Skip to main content

Jevons Paradox

Efficiency goes up. Costs go down. Demand goes up. AI does not shrink the workforce. It expands what is worth doing.

The Cut-to-the-Top Myth​

Name a company that won big by shrinking. Take your time.

Operational efficiency matters. Trimming fat keeps companies alive when margins tighten. But no one becomes a category leader by cutting. Winners invest, innovate, and grow. They treat technology as a lever to expand capability, not a scythe to cut headcount.

The smartest organizations already see this about AI. They are not asking "how many people can I replace?" They are asking "how much more can my people do?" Augmented intelligence, not artificial replacement.

Jevons Paradox is the economic mechanism that explains why the second question is the winning one.

The 1865 Observation​

In 1865, economist William Stanley Jevons was studying coal in England. Steam engines were getting more efficient. Less coal was needed per unit of work. Common sense said coal demand would fall.

It rose.

As steam engines became cheaper to run, more industries adopted them. New use cases appeared that were previously uneconomic. Energy became something you could spend on anything, rather than a precious resource you had to ration. Total coal consumption went up, even as the coal required for any single task went down.

That is Jevons Paradox in one sentence. When a resource becomes more efficient to use, total consumption tends to rise, because demand expands into everything that was previously too expensive to bother with.

The Spreadsheet Case​

Spreadsheets automated arithmetic. The obvious conclusion in the 1980s: accountants are cooked. Nobody will pay a human to add up columns when a machine does it for free.

That is not what happened.

Demand for financial analysis exploded. Freed from clerical arithmetic, accountants and analysts moved into forecasting, modeling, strategic advisory, and scenario planning. Entirely new categories emerged: FP&A analyst, financial data analyst, Excel-wielding strategic operators, CFO-as-a-service firms. Companies hired more finance professionals, with higher skill bars, doing higher-value work.

Efficiency went up. Cost went down. Demand for the human role went up, not down. That is Jevons Paradox in a modern labor market.

The Radiologist Prediction​

In 2016, Geoffrey Hinton said radiologists should stop being trained. In five years, he argued, deep learning would be better than any radiologist. It was a rational prediction rooted in what the technology clearly could do.

Ten years later, in 2026, the United States is training more radiologists than it did in 2016. AI got dramatically better at medical imaging. It now reads scans alongside human specialists, flags subtle findings, and helps triage cases faster. The result was not replacement. It was expansion of what radiology could cover, which expanded demand for human radiologists running AI-integrated workflows.

This is what Hinton missed. Capability is not a ceiling on demand. It is the thing that creates new demand.

What This Means for AI​

Jevons Paradox predicts the specific shape of what AI will do to the labor market. Fewer low-discretion routine roles. More high-context, high-accountability roles. Implementation commoditizes. Judgment, orchestration, and relationship work expand.

Concretely:

Entirely new categories of work​

Jobs that did not exist before. AI product managers. AI safety engineers. Prompt engineers. Applied AI practitioners. Agentic OS trainers. Chief AI Officers. These roles are being defined right now, by the people filling them for the first time. This is where the fastest wage growth of the decade is landing.

Long-tail services that were priced out​

Custom tutoring for every kid. Niche legal analysis for small-business cases. Personalized healthcare coaching. Boutique research. Every service that was previously only affordable to people with significant resources is becoming feasible for everyone. Which means more work for the humans who design, oversee, and personalize those systems.

Higher expectations, which create more work​

When AI raises the bar on quality, speed, and availability, "good enough" moves up. Which means more integration work, more oversight, more compliance, more trust infrastructure. A single customer service thread that used to end in one email now involves AI drafting, human review, sentiment analysis, retention modeling, and follow-up scheduling. The total amount of work goes up even as the labor per task goes down.

High-context and high-accountability roles expand​

Problem framing. Goal setting. Judgment calls. AI supervision, evaluation, governance. Customer-facing work where trust matters. Cross-functional coordination. These are the roles that move up the food chain as routine work moves down it.

The Skills That Compound​

Smart employees in the Jevons era look different from smart employees in the pre-AI era. Five capacities compound:

  1. Adaptability and flexibility. The shape of the work keeps changing. People who can pivot keep winning. People who want their job description to stay the same for ten years are choosing a shrinking bucket.
  2. Lifelong learning. Information is coming faster every quarter. People who treat learning as a recreational activity rather than a chore will outpace the rest.
  3. Critical thinking. AI is often right. "Often" is not "always," and "right" is not always "what we wanted." The human in the loop has to notice the gap and correct it. Strategy is the new execution.
  4. Creativity. With AI handling the mundane, there is more time for big-picture thinking. People who can imagine what should exist (not just optimize what already does) become unreasonably valuable.
  5. Trust, relationships, and taste. The things AI cannot give you: presence, judgment, care, legitimacy. Relationship capital is the hardest form of leverage in this economy.

These are the same capacities RIP To The Career Ladder names from the labor-market side. Jevons Paradox is the economic mechanism underneath: as AI commoditizes execution, the ceiling on human work rises, and the skills that let you operate at the new ceiling become irreplaceable rather than merely nice to have.

For Business Owners​

The trap is obvious once you can name it. You can spend this decade trying to cut your way to the top, or you can spend it expanding what your people can accomplish with AI in their hands.

Winners will hire applied-AI-native operators and train them deliberately, then push them into the work that was previously uneconomic: the long-tail services, the higher quality expectations, the new categories. Cutting to the bone gets you short-term margin and a shrinking business. Investing to the edge gets you a new moat.

The parking-lot metric (fewer employee cars) is a lagging indicator of losing. The capability metric (more that your team can credibly ship per quarter) is the leading indicator of winning.

For Practitioners and Operators​

If you are early in your career or between roles, the right mindset is not defensive. It is expansionary. The question is not "how do I avoid being replaced?" It is "how do I become the person who uses these tools to do ten times what any single person could do before?"

  • Build your Personal Agentic OS.
  • Practice spec writing and agentic strategy daily.
  • Get into the new job categories before they are crowded.
  • Pick a long-tail service that was previously uneconomic and own it with AI.
  • Develop the five capacities above. AI cannot provide them for you.

The Cars in the Parking Lot​

Leaders who are optimizing for fewer cars in the employee parking lot are reading the wrong scoreboard.

The right scoreboard is: how much more can your organization do this year than last, per person, per week, per dollar? That is Jevons growth. That is where winners get made.

The train is already pulling away from the station. The question is whether you are running toward innovation and expansion, or toward the HR office with a list.

Further Reading​