AI Jobs Transition (3): Cheaper work can create more work

AI Jobs Transition · 3/6

The standard automation story follows one side of the productivity equation. If AI lets one person produce twice as much, a company needs half as many people to deliver the same output.

That is correct only if output stays the same.

When a product becomes cheaper, faster, easier to customize, or more widely available, people often buy more of it. AI can reduce labor per unit and expand the number of units demanded at the same time. Employment depends on which effect is larger.

This is why demand elasticity belongs at the center of any serious AI jobs analysis.

Productivity creates a labor-saving effect and a scale effect

Imagine an agency that needs ten designers to deliver 1,000 projects a year. AI doubles output per designer, so the same business can now operate with five.

But suppose faster delivery and lower prices increase annual demand to 3,000 projects. The agency now needs fifteen designers at the new productivity level. AI reduced the labor required per project while total employment grew.

The first force is the direct labor-saving effect. The second is the scale effect from expanded demand.

Most claims about automation calculate the first and assume away the second. That works for a fixed internal workload. It fails when AI changes the price, quality, speed, or accessibility of something sold into a market.

Latent demand determines where markets expand

Demand is elastic when customers buy substantially more after prices fall. It is inelastic when quantity changes little.

Some work is constrained by events rather than price. Cheaper firefighting does not create more fires. Some healthcare demand depends on disease incidence, insurance reimbursement, staffing rules, or limited public budgets. Productivity can still improve service, but lower costs may not produce enough additional volume to preserve every job.

Other markets contain large amounts of latent demand. Small companies postpone custom software, design, analysis, legal advice, and marketing because the work is too expensive. If AI cuts the effective cost, these customers enter the market. Existing customers may buy more frequent analysis, more versions, or more personalized service.

This is the business opportunity hidden inside the labor-market framework. The most interesting AI market may not be where a company can remove the most labor. It may be where lower costs release the most previously unaffordable demand.

Better, faster, and more available also change demand

Price is only part of the story. AI changes waiting time, consistency, customization, and quality-adjusted output.

A medical service that becomes faster can treat people who previously abandoned the queue. A legal service that becomes easier to access can serve small disputes that were not worth pursuing. Software that becomes cheap to customize can move from a shared tool to a system tailored to each business.

These changes act like price reductions even when the sticker price stays the same. Customers receive more value for the same money.

For founders, this means a product strategy built only around cost reduction is incomplete. If the company keeps the old offer and simply uses fewer people to deliver it, the upside is capped by the existing market. Redesigning the offer around new frequency, new customer segments, or new quality levels can turn productivity into growth.

Elasticity is a hypothesis to test, not a number to trust

The OpenAI report estimates occupation-level demand elasticity with a GPT model using O*NET job descriptions and a standardized scenario: if the price of an occupation’s output fell by 10 percent, how much more would customers buy over two to three years?

That creates a structured prior, not causal evidence. Occupations do not always have one clean output or market price. Insurance, licensing, fixed budgets, and complementary bottlenecks can prevent theoretical demand from becoming purchases.

Companies can test the mechanism more directly:

1. Which customers reject the service primarily because of price or waiting time?
2. What work is currently done rarely, poorly, or not at all?
3. If delivery cost falls by 50 percent, does the offer reach a new segment?
4. Can distribution and onboarding scale with production?
5. Does higher volume create a new human bottleneck elsewhere?
6. Will customers buy more units, or simply expect a lower price for the same volume?

The answers matter more than a generic claim that the market is large.

Do not confuse market growth with worker protection

Even when total demand expands, the gains may not flow to the same workers. New companies can capture the growth, senior workers may supervise agents while junior roles disappear, and wages can change even if employment rises.

Demand elasticity describes whether output expands enough to offset labor savings. It does not tell us who owns the productivity gain, which skills become valuable, or how work quality changes.

That is why the framework should guide strategy, not reassure people. For workers, the question is where judgment moves when routine production becomes cheap. For companies, it is whether they can turn lower costs into a broader market instead of competing only on headcount reduction.

AI makes supply abundant. The economic outcome depends on whether someone redesigns the product, price, and distribution so that demand can meet it.

Sources

• Alex Martin Richmond, The AI Jobs Transition Framework, OpenAI Economic Research, April 2026.

AI Jobs Transition series

• Previous: The work AI cannot own is often more important than the work it can do
• Next: The AI capability gap is an operating problem