Jevons’ Paradox and AI: Why Making Work Cheaper Creates More Work, Not Less

In 1865, an English economist named William Stanley Jevons noticed something that nobody expected. The steam engine had just gotten dramatically more efficient — it used far less coal to produce the same amount of power. You'd think coal consumption would drop. Less coal per engine, less coal overall. Simple math.

But coal consumption skyrocketed.

More efficient engines didn't just save coal — they made coal-powered machines viable in places where they never were before. Factories that couldn't afford to run steam engines suddenly could. New industries emerged. New applications were invented. The cheaper coal became to use, the more uses people found for it.

Jevons called this "the rebound effect." Economists now call it Jevons' Paradox: when technology makes a resource significantly more efficient to use, total consumption of that resource tends to go up, not down.

This is the single most important framework for understanding what AI is about to do to human work.

The Replacement Narrative

The dominant story about AI and jobs goes like this: AI can write, code, design, analyze data, generate images, draft legal documents, and answer customer questions. Therefore, AI will replace the people who currently do those things. Writers, developers, designers, analysts, illustrators, paralegals, customer service reps — all on the chopping block.

It's a clean, logical narrative. It's also almost certainly wrong — not because AI can't do these things, but because it misunderstands what happens when you make something radically cheaper.

What Actually Happens When You Cheapen Work

Think about what happened when desktop publishing software made graphic design cheaper. Before Photoshop and InDesign, producing a professional-looking brochure required hiring a typesetter, a graphic artist, and a print specialist. It was expensive. So most small businesses didn't do it.

When the tools got cheap, design didn't disappear as a profession. The world didn't need fewer designers. What happened was an explosion of demand. Suddenly every business needed a logo, a website, social media graphics, branded email templates, pitch decks, trade show banners, YouTube thumbnails, app icons. Things that nobody would have paid a typesetter to produce in 1985 became essentials by 2005.

The tools got cheaper. The demand got bigger. The total amount of design work done by humans increased massively.

The same thing happened with video. Professional video production used to require a crew, expensive cameras, and an editing suite that cost more than a house. When smartphones and tools like Final Cut Pro made video production nearly free, video didn't become a dead industry. It became the dominant medium of the internet. More humans produce more video today than at any point in history.

Web development followed the same arc. When website builders made it possible to create a site without coding, the number of websites didn't plateau — it exploded from a few million to over a billion. And the demand for web developers went up, not down, because all those new websites created demand for customization, integration, optimization, and maintenance that the tools couldn't handle alone.

This is Jevons' Paradox playing out across every creative and knowledge industry. Make something cheaper to produce, and society finds a thousand new reasons to produce it.

Applying This to AI

AI is making a wide range of knowledge work dramatically cheaper. Writing that once took a copywriter four hours now takes thirty minutes of prompting and editing. A rough application prototype that took a developer two weeks now takes an afternoon. Data analysis that required a specialist can now be done by anyone who can describe what they're looking for.

The replacement narrative says: fewer copywriters, fewer developers, fewer analysts.

Jevons' Paradox says: radically more writing, more software, more analysis — because the cost barrier that prevented most of this work from happening just collapsed.

Think about how many small businesses have no marketing copy at all — no blog, no email sequences, no social media content — simply because they couldn't afford a copywriter. AI doesn't just replace the copywriter they had. It unlocks the marketing they never did.

Think about how many internal tools never got built at companies because developer time was too expensive to spend on "nice to have" projects. AI doesn't just replace those developers. It makes a thousand small software projects suddenly worth doing.

Think about how many nonprofits, churches, freelancers, and solo entrepreneurs don't have websites — not because they don't need one, but because the cost was too high. Make website creation nearly free, and you don't eliminate web design as a category. You bring millions of new websites into existence that otherwise wouldn't have existed.

The pie gets bigger. It always gets bigger.

Where the New Opportunities Are

If Jevons' Paradox holds — and it has held for 160 years across every technology that has dramatically reduced costs — then the right question isn't "what jobs will AI take?" It's "what work becomes possible now that AI has made the inputs cheap?"

Here's where to look:

The taste layer. AI can generate a hundred options. It can't tell you which one is right for your audience, your brand, your context. The ability to curate, judge, and direct — to have taste — becomes more valuable as generation becomes trivial. Art directors, creative directors, editors, and strategists don't become less important when production is cheap. They become the bottleneck, which means they become more important.

The customization layer. AI is good at producing generic outputs. Turning a generic output into something specific — tailored to a particular business, audience, market, or use case — is where human judgment thrives. The person who can take an AI-generated website and make it actually convert for a specific plumber in a specific city is doing work that AI enabled but can't finish.

The trust layer. As AI-generated content floods every channel, trust becomes scarce. People will pay a premium for human-verified information, human-reviewed recommendations, human-created work that carries a reputation behind it. Doctors, lawyers, financial advisors, consultants — anyone whose value comes partly from accountability — will find that AI increases demand for their judgment, not decreases it.

The new-category layer. This is the biggest one. Entire categories of work that don't exist yet will emerge because AI made them economically viable. Just as YouTube created the category of "content creator" and the App Store created the category of "indie app developer," cheap AI will create job categories that we literally don't have names for yet. Someone who manages a fleet of AI agents to run personalized marketing for 50 small businesses simultaneously. Someone who designs custom AI workflows for specific industries. Someone who audits AI outputs for accuracy in regulated fields.

We can't name all of these yet. That's the point. The new jobs created by a major technological shift are, by definition, hard to predict in advance. What we can predict is that they'll exist — because they always have.

The Real Risk Isn't Replacement

None of this means the transition will be painless. Jevons' Paradox describes the aggregate outcome — more total work, more total demand — but it doesn't promise that the transition is smooth for every individual.

The real risk isn't that AI eliminates the need for human work. It's that the work shifts faster than people can adapt. The copywriter who insists on doing everything the old way will struggle. The copywriter who learns to use AI as a force multiplier — producing ten times the output at higher quality — will thrive.

The opportunity goes to people who understand that AI is a tool that makes their work cheaper to deliver, which means there's more demand for it, which means they can serve more clients, enter new markets, and do work that wasn't economically viable before.

The Bottom Line

Every major technology that has made a type of work cheaper has ultimately created more of that work, not less. Steam engines didn't reduce the need for energy — they created the industrial revolution. Spreadsheets didn't eliminate accountants — they made financial analysis so accessible that every company now has an entire finance department. The internet didn't reduce the need for communication — it created an economy where communication is the primary activity of the workforce.

AI will follow the same pattern. Not because it's inevitable, but because the underlying economic logic is sound: when you drop the cost of something valuable, demand scales faster than supply shrinks.

The jobs of 2030 won't look like the jobs of 2020. They never do. But there will be more of them, they'll produce more value, and they'll be available to more people — because that's what happens when powerful tools get cheap.

Jevons saw it in 1865 with coal. We're about to see it again with everything AI touches.