Here's the math that should concern anyone with a white-collar job: AI companies have accumulated massive debt that traditional revenue streams—ChatGPT subscriptions, advertising, enterprise solutions—simply cannot repay. Tristan Harris articulated the inevitable conclusion: only building artificial general intelligence capable of replacing human workers justifies these expenditures.
This isn't speculation or fear-mongering. It's basic corporate finance. When the numbers don't work through customer acquisition, companies pivot to cost reduction. And in knowledge work, the largest cost is labor.
Analyst Kevin Driscoll frames this as the "intelligence curse"—a structural shift where economic output continues growing while requiring fewer people to sustain it. This represents something fundamentally different from historical automation cycles. We're not talking about marginal productivity gains. We're discussing whether entire professional classes remain economically necessary.
Driscoll emphasizes this isn't conspiratorial—it's where market incentives naturally lead. Companies aren't primarily building tools to augment human work. They're developing systems that reduce labor requirements. The business case is straightforward: margins improve dramatically when fewer people generate equivalent revenue.
Finance, insurance, law, media, and consulting face particular exposure. These aren't low-skill, low-wage positions vulnerable to historical automation. These are high-skill, high-wage roles that support surrounding economic ecosystems—housing markets, schools, local businesses. Their compression creates ripple effects that economic models barely capture.
The data already shows early signals of this decoupling. Corporate revenues grow while headcount stagnates or shrinks. Productivity metrics improve while employment participation rates face pressure. Top-line economic health masks ground-level deterioration.
Policy responses remain inadequate because they address symptoms rather than structure. Regulations target deepfakes, privacy violations, and AI guardrails—important issues, but tangential to the core question: what happens to an economy when human labor becomes economically peripheral to knowledge work?
The counterargument holds that AI will create new categories of work, just as previous technological revolutions did. Perhaps. But the timeline matters enormously. Agricultural mechanization unfolded over generations. Software automation took decades. AI-driven labor replacement could compress into years, leaving insufficient time for economic adaptation.
Moreover, the new jobs created by previous revolutions generally required less skill and paid less than the jobs they replaced. Factory workers earned less than artisans. Data entry paid less than bookkeeping. If AI follows this pattern with knowledge work, we're discussing substantial wealth destruction across the professional class.
The investment levels tell you everything about industry expectations. Companies don't deploy capital at this scale for incremental improvements. They're betting on fundamental transformation. And when revenue can't justify the bet, cost reduction becomes the path to returns.
Cui bono? Shareholders of AI companies that successfully reduce labor costs. Capital owners broadly as margins expand. Everyone else should be asking much harder questions about what comes next.
