The ivory tower is vibrating with excitement, and that should be your first warning sign. For the past eighteen months, mainstream economists have undergone a religious conversion. After a decade of mourning the "productivity paradox"—that awkward reality where computers were everywhere except in the GDP stats—the dismal scientists have finally found their new god: Generative AI.
They are heralds of a new golden age. They cite early studies from MIT and Stanford showing 35% jumps in call center efficiency. They point to Goldman Sachs reports claiming AI could drive a 7% increase in global GDP. They are obsessed with the idea that we have finally found the "General Purpose Technology" that will fix our stagnant standard of living. Read more on a connected subject: this related article.
They are missing the point so spectacularly it borders on professional negligence.
Economists are treating AI as a faster shovel. They believe we are just going to do the same things we’ve always done, just more efficiently. They are looking at the wrong metrics, asking the wrong questions, and preparing for a world that isn't going to exist. The "AI bug" they've caught isn't a vision of the future; it’s a fever dream of the past. More reporting by The Motley Fool delves into comparable views on the subject.
The Efficiency Trap: Why 40% Better is Actually 100% Irrelevant
The current obsession is task-level productivity. Can an AI write this email? Can it debug this Python script? Can it summarize this 50-page PDF?
When researchers like Erik Brynjolfsson or groups like the NBER look at AI, they measure how much faster a human can complete a pre-defined unit of work. This is a classic industrial-era mistake. In a factory, if a machine makes a widget 20% faster, you have 20% more widgets. In the knowledge economy, if you make an email 20% faster, you just get more emails.
We are optimizing for the production of "noise."
I have watched Fortune 500 companies pour tens of millions into "internal productivity tools" designed to help their middle managers churn out more decks. The result isn't a leaner, more profitable company. It’s a company drowning in high-quality, AI-generated garbage that no one has the time to read. We are witnessing the industrialization of bureaucracy.
If your "productivity" gain results in an infinite supply of something with zero marginal value, your economic model is broken. Economists are cheering for a 4% GDP bump that might consist entirely of digital hallucination and administrative bloat.
The Substitution Fallacy
The "consensus" view suggests that AI will augment the low-skilled and replace the mediocre. This is the "Rising Tide" argument: AI raises the floor of human capability.
It’s a comforting thought. It’s also wrong.
In reality, we aren't seeing augmentation; we are seeing the total collapse of the entry-level career path. Economists look at a junior analyst using AI to do a senior analyst’s work and call it "upskilling." I call it the destruction of the apprenticeship model.
When you automate the "grunt work," you remove the training ground where expertise is actually built. You cannot have a 10-year veteran without someone who spent years doing the 2-year level tasks. By automating the bottom of the pyramid, we are creating a massive, invisible talent debt that will come due in a decade.
The data won't show this for years. For now, the charts look great. Costs are down. Output is up. But the "human capital" variable in your $Y = A \cdot K^\alpha \cdot L^\beta$ equation is rotting from the inside out.
The Rebound Effect (Jevons Paradox)
Economists love to ignore William Stanley Jevons, but his 1865 observation is the only thing that matters right now. Jevons noted that as steam engines became more efficient at burning coal, coal consumption didn't go down—it skyrocketed. Because coal was more efficient, it became more useful for more things, and demand exploded.
AI is the "coal" of the 21st century.
As the cost of generating "intelligence" or "content" drops to near zero, we won't save time. We will simply find increasingly complex and useless ways to spend it. We are already seeing this in software engineering. AI makes coding faster, so managers demand more features, more complex architectures, and more frequent deployments. The "saved" time is immediately consumed by the increased complexity that the efficiency enabled.
The economist's dream of a 15-hour workweek is a fantasy because they don't understand that work expands to fill the available computational power. We aren't becoming more productive; we are just running the treadmill faster.
The Great Deflationary Lie
Central banks are watching AI with a mix of hope and terror. They hope AI will be a "supply-side shock" that kills inflation forever. If it costs less to produce goods and services, prices go down, right?
Not in a world of platform monopolies.
In our current market structure, the gains from AI productivity aren't being passed to the consumer in the form of lower prices. They are being captured as rent by the three or four companies that own the compute and the models.
$Profit = Price - Cost$
If AI lowers the cost, but the "Price" is controlled by an algorithmic cartel or a dominant platform, the "productivity gain" is just a massive wealth transfer from the labor class to the GPU-owning class. Economists talk about "Total Factor Productivity" as if it’s a public good. It isn't. It’s a proprietary asset.
Stop Asking "Will AI Replace Jobs?"
This is the most tired, lazy question in the history of the profession. The answer is "yes, and it doesn't matter."
The real question is: "What happens when the value of human cognition hits a floor?"
For two centuries, the economy has been based on the scarcity of human intelligence. We paid people because they knew things or could figure things out. When that scarcity vanishes, the entire structure of the labor market doesn't just "shift"—it dissolves.
We are moving toward a "Post-Cognitive Economy." In this world, the ability to do the work is worthless. The only things that retain value are:
- Responsibility: Who goes to jail when the AI kills someone?
- Physicality: Can you move an atom in the real world?
- Verification: How do we know this isn't a lie?
Economists are still trying to figure out how to tax AI to fund UBI. They should be figuring out how to value "truth" in a world where "facts" are a cheap commodity.
The Strategy for the Sane
If you are waiting for the "AI revolution" to show up in your paycheck or your company’s bottom line because you’re "using it more," you’ve already lost. You are just a more efficient cog in a machine that is rapidly depreciating your specific type of cog.
The only way to win is to stop playing the productivity game.
- Audit for Complexity, Not Efficiency: Don't ask how AI can make a process faster. Ask if the process needs to exist at all now that its output is effectively free.
- Own the Inputs: If you aren't owning the data or the hardware, you are just a tenant on Microsoft or Google’s farm. Your "productivity" is their profit margin.
- Bet on the High-Touch: As AI scales, anything that can be an LLM output will become a low-value commodity. The "unproductive" parts of business—the long lunches, the hand-holding, the physical presence—are the only things that will remain luxury goods.
The economists are right that the "bug" is spreading. They’re just wrong about it being a feature. It’s a total system rewrite.
Stop measuring how much faster you're running. Start looking at where the floor is moving.
Check your internal metrics for "activity" versus "outcome" before you approve your next AI implementation budget.
