Does AI Productivity Boost Lead to More Hiring? Jevons Paradox Explained

Mar 26, 2026

I kept seeing the same argument everywhere: “AI will make engineers 10x more productive, so companies will fire 90% of them.”

The logic seemed sound. If one engineer can now do the work of ten, why would you need ten engineers anymore?

But then I stumbled upon a 150-year-old economic theory that completely flipped my understanding. And the more I researched, the more I realized this “obvious” conclusion was dead wrong.

The Problem: The Efficiency-Reduction Assumption

The argument goes like this:

AI makes workers efficient
      ↓
Same output needs fewer workers
      ↓
Companies fire excess staff
      ↓
Labor demand crashes

This linear thinking assumes demand is static. If we need 100 features and AI makes each feature 4x cheaper to build, we only need 25% of the engineers.

I believed this too. Until I discovered what happened with ATMs.

The ATM Paradox

In the 1970s, banks introduced ATMs. The obvious prediction: bank tellers would disappear. Why pay humans to dispense cash when machines can do it?

Here’s what actually happened:

ATMs introduced
      ↓
Expected: Fewer tellers
Actual: Bank branches increased 40%
      ↓
More tellers hired
Reason: Lower cost per branch → More branches opened

Wait, what?

The key insight: Lower cost per branch made it profitable to open MORE branches. The number of bank branches in the US grew from ~20,000 in 1970 to over 80,000 by 2010. And the number of bank tellers? It grew too.

I was confused. How does efficiency lead to MORE resource consumption?

Enter Jevons Paradox

In 1865, economist William Stanley Jevons observed something odd about coal consumption in England:

Steam engines became more efficient
      ↓
Coal consumption INCREASED, not decreased

The logic:

Efficiency Gain → Lower Cost → Higher Demand →
Greater Total Resource Use

Why? Because cheaper steam power made it profitable to use steam engines in applications that were previously too expensive. The total market for steam power exploded.

Jevons Paradox: When efficiency lowers the cost of using a resource, total consumption of that resource may increase rather than decrease.

Applying This to AI and Software Engineers

Let me walk through the logic with real numbers.

Scenario 1: Static Demand (The Common Assumption)

Company needs 100 features
Each feature costs $100K to build
Total budget: $10M

AI makes development 4x cheaper
Each feature now costs $25K

Do they:
A) Build same 100 features, save $7.5M, fire 75% of staff?
B) Build 400 features with same budget?

Most AI-doom predictions assume Option A. But that’s not how markets work.

Scenario 2: Dynamic Demand (The Reality)

What actually happens:

Phase 1: Cost Reduction
├── AI reduces cost per feature by 50-75%
├── Projects previously "too expensive" become viable
└── Companies can now afford 4x features

Phase 2: Demand Expansion
├── Competitors gain AI leverage → You must keep up
├── New market segments become accessible
├── Previously unprofitable products become profitable
└── "Nice to have" features become "must have"

Phase 3: Net Labor Effect
├── More projects = More project managers
├── More features = More QA, more DevOps
├── More markets = More product managers, more sales engineers
└── Net result: Labor demand INCREASES

Historical Evidence: The Spreadsheet Revolution

When VisiCalc (the first spreadsheet) and later Excel were introduced, everyone predicted the end of accountants.

Reality: Accountant jobs grew 4x over the following decades.

Why? Analysis became cheaper. Companies that previously analyzed finances once per quarter started analyzing monthly, weekly, daily. New types of analysis became viable. Financial modeling expanded. And yes, more accountants were needed.

A Simple Model

Let me show you the math:

# Traditional view (static demand)
def labor_demand_old(productivity_gain, current_workers):
    return current_workers / productivity_gain
    # Fewer workers needed

# Jevons Paradox view (dynamic demand)
def labor_demand_new(productivity_gain, current_workers, market_elasticity):
    # Efficiency lowers cost → demand rises
    new_projects_viable = productivity_gain * market_elasticity
    return current_workers * new_projects_viable * 0.6
    # Net growth (after adjustment)

# Example with 4x productivity:
print(labor_demand_old(4, 100))    # Output: 25 workers (job loss)
print(labor_demand_new(4, 100, 1.5))  # Output: 90 workers (net growth)

The 0.6 factor accounts for some displacement, but the key insight is market_elasticity. When software becomes cheaper, companies don’t just make the same software cheaper—they make more software.

The Ice Cream Shop Mental Model

This helped me understand it intuitively:

Imagine you own an ice cream shop:

New machine makes ice cream 4x faster

Option A: Keep same output, fire 3/4 of staff
         → Save money, but competitors with machine outsell you

Option B: Sell 4x ice cream, keep or expand staff
         → What actually happens

Why Option B wins:
├── Prices drop (competition)
├── More people buy ice cream (demand rises)
├── New flavors become viable (variety expands)
├── Ice cream catering becomes profitable (new markets)
└── Net result: More staff needed across expanded operations

The Reddit Debate That Clarified Everything

I found a fascinating debate on Reddit about this exact topic. Here are the key insights:

The Efficiency Argument:

“When the price for a product comes down people buy more. Price comes down due to competition and typically companies make more at the price equilibrium point. The Jevons Paradox I mentioned is real.”

The Counter-Argument:

“I didn’t say demand is fixed. I said the number of people creating the demand is limited to the amount of people on earth. Meaning if you increase output quickly the demand doesn’t necessarily follow even if the prices are falling.”

The Strategic Question:

“If engineers can produce 4 times as much value for the same cost, are you going to fire a 3/4 of your staff, or make 4 times as much stuff?”

The Scale Perspective:

“The software license market is only a few hundred Billion. The labor market is 11T in the US. This was always the intent.”

The counter-argument is valid in isolation—there IS a population constraint. But it misses how efficiency transforms markets. The $11T labor market (vs. the few hundred billion software market) represents vast untapped demand that AI-powered efficiency can unlock.

Why This Matters

For Individuals

Skills that combine domain expertise with AI leverage become more valuable
The question shifts from “Will I be replaced?” to “What can I build now that was impossible before?”
Career opportunities expand at the margins—new industries, new roles

For Companies

Competitors with AI leverage will ship faster
Standing still while others gain efficiency means losing market share
The winning strategy is building more, not hiring less

For Policy Makers

Historical precedent suggests efficiency gains expand markets
Labor market disruption is real but net-negative job loss isn’t inevitable
Focus on transition support, not protectionist policies

Common Mistakes in This Debate

Mistake 1: Static Demand Assumption

Assuming demand for software is fixed. It’s not. Cheaper production creates new demand.

Mistake 2: Ignoring Time Horizons

Short-term: Some displacement occurs
Medium-term: New roles emerge
Long-term: Market expansion dominates

Mistake 3: Conflating Efficiency with Automation

AI tools amplify human capability (efficiency)
Full automation is different (replacement)
Most current AI is efficiency-enhancement, not full automation

Mistake 4: Missing Second-Order Effects

More software = More security needs
More features = More customer support
More automation = More oversight requirements

What History Tells Us

I found consistent patterns across multiple technological shifts:

Technology         Expected Impact         Actual Result
────────────────────────────────────────────────────────────
ATMs               Fewer bank tellers      More branches, more tellers
Spreadsheets       Fewer accountants       4x accountant jobs
Computers          Paperless office        More paper, more admins
Cloud computing    Fewer IT staff          More IT complexity, more staff
E-commerce         Death of retail         New retail models, more workers

Each time, the “obvious” prediction was wrong. Not because technology didn’t change things, but because efficiency expanded markets rather than shrinking them.

The Key Takeaway

Jevons Paradox provides the economic framework to understand why AI productivity gains may lead to more hiring, not less. The key insight is that efficiency doesn’t just reduce costs—it expands possibilities.

When software becomes cheaper to build, companies build more software, enter more markets, and create more products. This expansion requires more human expertise, not less.

Don’t bet against 150 years of economic evidence. Efficiency expands markets. AI will likely create more work than it eliminates—the question is whether you’re positioned to do the new kinds of work that emerge.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!