Dear CIO,

Every time a new technology arrives, someone writes an article about Jevons Paradox. Here is mine, but with one important caveat. We do not know that AI will follow the same path as previous waves of efficiency revolutions did.  History does not repeat on command. AI is different from steam engines, factory automation, or cloud computing, and knowledge work is not the same as manufacturing. Regulation, trust, organizational capability, capital investment, and countless other factors will shape what happens next. Still, if history is any guide, I would not bet on AI simply eliminating work. Instead, I would bet that it changes the nature of work, expands what organizations attempt, and shifts value toward the people and systems that learn to use it well.

Best Regards,
John, Your Enterprise AI Advisor

Why history suggests AI will expand knowledge work more than it replaces it.

William Stanley Jevons, a 19th-century British economist, was not writing about artificial intelligence or even factories. He was writing about coal. The simple version is this: when we make something more efficient, we often use more of it, not less. More efficient steam engines did not decrease coal consumption. They made coal-powered activities cheaper and more practical, which increased demand for coal overall. We have seen that pattern play out repeatedly.

Take the automotive industry. Over the past century, robots, automated tooling, sensors, and sophisticated production systems have dramatically reduced the amount of human labor required to build a vehicle. Yet automation did not shrink the industry. It allowed manufacturers to build more vehicles, with more features, for more customers than ever before. Toyota is a perfect example of this. The Toyota Production System was never built around replacing people with machines. One of its central ideas, jidoka, often translated as "automation with a human touch”, was about designing systems that could detect problems, stop when something was wrong, and give people the ability to improve the process. Automation here was not the destination, but rather better capability. As production became more efficient, the work became more sophisticated. Today's automaker is coordinating global supply chains, managing software platforms, integrating batteries and sensors, complying with regulations across dozens of countries, and shipping vehicles that increasingly resemble rolling computers. Efficiency is what enabled it. As complexity grew, so did demand for different kinds of expertise: robotics technicians, industrial engineers, software developers, quality specialists, supply chain experts, and countless others whose jobs barely existed in previous generations of manufacturing.

The same pattern appeared in software. Cloud computing made infrastructure dramatically cheaper and easier to consume, and DevOps shortened deployment cycles and reduced the friction of delivering software. None of that reduced the demand for computing. McKinsey estimates cloud adoption could generate roughly $3 trillion in EBITDA value for Global 2000 companies by 2030. That is a Jevons-like effect: cheaper computing led to more computing, faster deployment led to more deployment, easier infrastructure led to more infrastructure, and better software delivery led to more software demand.

AI may represent a similar shift for knowledge work. If AI makes writing, analysis, coding, support, research, and decision preparation cheaper, the result may not be less knowledge work. It may be much more of it, but that does not mean everyone is safe. Work that consists primarily of routine, low-context tasks, like summarizing documents, formatting reports, routing information, producing generic first drafts, or following predictable workflows, will likely face increasing pressure. The advantage shifts toward people who bring judgment, domain expertise, systems thinking, and ownership of outcomes. AI amplifies those capabilities far more effectively than it replaces them. That is where the automotive comparison becomes useful. The robot did not eliminate the factory, but rather changed the factory. DevOps did not eliminate operations, but rather, changed operations, development, security, testing, reliability, and the relationship between business and technology. AI will likely do the same to knowledge work.

Of course, none of this guarantees a painless transition. Jevons-like effects create winners and losers. Total demand can grow while particular jobs disappear. Entire categories of work become less valuable even as new ones emerge. More value may accrue to platforms, infrastructure, data, and organizations that learn fastest. That is why I find both extremes unconvincing. Neither "AI will eliminate jobs" nor “AI will create jobs” captures what is actually happening. A better way to think about it is this: AI reduces the labor required for many forms of knowledge work, but if history offers any guidance, lowering that cost may increase the total amount of knowledge work organizations decide is worth doing. If that is true, the future belongs to people who understand systems rather than just tasks. The lesson from Toyota wasn't that automation replaces humans. It was that automation works best when it strengthens quality, learning, and continuous improvement. The lesson from cloud and DevOps was not simply lower costs. It was that making computing easier unlocked entirely new ways of building software. AI may ultimately teach us the same lesson. We will not know until history has had its say, but if I had to place a bet, it wouldn't be on the end of knowledge work. It is the beginning of a much larger redesign.