There is a common assumption in AI discussions: if AI writes code, demand for developers will drop because less human coding is needed.
That assumption misses a key economic dynamic.
The real thesis is this: AI reduces coding effort per unit, but that lower cost makes many previously uneconomical projects affordable. As more projects become viable, total demand for software rises.
The Jevons Paradox, in plain terms
In the 19th century, economist William Stanley Jevons observed that improvements in coal-use efficiency did not reduce total coal consumption. They increased it.
Why? Because efficiency lowered cost, which expanded usage across more applications.
More affordable input → more use cases → more total demand.
The coal example mapped to AI coding
The historical pattern maps directly to software:
- Then: better steam-engine efficiency lowered effective cost of coal use, and coal consumption rose.
- Now: AI-assisted coding lowers effective cost of software creation, and software demand can rise.
This is not contradictory. It is exactly what Jevons Paradox predicts in systems where reduced unit cost unlocks broader adoption.
What this means for developers
Yes, AI changes the mechanics of coding. But that does not automatically mean less work overall.
As software becomes cheaper to produce, more organizations pursue software projects they previously deferred. That expands total system demand for product thinking, architecture, integration, QA, security, and continuous iteration.
The work shifts. The volume can still grow.
What this means for growing businesses
This is where the shift is most important.
For years, custom software was often out of reach for smaller or mid-sized businesses. Off-the-shelf tools were the default, even when they were a poor operational fit.
AI-assisted development changes that affordability threshold.
More growing businesses can now justify custom-tailored solutions that match their real workflows, instead of forcing workflows into generic software constraints.
The strategic opportunity
Businesses that treat this as an architecture and operating decision—not a hype cycle—can compound advantage faster:
- Higher workflow fit
- Lower process friction
- Faster iteration
- Better scalability from systems designed for their reality
Bottom line
The key question is not “Will AI reduce coding effort?” It clearly will.
The key question is “What happens to total demand when software becomes cheaper to produce?”
If Jevons is any guide, demand rises.
And for growing businesses, that means something powerful: custom software is increasingly within reach.