Jevon's Paradox and the AI Productivity Puzzle
Or why we should worry about 19th century coal innovations
I love paradoxes.
Jevons’ Paradox states that making something more efficient doesn’t mean we use less; instead, we use more.
Jevons figured this out in 1865 for coal consumption in the UK. As steam engines became more fuel-efficient, people assumed coal demand would drop. But the opposite happened—coal became so efficient that it powered more industries, expanding the economy and driving demand through the roof. Efficiency didn’t lead to conservation. It led to acceleration.
I’ve been thinking about this since the DeepSeek announcements. Swap out “coal” for “AI,” and you get a pretty good idea of where we’re headed. There’s this persistent fear that AI will eliminate jobs and make human labor redundant. But if Jevons’ Paradox holds, AI won’t lead to a jobless dystopia. It’ll supercharge productivity, create new industries, and—counterintuitively—increase the total work.
And as we find more efficient ways to run these LLMs, we’ll run them more.
Let’s break it down. AI makes everything faster, cheaper, and more scalable. If we follow Jevons’ logic, that won’t mean fewer workers—it’ll mean:
1. More Growth, Not Less Work – More Inventory, More Experimentation, More Complexity
AI isn’t just cutting advertising costs—it’s expanding what’s possible. Programmatic bidding, dynamic creative optimization, and AI-driven media planning reduce inefficiencies, allowing brands to launch more campaigns, test more variations, and reach more audiences. But instead of shrinking ad teams, this explosion of possibilities means more strategists, data scientists, and creative specialists are needed to fine-tune messaging, optimize engagement, and keep up with the ever-growing complexity of AI-driven campaigns.
Take retail media networks—AI has made them scaleable, turning every major retailer into an ad platform. More efficiency hasn’t led to a slowdown; it’s triggered an arms race in commerce-driven advertising, forcing brands and agencies to invest in AI-powered tools to stay competitive.
2. New Problems to Solve – Entirely New Media Markets
The internet didn’t just digitize old media—it created new economies around content, influencers, and on-demand entertainment. AI is following the same pattern, spinning up entire sub-industries that didn’t exist a few years ago.
Take synthetic media. AI-generated content is turning creative production into a scalable, automated process, but that raises new challenges: How do we ensure brand safety in a world of deepfake influencers? How do advertisers navigate copyright and ownership when AI remixes assets on the fly? These aren’t problems we had five years ago, but they’re shaping entire business models now.
And then there’s contextual advertising. With privacy changes killing off third-party cookies, AI-powered content analysis fills the gap. The result? A new wave of startups specializing in AI-driven brand suitability, contextual targeting, and sentiment analysis—industries that wouldn’t exist without AI-driven ad disruption.
3. Work Will Look Different – Reshaping, Not Replacing Roles
AI isn’t eliminating media jobs—it’s evolving them. Just as the rise of digital advertising didn’t wipe out marketers but shifted them from print to programmatic, AI is transforming how advertising professionals work.
For example, media buyers aren’t disappearing—they’re becoming media strategists who oversee AI-driven bidding algorithms. Copywriters aren’t being replaced—they’re becoming AI editors, fine-tuning machine-generated messaging for different audience segments. Creative directors are leaning on generative AI to spin up hundreds of ad variants, but their role in storytelling and brand vision is more critical than ever.
Look at ad measurement. AI is automating lift studies, MMM, and attribution modeling at a scale that is impossible to manually perform. But instead of making researchers obsolete, it’s forcing them to upskill—moving from raw data crunching to higher-level strategic analysis and insight generation.
Jevons’ Paradox tells us AI won’t just shrink the labor market—it’ll transform it. But the big unknown is who will benefit from that transformation.
• For businesses, AI brings efficiency, but efficiency alone doesn’t guarantee shared prosperity.
• For workers, adaptability is the new job security. The ability to work with AI, not against it, will define career stability.
• For policymakers, the challenge isn’t stopping automation—ensuring AI lifts more boats than it sinks.
If history is any guide, AI won’t be the end of jobs—it’ll be a massive shift in what counts as work. Jevons’ Paradox tells us that efficiency gains don’t mean society slows down—they mean we speed up. AI will make businesses bigger, make some jobs disappear, and create new industries we haven’t even considered. (Who knew retail media was going to be a thing?)
The real risk is not AI itself but failing to adapt fast enough. The future isn’t about fighting automation—it’s about keeping up with what happens next.