Follow one paycheck. Bob earned about $120,000 a year. To his employer, that was simply the cost of getting his work done. When AI takes over Bob's role, that cost doesn't vanish—the work still gets done, for customers who still want it, but now at a fraction of the price. The roughly $105,000 the company no longer pays Bob doesn't disappear either: it becomes profit.
New profit owes new tax. And the cheapest way to erase that tax bill is to buy Human Credits—including ones earned by a provider now serving Bob. So the company (or another corporation with a tax bill to offset) funds the provider that delivers Bob's housing, healthcare, and food. A portion of what used to be Bob's paycheck flows back to him—no longer as wages, but as guaranteed security. The employer still comes out well ahead, and serving Bob is itself a new kind of job.
That is the whole engine, at the scale of a single worker. Now multiply Bob across the economy. US labor compensation runs about $16 trillion a year; as AI displaces a growing share of it, that money becomes corporate profit, and the Human Credit captures a fair slice of the windfall by taxing it at the existing rate and closing the loopholes that let it escape, with legacy entitlement spending redirected as the system matures. A bounded, temporary deficit bridges the early years—a bridge, not a permanent program. We do not pretend that away: we already run large deficits for far less, every alternative (doing nothing, or a permanent universal cash program) is worse, and this one closes itself as the levers below engage.
There is a deeper reason business should want this. The paycheck above quietly assumes Bob's old company still has customers. But if AI strips income from displaced workers everywhere, those customers vanish—and the firms automating the work lose the very market they sell into. By keeping Bob a consumer, the Human Credit keeps that market alive. It is, in effect, demand insurance.
These inputs are illustrative and deliberately ranged—enough to show the mechanism is workable, not a final score. A rigorous model is exactly what we invite OMB and think tanks to build.
| Assumption | Illustrative value |
|---|---|
| US aggregate labor compensation | ~$16T / year |
| Population / employed workers | 340M / 162M |
| Per-person annual support | Budget ~$30k · Generous ~$50k |
| Corporate windfall from displacement | ≈ displacement share × $16T (saved labor → profit) |
| Effective corporate tax rate (credits drawn against) | ~13% today → 21% as avoidance closes → ~25% backstop |
| Entitlement redirection | ~$4.17T, phasing in only at higher displacement (end game) |
| Re-employment into the human-services economy | ~18–20% of displaced workers |
| Cost to deliver each outcome | Declines over time (disinflation + tapering provider margins) |
A note on how to read these. Each lever is additive, and in combination they are powerful—plausibly more than the sum of their parts. Their individual magnitudes, however, are genuinely hard to predict; pinning down precise contributions is the work of rigorous modeling. That uncertainty is exactly why the design leans on many levers rather than betting on any one—and why a tenth exists as a hard backstop.
The first nine engage progressively as displacement rises (charted later in this section); the tenth is held in reserve, used only if the combined effect falls short.
Because funding and need both scale with AI displacement, the economics are best seen as a continuum, not a single snapshot. Early on, need outruns capacity and the bounded transition deficit bridges the gap. As avoidance closes, disinflation compounds, and entitlements redirect, the system reaches break-even and then runs a surplus—the AI Dividend—at high automation. The figure below decomposes it: without the levers, the deficit only deepens with displacement; with them engaged, it becomes a bounded bridge that turns to surplus. (The table beneath gives the same net position by support tier.)
| AI displacement | Budget tier (~$30k) | Generous tier ($50k) |
|---|---|---|
| 10% | −$0.2T | −$0.8T |
| 30% (transition valley) | −$0.6T | −$2.3T |
| 50% | +$1.8T | −$0.5T |
| 70% | +$2.9T | ≈$0 |
| 90% | +$3.8T | +$0.5T |
The figures here are illustrative and deliberately ranged—a workable model meant to demonstrate due diligence and feasibility, not a final score. Rigorous economic modeling by qualified economists is needed to validate them; we invite economists, think tanks, and policy researchers to apply rigor to these assumptions.
The continuum has a turning point: the deficit is deepest at moderate displacement (around 30% in this illustration), then narrows toward break-even and surplus. The reason is that its two underlying curves scale differently. Support cost rises roughly linearly with displacement—each additional slice of displaced workers adds about the same need. Funding capacity, by contrast, is convex—slow at first, then accelerating—because the most powerful levers ramp up or switch on only as displacement deepens.
Three forces are weak before the turn and strong after it:
So the deficit is deepest where need has peaked but the structural levers have not yet fully engaged; past that point, funding and falling costs overtake the rising need. The precise location of the turn is an artifact of the assumptions above—how fast avoidance closes, when entitlements redirect, the rate of disinflation—not a prediction. What is robust is the shape: a gap that widens, then closes. The figure below shows how each lever is assumed to engage as displacement rises.