Explainer
The Demand Elasticity Effect
Why making work cheaper can create more jobs, not fewer — and where this does and doesn’t apply.
When AI makes work cheaper, the question isn’t just how many jobs disappear — it’s how many new ones become worth doing. The answer depends on two things: how much demand grows when costs fall, and how many tasks in a job AI can actually improve.
Jevons Paradox, Applied to Labor
In 1865, economist William Stanley Jevons observed something counterintuitive: as steam engines became more fuel-efficient, total coal consumption increased rather than decreased. More efficient engines made steam power economical for uses that were previously too expensive, and the resulting explosion in demand for steam-powered machinery more than offset the per-engine fuel savings.
The same dynamic applies to labor markets. When AI dramatically reduces the cost of producing something, two things happen simultaneously: each worker produces more output (fewer workers needed per unit), and the total market for that output may grow (more units demanded at lower prices). The net effect on employment depends on which force is larger.
This is not a theoretical curiosity. It has played out repeatedly across technologies. And early evidence suggests it is already playing out with AI in sectors where demand is elastic.
Coal
1865
Efficient engines = more total coal consumed
ATMs
1970s
Cheaper branches = more tellers employed
Excel
1980s
Automated calculation = more accountants hired
Desktop Publishing
1990s
Cheaper publishing = more designers needed
See the Effect Across Industries
Most displacement projections assume a fixed demand for output: AI automates tasks, fewer workers are needed, end of story. But that ignores the demand side. In some industries, lower costs unlock entirely new markets. In others, they don’t.
Why does low elasticity mean more displacement? When demand is inelastic, the total amount of work people want done barely changes even when it gets cheaper. Nobody needs more payroll processing, toll collection, or compliance paperwork just because AI can do it for less. So when AI makes each worker 3x more productive, you simply need a third of the workers. The efficiency gain translates almost entirely into job cuts, not growth.
Why does high elasticity mean expansion? When demand is elastic, cheaper output creates proportionally more buyers. Custom software at $500K serves Fortune 500 companies. At $50K, it serves every mid-sized business. The total market grows faster than per-worker productivity, so more workers are needed, not fewer. Toggle below to see the difference.
The Elasticity Effect by Industry
See how demand elasticity changes the displacement story. Toggle to stretch or compress each industry’s outlook.
The Two Conditions
Not every job that AI touches will see demand expansion. It requires two conditions to hold simultaneously:
Demand elasticity
Does cheaper output create proportionally more demand? When custom software drops from $500K to $100K, many more companies will build it. When toll processing drops in cost, nobody drives more to generate extra tolls.
High task exposure
Are enough tasks in the job AI-augmentable to meaningfully change the cost equation? A 5% efficiency gain won’t move the needle. A 5x productivity multiplier changes which projects are viable.
When both conditions hold, the Jevons effect dominates: AI exposure leads to more hiring, not less. The matrix below maps this framework across job categories:
Low demand elasticity
Cheaper doesn’t mean more demand
High demand elasticity
Cheaper unlocks much more demand
High AI exposure
Many tasks automatable
Displacement
AI handles many tasks, but demand doesn't grow when it gets cheaper. Net effect: fewer workers.
Back-office data entry, toll processing, basic bookkeeping
Expansion
AI handles many tasks AND lower costs unlock new demand. Net effect: more workers, often at higher wages.
Software engineering, creative services, legal aid, data analysis
Low AI exposure
Few tasks automatable
Minimal impact
Few tasks automatable, demand doesn't change much. AI has limited effect on this work.
Plumbing, electrical work, childcare
Modest growth
Few tasks directly automatable, but AI tools at the margins unlock some new demand.
Healthcare (with AI diagnostics), skilled trades (with AI estimation)
The Agentic Coding Example
Take a simplistic but instructive example. A mid-sized company has a project they want to build software for. Before AI, it takes 50 engineers to fully resource the effort, but the project doesn’t provide the ROI to justify 50 engineers compared to other initiatives. Or the company knows its expertise isn’t in building software, so it’s not even worth starting. The company hires 0 engineers and doesn’t start the project.
Now, AI agents make it possible for this to be a 10-engineer problem. The ROI calculus immediately changes. So instead of hiring 0 engineers, the company hires 10 with AI agents. Net effect: +10 engineers, not -40.
This has compounding implications. Coding can now have impact for internal workflow automation, systems integration, data analysis, and customer-facing product innovation. By bringing down the cost of writing code, we can just begin to use it for far more.
By making engineering cheaper, AI doesn’t just accelerate existing projects. It makes entirely new categories of projects viable. The demand for software is nowhere near saturated.
See It Yourself
Adjust the productivity multiplier and demand elasticity to see how workforce size changes across different sectors. Try the presets to see why the same AI gains produce opposite employment effects in different industries.
Demand Response Calculator
Adjust AI productivity gains and demand elasticity to see how workforce size changes. When demand is elastic, cost reductions create more total work.
0
Net worker change
per 100 original workers
300%
Demand vs. before
-67%
Cost per unit
How this works: AI reduces cost per unit of output by the productivity multiplier. Demand responds to lower prices based on elasticity (elasticity of 1.0 means 10% lower price = 10% more demand). Net workforce = new demand divided by new output-per-worker. When elasticity > 1 and productivity gains are large, the demand increase outweighs the per-worker efficiency gain, resulting in net job growth.
Where Expansion vs. Contraction Is Likely
Not every sector will follow the same path. The elasticity of demand varies enormously, and so does the magnitude of AI-driven productivity gains.
Software & Engineering
Vast unmet demand for automation, custom tooling, integration, and data analysis. Companies that couldn't afford software projects now can. The long tail of software needs is enormous.
Creative Services
Bespoke creative work (custom video, illustration, copywriting) was unaffordable for most SMBs. Lower production costs expand the addressable market for creative services.
Legal Services
Large latent demand for affordable legal services, but regulatory friction, licensing requirements, and liability concerns slow expansion. Contract review and compliance could expand; litigation less so.
Customer Service
Consumers don't want more support interactions at lower prices. Some expansion in proactive service, but automation of existing volume dominates.
Data Entry & Back-Office
Nobody wants more data entry. When AI automates it, the work simply goes away. This is the textbook case where efficiency reduces total labor.
Healthcare
Enormous unmet demand for healthcare, but constrained by regulation, licensing, and institutional inertia. AI diagnostics could unlock demand if regulatory frameworks adapt.
The Reallocation Caveat
Even where demand elasticity predicts net job growth, the transition will not be smooth. Expansion happens at the sector or economy level, but displacement happens at the firm and individual level. The worker displaced at Company A is not automatically hired at Company B.
Real frictions in the transition
- •Skill mismatch — 10 AI-augmented engineers require different skills than 50 traditional engineers. The displaced 50 can’t all become the 10 new hires without retraining.
- •Geographic mismatch — New demand may emerge in different cities, regions, or countries than where displacement occurs.
- •Timing asymmetry — Displacement is fast (firms can cut costs this quarter). New demand creation is slow (new projects need planning, hiring, onboarding). The gap between destruction and creation can be painful.
- •Wage adjustment — Even when more jobs exist in aggregate, some workers transition to lower-paying roles during the reallocation period.
These frictions are real, and policy matters. But the core dynamic holds: in sectors with elastic demand, AI-driven cost reductions will create more total work than they eliminate. The historical record on this is remarkably consistent. The challenge is managing the transition, not preventing it.
The default assumption — that automation always means fewer jobs — is wrong in any market where demand is elastic. The question for each sector is not “how many tasks can AI do?” but “how much more of this work would the world buy at a fraction of the current price?”
Continue exploring