Why Your Pay Can Stand Still

The work has changed. The payslip has not.

A team can now finish its monthly reports several days earlier. The figures are checked. Customers like the clearer explanations. The manager thanks everyone for embracing the new tools. At the pay review, the answer is familiar: there is no room in the budget.

That answer might describe a real constraint. It might describe a choice. Often it contains some of each, and the worker has too little information to separate them.

What makes the situation so frustrating is that the gain seems visible. You used to complete this much work. Now you complete more. Surely some of the difference should come back to you.

I think it should. But “should” carries an argument about fairness. It is not a description of a payment system that already exists. A productivity improvement does not place a raise in a bank account. Someone must recognize a claim, agree to terms, and make the payment.

We need to understand that route if we want to improve wages. We also need to recognize its limit. Even a much stronger wage bargain leaves out people who are not employed by the businesses generating the gains. Universal ownership addresses that wider problem. Everyone should hold a meaningful claim on productive capital, including people whose bargaining position at work is weak and people outside paid work altogether.

To see why, start with the money already on the table.

More output is only the beginning

Suppose a small design firm can now prepare more proposals with the same staff hours. That may be a valuable improvement. It has not yet established that customers want more proposals, that they will pay for them, or that the firm earns more after its added costs.

If the team was turning away good work, the extra capacity may lead to more sales. If demand was thin, the firm may simply finish its existing assignments earlier. Staff could use that time to improve the work, learn, or go home. The owner could also fill it with new demands that produce little value. Capacity creates a choice before it creates an income.

Now imagine that competitors obtain similar tools. They may lower their prices to attract customers. The firm could produce more designs while receiving less money for each. Customers would gain. The firm’s profits might rise, fall, or stay much the same, depending on how much it sells and what it costs to deliver.

This matters because “the company is more productive” and “the company has more cash available” are different claims. A pay discussion becomes stronger when it identifies the actual gain. Perhaps revenue rose. Perhaps an outside expense disappeared. Perhaps overtime fell. Perhaps the firm retained a customer it was likely to lose. Each claim requires its own evidence.

The software also has a price. There may be costs for integration, training, better data, or additional review. Some are temporary. Others continue. An owner who paid for these things has a legitimate interest in earning a return. So does a worker whose knowledge and effort made the system useful. Recognizing one contribution does not erase the other.

The trouble begins when one party’s contribution becomes an enforceable claim and the other’s becomes a thank-you.

The owner may already hold the right to profits through shares in the company. The vendor has a contract for its fees. The worker has a contract for pay, which may contain no link to the improvement. Unless that agreement changes, the extra contribution can leave the worker’s cash income exactly where it was.

No villain is required for this outcome. The existing rules can produce it quite efficiently. That is precisely why changing the rules matters more than asking every owner to have a generous afternoon.

What the pay records show

Early labor-market evidence makes the gap concrete. Anders Humlum and Emilie Vestergaard linked Danish adoption surveys to administrative records. Their March 2026 working-paper revision found no average effect on earnings or recorded hours within its study design, ruling out effects larger than 2 percent two years after ChatGPT’s launch. Work had changed through new tasks and reorganization even while those averages barely moved.¹

A U.S. working paper revised in August 2026 found a different warning. In ADP payroll data through June 2026, employment among workers aged twenty-two to twenty-five in AI-exposed occupations was 19 percent below the path implied by growth among less-exposed peers. The gap arose mainly through weaker hiring. The authors found no widespread displacement and described the results as descriptive, with some trends beginning before generative AI. The gap cannot simply be labeled jobs lost because of AI.²

These studies follow different people in different settings. Together they give us reason to watch actual pay and entry routes closely, rather than infer either from faster task completion.

Be precise about the gain you can buy with

Economists use labor productivity to describe output per hour worked. It measures the result of a production process, including the tools, organization, and other resources used with labor. It is not a moral score for the person working. The U.S. Bureau of Labor Statistics also distinguishes pay from total compensation, which includes benefits, and distinguishes money amounts from their purchasing power after price changes.³

These distinctions help us read a claim without turning the discussion into a contest of charts.

Take a monthly wage of $3,000 in a simple example. A 3 percent raise takes it to $3,090. If the prices relevant to the comparison rise by 5 percent, the new wage buys about 1.9 percent less than the old one did. The worker received more dollars and lost purchasing power. Both statements are true.

A price index describes a basket of purchases, not the exact bills of every household. A person facing a sharp rent increase may feel less benefit than someone whose housing cost barely changed. A lower price for software is welcome. It does not cancel a higher price for the place where you sleep.

This is why I want the AI dividend judged in ordinary life. Can people afford essentials? Do they have more secure income? Can they take time to care for someone without their finances collapsing? A production statistic helps us understand what may be possible. The household tells us whether the possibility arrived.

We also need to compare like with like. An average can rise while many people’s pay stands still. Benefits can become more costly to an employer without leaving more cash in a worker’s pocket. Output prices and consumer prices can move differently. The BLS has shown why the choice of price adjustment affects comparisons between productivity and compensation.⁴

None of these measurement issues makes a worker’s complaint disappear. They tell us which complaint we are making. “The typical worker has received little of the gain” requires a different measure from “the firm’s total labor costs have risen.” A company can answer the second and leave the first untouched.

There is a small power in refusing that substitution. Ask for the measure that addresses the claim. Keep the units visible. Annual salary, hourly pay, total compensation, and take-home income are related, but a change in one should not be smuggled in as an answer about all the others.

What happens if you say no

Two workers can make equally strong contributions and obtain different raises because their alternatives differ.

One has a credible offer from another employer. The other lives near the only workplace that offers hours compatible with caring for a parent. The second person’s needs do not reduce the value of the work. They may reduce the practical ability to walk away.

An outside option is what you can realistically do if you reject the current terms. The word realistically does a lot of work. A vacancy hundreds of miles away is not automatically an option. A higher hourly rate attached to unreliable hours may not be an improvement. A person cannot pay next month’s bills with an employer’s theoretical willingness to interview.

Labor-market power can exist even where there are several employers. A 2022 U.S. Treasury report explains how the time, uncertainty, and costs of finding and changing jobs can limit wage competition. Workers do not instantly move among perfectly interchangeable positions. Employers can have room to set pay below what a model of frictionless competition would suggest.⁵

Put that mechanism back inside our design firm. The worker who improved its process may be valuable. Management may nevertheless believe she will stay at the current salary because leaving is difficult. Her contribution and her power to claim payment have come apart.

AI can change this balance in either direction. If it helps her offer a service independently, reach new customers, or qualify for more roles, her options may improve. If similar tools let many more people supply what she sells, customers and employers may gain alternatives. The same technology can help her do better work while making the market for that work more crowded.

The relevant question is therefore more specific than “Have I learned AI?” What can you now do that someone needs, and what credible choices does that ability give you? A tool subscription is access to capacity. It is not, by itself, a durable claim on the income that capacity may generate.

This is not an instruction to keep one’s knowledge secret or make colleagues dependent. A workplace in which everyone hides improvements becomes worse for everyone who has to use it. It is an argument for agreeing how contributions will be recognized before cooperation turns into an unpriced obligation.

Collective action can also change the discussion. Workers who raise a shared concern together may be able to negotiate terms that an isolated employee cannot obtain. The available forms and legal protections vary by place. The basic purpose is clear enough: make the people carrying the change a party to its terms.

An individual can prepare evidence, develop skills, and seek alternatives. Those actions are useful. They should not become a story in which every low wage proves that someone failed to negotiate with enough confidence. The person with the least room to refuse is often already doing a great deal to hold life together.

The promise of cheaper things

There is a serious optimistic reply to all this. Even if wages do not rise much, AI may make many goods and services cheaper. People could become better off through what they can buy.

I want that gain. A less costly service can reach people who were previously priced out. Better tools can help a small business compete. A household does not need to receive a dividend payment to benefit from a lower bill.

But this route depends on the saving reaching the customer. If competition is weak or switching is hard, a seller may retain more of the gain. If the cheaper service depends on an expensive bottleneck elsewhere, the final bill may not fall much. The path from capability to affordability needs the same close attention as the path from capability to pay.

There is also a distribution problem within the household. Suppose a worker saves a modest amount on several services while losing a large share of paid hours. Those savings are real. They do not necessarily cover the income loss. A national total can look favorable while a particular family is in trouble.

And consider time. A system may reduce the hours needed to deliver a service, yet the employee’s schedule stays long because the target rises. The customer gains speed. The owner gains revenue. The worker gains more work. That is one possible bargain; it is not a technical necessity.

A better bargain can share benefits through shorter paid working time, higher compensation, training, or ownership. Which combination makes sense depends on the work and the people involved. A parent might value a predictable afternoon more than a small bonus. A person behind on bills may need cash. Workers should have a voice in choosing among benefits that will shape their lives.

Still, I do not want every improvement in a person’s life to require a fresh appeal to an employer. That keeps the source of security too narrow.

Why a wage bargain cannot include everyone

Imagine the design firm becomes extraordinarily successful. Its staff negotiate excellent pay and a fair profit-sharing plan. That is an achievement worth defending.

Now ask about the person who left work to provide unpaid care. Ask about the young adult who cannot find an entry role, the person whose health limits paid work, or the worker employed in a business that gains little from the technology. None receives a share merely because this particular firm’s bargain is good.

Employee ownership can improve the relationship inside a company. It cannot, on its own, put every person in the capital-owning class. A universal claim needs a route that follows the person beyond one job and reaches people without a job.

This is the point at which the discussion must become more ambitious. We should want workers to bargain well. We should also want everyone to hold real rights to a share of productive wealth. The second goal does not have to wait until the first route has failed completely.

Ownership means more than being sent an encouraging certificate. It requires clear rights to income or value, rules about control, and protection against having those rights quietly diluted or withheld. Later chapters examine how such claims could be built and governed. Here the essential point is that they create another route between expanding production and ordinary security.

Capital income carries risk. Profits can fall. Assets can lose value. Broad ownership therefore cannot be an excuse to strip away wages, public services, or an adequate social floor. Nor should people be required to gamble their security on the fortunes of the single company that employs them. A serious design spreads risk and makes the claim understandable.

This demand is compatible with wanting technology to advance rapidly. Indeed, it changes what advancement feels like. If a more capable system can raise the value of a broadly held productive base, people have a direct reason to welcome its success, alongside their other concerns. A share will not answer every question about safety or the use of power. It answers a central question about who benefits.

Ask where the answer is missing

At the next pay review, a worker can make a more exact argument than “AI has made me productive, so pay me more.” She can identify the completed work, the change in quality or cost, and the responsibility she now carries. She can ask how those changes enter the firm’s pay decisions. If the answer is “they don’t,” that is useful information about the bargain.

A team can ask what happened to saved time. Did it reduce overtime? Did it become more orders? Did a vendor absorb the saving? Which part is measured and which part is assumed? The answers may reveal room for a better agreement, or a need for better evidence.

But the public question is larger than any review meeting. Why should a person have to prove continuing scarcity in the labor market to share in a society’s increasing capacity to produce? Why should the ownership route remain easy to inherit and difficult to enter?

We can improve the wage bargain and broaden the ownership bargain together. The first concerns fair terms for work. The second concerns membership in the prosperity that work, technology, and accumulated public knowledge make possible.

The payslip may tell you that your pay has stood still. It cannot tell you that your claim must stand still with it.

Sources and notes