The Least Dishonest Answer

My instincts run against this argument.

Adam Smith first. Then Hayek, Friedman, and Sowell. I trust markets over management, prices over plans, entrepreneurs over bureaucrats. I believe self-interest and competition, ordered through exchange, law, trust, and reputation, can create more good than centralized management ever will.

But reality gets a vote.

And the signal is no longer theoretical.

Cloudflare said it would cut about 20% of its workforce, more than 1,100 jobs, as it restructures around an "agentic AI-first operating model." ClickUp cut 22% of its workforce while rebuilding around AI and what its CEO called "100x output." Block has also been reported as cutting more than 4,000 jobs in an AI-driven restructuring, though I would treat that one with more caution than the first two. The safe conclusion is enough: serious companies are starting to reorganize around the belief that they can produce more with fewer people. (Reuters)

If you are not seeing the signal now, you are not paying attention.

The firm sees efficiency. The economy sees productivity. The worker sees replacement.

All three can be true at the same time.

A company can use AI, reduce headcount, increase margins, serve customers faster, and become more valuable. On paper, that is progress. It is also a man getting called into a meeting and realizing the skill set that carried his mortgage, his family, his status, and his sense of usefulness has just been repriced by a system that does not need a salary, healthcare, PTO, payroll taxes, HR, or a performance review.

The long run may be fine.

The long run is not where people buy groceries.

That is why Keynes is back in the room.

Not as the architect of the future. As the emergency doctor when the system is bleeding.

Smith still gives us the starting point: self-interest and competition can create public good when exchange is ordered by trust, law, and markets. Hayek tells us why prices and wages matter, because they carry information no planner can hold. Friedman warns that government intervention comes with distortions, bad incentives, and bills that usually arrive late. Sowell drags every theory back to the person inside the tradeoff: compared to what, at what cost, with what incentives, and who pays?

That is the frame I trust.

Which is why this is uncomfortable.

AI may create a Keynesian employment crisis inside a world where Smith, Hayek, Friedman, and Sowell remain more right in the long run.

The mechanism is not mysterious. Firms respond to incentives. Anyone who has carried payroll knows headcount is never just a line item. It is responsibility, drag, trust, risk, management, culture, and care. A person is not a unit of output. But inside a firm, a person is also cost and complexity.

AI infrastructure is expensive too. Compute, model access, integration, security, and technical talent all cost real money. But those costs scale differently. Once the system works, the marginal cost of more output can fall hard.

So the firm starts asking a different question.

The old question was: how many people do we need to produce this output?

The new question is: how much of this organization can be rebuilt around fewer people and more powerful systems?

That is not just job replacement. It is organizational replacement. The company that used to add people to scale can now add compute, agents, software, and automation. A department that used to grow with revenue can stay smaller. A manager who used to manage people can manage systems. Work that used to justify a role becomes a workflow.

One company doing that is operational efficiency.

Thousands doing it at once is macroeconomics.

This is where the free-market answer starts to feel both true and insufficient. The market should discover what comes next. Prices should move. Wages should adjust. Entrepreneurs should build new firms. Capital should flow to new uses. Labor should move toward new value.

But a laid-off 48-year-old project manager with a mortgage does not live inside an elegant equilibrium. He lives inside next month's cash flow. He has a family, a car payment, a resume, and the private humiliation of wondering whether the economy still needs what he spent twenty years becoming.

This is where ideology gets thin.

If you say the answer is to stop firms from automating, you are wrong about the future. You cannot freeze the economy in its current shape and call that compassion. If government starts telling companies they cannot use AI to replace labor, cannot reduce headcount beyond some approved threshold, or must preserve roles after the work no longer requires them, then the state has started deciding how firms are allowed to produce.

That is Hayek's nightmare with better software.

It protects yesterday's job titles by taxing tomorrow's abundance. It favors incumbents. It punishes productivity. It turns firms into employment warehouses. It makes politics the allocator of work.

But if you say the market will absorb every shock without a floor underneath people, you may be wrong about the present.

Software can reprice a workflow in months. A person may need years to retrain, relocate, rebuild, or find the next place where his labor matters. Maybe the labor market recovers. Maybe it does not recover in any form we would recognize. Nobody knows that yet, and this essay cannot carry the whole argument. What is clear is that firms move at the speed of incentives. Families move at the speed of reality.

That gap is the Keynesian moment.

And there is no clean lever.

A token tax sounds clean until it punishes productive AI use, pushes usage offshore, advantages incumbents, or gives the state a new claim over cognition itself. An automation tax sounds fair until government has to decide what counts as automation and every lobbyist in the country starts carving exemptions. Layoff restrictions sound humane until companies become employment warehouses and bureaucrats start deciding how firms are allowed to produce.

Then there is Universal Basic Income, or UBI.

I do not like UBI as a theory of human purpose. Income is not meaning. The state cannot direct deposit dignity into a life. A man without work, burden, responsibility, and usefulness is not automatically free because his bills are paid. He may become softer, more distracted, more hollow.

But if AI breaks the labor market faster than new work appears, some kind of income floor may be the least destructive intervention.

That is the painful concession.

I would rather give people cash than give government permanent control over how firms are allowed to produce.

Cash support does not tell companies which jobs must exist. It does not require bureaucrats to approve productivity. It does not freeze the economy in its current shape. It lets the market keep discovering while trying to keep households from falling through the floor during the transition.

That does not make it clean. It is expensive. It can create dependency. It can weaken work incentives. It can become politically impossible to unwind. It can feed inflation if designed badly. Friedman would be screaming warnings from the grave, and he would be right.

But compared to what?

Compared to freezing firms in old labor structures, I would rather risk the distortion of income support. Compared to letting millions of people absorb a software-speed displacement shock with no floor, I would rather admit the emergency is real. Compared to handing the state permanent power over production, I would rather preserve market discovery and cushion the human beings getting crushed by the transition.

The people who want to stop automation are wrong about the future.

The people who think markets will absorb every shock without a social floor may be wrong about the present.

That is the narrow road.

Use Keynes for the shock.

Use markets for discovery.

Do not let either side lie about the cost.

The market may discover what comes next. It will not do it gently.