how the 1920s handled safety and regulatory challenges. Car-related death rates in 1920 were, on a per-mile basis, 20 times higher than today. The response was neither to abandon the technology nor to leave it entirely to market forces. It combined industry self-regulation, municipal experimentation (traffic lights in Los Angeles and Detroit), state-level requirements (New Jersey’s driver’s licenses in 1924), and federal legislation (the Air Commerce Act of 1926) into a multi-layered safety architecture that dramatically improved outcomes without stifling adoption.

This model—layered, adaptive, and built through collaboration between industry pioneers and public authorities—is precisely what AI governance currently lacks and most urgently needs. The contrast with today is sobering. The 1920s saw meaningful federal aviation legislation within six years of commercial air travel becoming viable. We are now more than five years into the era of large-scale AI deployment, and coherent federal AI legislation in the United States remains elusive. The European Union has moved further with its AI Act, which creates risk-tiered requirements for AI systems, but even that framework carries significant compliance ambiguity as implementation proceeds.

Business leaders should not wait for governments to set the terms of AI safety. The automobile industry’s history shows that companies that took safety seriously before regulation required it gained customer trust and competitive advantage. Today’s AI leaders who proactively develop auditable, explainable, and ethically governed AI systems are building organizational capabilities that will matter more, not less, as regulatory frameworks inevitably mature.

What Business Leaders Must Do Now

The most productive reading of Anders’ historical comparison for practicing executives is not “don’t worry, things worked out before.” It is: “here are the specific mechanisms through which societies navigate technological disruption; now build those mechanisms deliberately rather than hoping they emerge organically.”

First, invest in adjacent workforce capabilities. The 1920s automobile economy’s spillover employment didn’t happen by accident; it happened because entrepreneurs, educators, and municipal governments recognized and responded to new demand signals. Today’s AI wave will similarly create massive demand for roles that barely exist today: AI auditors, algorithmic ethicists, human-AI collaboration specialists, data governance officers, and AI systems trainers. Organizations that begin building these talent pipelines now, rather than waiting for displacement to become acute, will be better positioned to retain institutional knowledge and maintain their social license.

Second, treat AI safety as strategy, not compliance. The executives at Radio Corporation of America who invested in broadcast quality and programming standards were not merely responding to regulation; they were building the platform credibility that gave RCA competitive dominance for decades. Similarly, companies that invest seriously in AI safety, explainability, and bias reduction today are building the trust architecture that will be enormously valuable as AI systems take on higher-stakes functions in healthcare, finance, legal reasoning, and education.

Third, resist historical fatalism. Anders quotes Henry Ford: “It is easier to go along with progress than to try to hold things back.” Ford was right about the direction of change; he was catastrophically wrong about his own company’s management of it. Ford’s resistance to labor organizing and his eventual failure to adapt to consumer preferences nearly destroyed the company he built. Going along with progress is not a substitute for managing it strategically.

Research by Raffaella Sadun and colleagues at Harvard Business School has consistently found that firms led by executives with strong management capabilities significantly outperform peers during technological transitions, precisely because good management mediates the relationship between technology and organizational outcomes (Bloom, Sadun, and Van Reenen, “Americans Do IT Better,” American Economic Review, 2012). Technology doesn’t transform organizations automatically; leaders do.

The Distribution Question Business Leaders Cannot Ignore

One dimension of the 1920s that Anders’ piece does not develop adequately is the role of organized labor in shaping how technological gains were distributed. The 1920s were a decade of significant labor conflict alongside technological dynamism. That conflict, though often difficult, ultimately produced the wage gains and working-condition improvements that allowed the emerging middle class to purchase the goods that mass production was creating. Henry Ford’s famous five-dollar workday was not philanthropy; it was a recognition that workers needed to be consumers.

The current era is characterized by organized labor’s relative weakness compared to a century ago, and this is precisely why job-creation optimism deserves scrutiny. If AI productivity gains flow primarily to capital owners and top management, the social-uplift dynamic that Anders credits with easing 1920s adoption anxieties may not materialize. Business leaders who want the AI transition to proceed with minimal social disruption have a material interest in ensuring that its benefits are distributed broadly enough to maintain the social trust on which their enterprises depend. This is not merely an ethical consideration; it is a strategic one.

History as Compass, Not Map

George Anders has written a valuable piece of historical journalism, and its core message is well-founded: technological disruption has precedents, adaptation is possible, and the combination of market forces and regulatory action has historically produced better outcomes than either alone. These are worth internalizing.

But history is a compass, not a map. It indicates the general direction of travel; it cannot tell today’s leaders the specific terrain they will cross, the obstacles they will encounter, or the pace at which they must move. The 1920s took a decade to unfold and then unraveled into depression. AI is unfolding in years and will reshape the productive capacity of the global economy for generations.

Business leaders who take the historical lessons seriously will do more than find comfort in them. They will invest in workforce transition before displacement forces the issue. They will build regulatory relationships rather than treating AI governance as adversarial. They will distribute the gains of AI-driven productivity broadly enough to maintain the social license that allows continued technological advancement. And they will lead actively through the transition rather than assuming that because the automobile age worked out, the AI age will work out on its own.

The 1920s are instructive precisely because they required leadership. The decade produced both Henry Ford’s assembly-line revolution and the Great Crash, both the transformative power of radio and the dangerous concentration of speculative capital. What determined outcomes was not the technology itself but the decisions leaders made about how to deploy it, regulate it, and share its benefits. A century later, those decisions still belong to the people in the room.