Sam Altman got the diagnosis right. But the cure won’t come from politics — it will come from the boardroom.
Last week, Sam Altman published a 13-page policy paper. The title: Industrial Policy for the Intelligence Age: Ideas to Keep People First. The message is as simple as it is consequential: AI is shifting the distribution of value from labor to capital. And if no one intervenes, it will destabilize the foundations of modern economies.
Altman calls for robot taxes, a public wealth fund, and a fundamental reorientation of the tax base away from wage income toward capital returns. It sounds like New Deal 2.0 — and that’s exactly what he means. The comparison to the Progressive Era and the 1930s is no rhetorical accident.
Altman has named the macro problem clearly. But he’s addressing it at the wrong level. The decision that actually shifts the balance between labor and capital is not made by policymakers. It is made every day inside companies.
What Altman is really describing
The core argument in Altman’s paper is macroeconomically precise: AI dramatically increases capital productivity. At the same time, the share of value creation flowing to workers as wages is declining. This is no longer a forecast — it is a measurable shift.
Policymakers could rebalance the tax base by increasing reliance on capital-based revenues — such as higher taxes on capital gains at the top, corporate income, or targeted measures on sustained AI-driven returns — and by exploring new approaches such as taxes related to automated labor. — From the OpenAI paper, April 2026
What Altman describes is the classic mechanism of factor substitution: when capital (here: AI) becomes cheaper, companies replace labor with capital. That increases margins. But it also — aggregated across all companies — reduces the labor share of the economy.
The effect is not new. It is as old as industrialization. What is new is the speed and reach. AI makes substitution possible in knowledge work — precisely the domain where qualification was previously thought to be a shield against automation.
“AI will break the economy as we know it. The question is whether we rebuild it better.” — Sam Altman, Axios interview, April 2026
The macro problem: a trap with two locks
The macroeconomic problem has its own logic — one that remains invisible at the company level. Every individual firm acts rationally: cut costs, increase margins, protect competitiveness. That is efficiency. The problem emerges when everyone does the same thing simultaneously.
When wage income falls, purchasing power falls. When purchasing power falls, demand weakens. When demand weakens, higher margins stop helping. This is not theory — it is the classic Keynesian demand trap, this time triggered by technological disruption rather than financial crisis.
Altman’s answer: redistribution through taxation. A public wealth fund, seeded by AI companies, gives every citizen a stake in productivity growth. Demand is preserved even as the labor share declines.
Economically coherent. Politically? Not in this decade. The Biden administration already failed to tax unrealized capital gains. The current administration has other priorities. And Europe — the most regulation-friendly terrain — is fighting for competitiveness, not redistribution.
The company level: two paths, one fork
Because the political path is blocked, the real decision falls where it has always fallen: inside companies. And there, the choice is not binary — but the direction is.
Path 1 — Efficiency: Automate → Reduce headcount → Increase margins. Works short-term. The company isn’t becoming more productive — it’s becoming smaller.
Path 2 — Growth: Augment → Upskill people → Increase output. Builds capacity. Requires longer time horizons. Creates durable competitive advantage.
Path 1 is the default. It is easier to justify (cost reduction is measurable), faster to execute, and simpler to explain in an investor report. It is also the path that, at macro scale, springs the trap Altman describes.
Path 2 demands a different starting question. Not: what can we automate? But: what can we do with AI that we could not do before?
The blind spot: productivity is being measured wrong
Many companies believe they are increasing productivity. In reality, they are shrinking in a controlled way. The difference is critical — and invisible in most management dashboards.
Classic efficiency metrics measure output per employee. When headcount falls and output stays flat, the metric improves. That looks like productivity. It is actually just a smaller system producing the same amount as before.
Real productivity growth means the system produces more — not that fewer people produce the same. The distinction sounds technical. The strategic consequences are fundamental.
Companies using AI for efficiency are buying time. Companies using AI for growth are buying future.
What this means for leadership
Altman’s paper suggests companies should use AI efficiency gains to retain workers, retrain them, and invest in four-day workweeks without pay cuts. For a tech CEO, this is an unusually conservative argument. It is also not an altruistic one: it is about not eroding the demand base that his own products depend on.
At the company level, this translates into three concrete leadership decisions:
- Every automation decision needs a growth hypothesis. Any decision to automate a process should include an explicit answer to: what do the freed-up capacities do instead? If the answer is ‘nothing’, it is cost reduction — not productivity growth.
- Upskilling is not an HR initiative — it is capital allocation. Employees who can work with AI tools are more productive than employees replaced by AI tools. The ROI of Path 2 is harder to measure but structurally more durable.
- The next generation of management systems will not be built around cutting costs — they will be built around making the right decisions and executing them. Which opportunities can we now capture that were out of reach before? Which customer problems can we finally solve? These are the questions that frame AI as a strategic instrument rather than a cost lever. The companies that answer them well will not just protect their margins. They will protect the market they sell into.
Altman is right — but he is waiting for the wrong actor
Sam Altman describes a real problem with real urgency. His paper is not a PR exercise — it is a serious attempt to start a debate that politics has not yet engaged with honestly.
But the solution he proposes assumes governments will act quickly and coherently. That is — especially in the political climate he himself helps shape — a heroic assumption.
The actual shift between labor and capital does not happen through legislation. It happens through thousands of individual decisions inside companies — every day, in every budget conversation, in every automation project.
That is not bad news. It means the solution lives there too.
You can automate work. You can’t automate demand. The companies that understand this will make the right decisions — and build the systems to execute them.
Sources: OpenAI, “Industrial Policy for the Intelligence Age: Ideas to Keep People First” (April 2026) · Axios, Behind the Curtain: Sam’s Superintelligence New Deal (April 2026) · Fortune (April 2026) · IMF World Economic Outlook 2017 · OECD Compendium of Productivity Indicators 2025 · Karabarbounis & Neiman, The Global Decline of the Labour Share, QJE 2014