How AI’s Efficiency Boom Could Actually Create More Jobs (Not Less): The “Jevons Paradox” Reborn in the AI Era
In the AI-age debate over whether machines will steal our jobs, one concept from 19th century economics is quietly making a comeback: the Jevons Paradox. This principle suggests that as a resource becomes cheaper or more efficient to use, we often end up consuming more of it overall—not less. In the context of AI, that means greater automation doesn’t necessarily translate to mass unemployment—it may fuel fresh waves of demand, new job roles, and broader economic transformation.
In this article I’ll walk you through how that works, where the risks are, and what this could mean for your career, organisations, and society. Along the way, I’ll bring in some recent insights, challenge some assumptions, and offer a fresh angle you might not see elsewhere.
Efficiency, Demand & Rebound: The Core Idea
William Stanley Jevons first observed, in the mid-1800s, that when steam engines became more efficient, coal usage actually increased. The logic is simple: if you can do more with less cost, more people adopt the technology and use it more intensively. Jevons observed this in Britain’s coal use and dubbed it a “paradox” to the intuitive assumption that better efficiency reduces consumption. (en.wikipedia.org)
Fast forward to today. AI promises to make tasks faster, smarter and cheaper—interpreting images, generating content, analysing data. So, by the same principle, when AI lowers the “cost” of certain cognitive services, more people and organisations will demand them. That rising demand can spawn new sectors, new workflows and entire classes of jobs that didn’t exist before.
Indeed, in software and tech circles, this is already being discussed in terms of the “10× developer paradox” or how AI tools might expand development work rather than shrink it. (medium.com)
In effect, we may see a rebound effect: efficiency yields lower cost per task, but the volume of tasks grows so strongly that total demand—and total workload—goes up.
Why Radiologists Are Still in Business
One of the more surprising examples comes from healthcare. Back in 2016, AI pioneer Geoffrey Hinton claimed that radiologists might become obsolete within five years. But nearly a decade later, demand for radiologists is higher than ever.
How come? Because AI made scans cheaper and faster, which drove up the number of scans being done. And more scans means more complex interpretations, follow-ups, treatment planning—areas where human doctors are still needed. The cheaper imaging enabled new usage, which in turn strengthened demand for human expertise.
That perfectly illustrates Jevons’ principle: improved efficiency unlocks latent demand.
It also shows a key nuance: AI often automates only a portion of a workflow. Humans still play roles in quality control, oversight, edge cases, and making decisions about what to do next. So the new human tasks may shift rather than disappear entirely.
Observations like these suggest that we need to stop thinking in substitution alone—and start thinking in amplification and recomposition.
The AI Job Debate: Collapse or Transformation?
Many people today fall into one of two camps:
- The doomsters: AI is going to displace millions of jobs, render large swathes of human labour redundant, and trigger widespread unemployment.
- The tech optimists: AI is overhyped, and while it may change work, it won’t fundamentally upend our economic systems.
But both sides miss what history and economics tend to teach us: transformations like this are seldom zero-sum. They tend to reconfigure labour, shift skill demands, and reorganise industries.
If you believe in the Jevons Paradox, AI’s efficiency gains are a feature, not a bug, of a broader growth cycle.
Already we’re seeing early signs of this in sectors such as financial services. In equipment finance, for example, applying automation to underwriting and application processing is making it easier to scale operations—and thus creating demand for more staff to manage the expanded volume of business. (monitordaily.com)
Still, to be realistic: not every job will survive. Some roles—especially highly repetitive, low-context ones—face real risk of commoditisation or elimination. But many of the new or transformed roles will require higher-order skills: judgement, oversight, adaptability, domain expertise plus AI fluency.
In fact, recent academic research has looked at the tension between substitution and complementarity of AI and human skills. One study found that across 12 million job listings in the U.S. between 2018 and 2023, the demand for AI-complementary skills (digital literacy, adaptability, teamwork) rose significantly compared to roles heavily reliant on skills that AI could more easily replace. (arXiv)
That aligns with another study which found that workers in higher-skilled non-routine jobs are more “susceptible” to automation—but interestingly, those workers often also see wage growth in areas where their roles become more AI-augmented. (arXiv)
Thus the narrative is not simply “humans vs machines,” but “humans + machines” evolving.
The Dual Paradox of AI
Here’s a fresh twist I like to bring in—one that’s getting attention in scholarly circles: the idea of a dual paradox. Not only might AI trigger a Jevons-style rebound in labour demand, but it might also provoke surge effects in resource consumption (energy, hardware, materials) and generate distributional tension over who captures the gains.
A recent paper, Artificial Intelligence and the Dual Paradoxes, argues that AI’s push for efficiency may lead to increased use of computational resources, greater environmental impact, and a simultaneous reshaping of labour dynamics. (arXiv)
So there’s a twofold tension:
- Efficiency → expansion → more resource usage (data centres, energy, chips)
- Efficiency → skill shift → new labour markets, but also potential inequality or dislocation
The catch is that the rebound in resource use may outpace improvements in sustainability unless policies and design guardrails are built in. In AI domains especially, where each marginal model gets more complex, you see that efficiency gains in one generation often pave the way for even more ambitious, resource-hungry versions in the next. (arXiv)
From a social lens, the winners in this new world may be those who design, govern, maintain and integrate AI systems, rather than those who merely operate them. The fracturing of labour may deepen: some workers will win big, others will struggle to keep pace without upskilling.
So while the Jevons lens gives us optimism about job growth rather than collapse, the dual paradox invites us to think about how sustainable and equitable that growth will be.
But Let’s Not Sugarcoat It: Risks, Limits & Realities
While the Jevons Paradox analogy is powerful, we must also be cautious. There are real limits and caveats:
- Non-elastic demand: Some jobs or sectors are “price inelastic”—you can’t just increase demand by cutting cost. If AI improves productivity in tightly constrained industries, organisations might not expand output sufficiently to absorb displaced labour. (Forgepoint Capital)
- Last mile complexity: Many tasks are messy, contextual, unpredictable. The remaining 10% of edge cases, ambiguous judgements, ethical considerations—automating that is often orders of magnitude harder.
- Skill mismatch & friction: The new jobs AI generates often demand different skills than what the displaced labour force has. Retraining, transitions, credential gaps—these barriers can slow adoption.
- Inequality & capture: If gains accrue to AI platform owners, capital investors, and skilled technologists, inequality may widen. Without mechanisms to redistribute value, social tensions could mount.
- Resource & environmental constraints: The rebound in compute, data centre energy demand, hardware supply chains, e-waste—all these are real physical limits. The dual paradox warns us that efficiency gains alone won’t solve these.
- Overhype and overinvestment: If many AI ventures rely on speculative assumptions of rebound, some may falter. Not every sector is equally capable of scaling.
So the Jevons framing is a lens, not a guarantee.
What Should Individuals, Organisations & Policymakers Do?
Individuals / Professionals
- Cultivate AI-complementary skills: judgment, domain expertise, system design, ethics, adaptability
- Stay curious: experiment with AI tools, learn prompt engineering, automation design
- Be comfortable with evolving roles: managing AI agents, supervising pipelines, interpreting outputs
- Build resilience—multiple skills, flexibility, continuous learning
Organisations & Employers
- Think in augmentation, not just automation
- Design workflows where humans and AI collaborate, rather than replace
- Invest in internal reskilling programs
- Evaluate what to automate versus what to leave human
- Monitor the resource footprint of AI adoption—not just wins
Policymakers & Society
- Incentivise sustainable AI infrastructure (renewable energy, efficient hardware)
- Support retraining, transitional safety nets
- Encourage inclusive access to AI tools to reduce centralisation
- Regulate or tax resource-intensive AI where necessary
- Monitor labour trends and ensure that gains from AI are broadly shared
The Final Word
When people hear AI + jobs, they often frame it as one replacing the other. But borrowed from 19th century economics, the Jevons Paradox offers a richer narrative: greater efficiency can unlock fresh demand, new roles, and expanded systems. In many cases, AI may not shrink total labour—it may reshape it.
Yet that’s only part of the story. The dual paradox reminds us that efficiency has costs beyond labour: energy, hardware, inequality, and environmental externalities. How we guide this transformation—through design, governance, policy and culture—will shape whether this new wave is generative or destructive.
So don’t despair, but don’t be naive either. The future isn’t just about machines doing more—it’s about how humans and machines become a collective, reimagined society.
