The future of work isn't forecast.
It's built.
AIPE: AI Productivity & Economics. R²'s research program on the future of work and society with AI. It runs on two instruments: an atlas of what people believe is coming, and a model of what AI can actually deliver.
What is left to humans after AI?
A meta-analysis of what economists, philosophers, historians and storytellers believe is coming.
The Automation Dividend.
Where the working hours actually come back, industry by industry, once you price the review.
Belief is causal.
Capital allocates, policy drafts, founders build, and students choose their careers, all of it bending toward the future they expect. Expectations about AI don't just predict the future of work. They pour its foundations.
So the most consequential thing a commercialisation lab can study isn't a forecast. It's the belief that is already doing the building, and whether it points somewhere worth going.
Belief builds the world that proves it.
A map of what people believe is coming.
We read the whole field: economists, philosophers, historians, political economists, and the storytellers who rehearse our futures. Then we synthesised where they agree, and, more tellingly, where they split.
We don't resolve the disagreement. We chart it. And one image carries the whole economic argument: a rising tide on a ladder of work.
Full interactive document · opens in place
2025Today, the routine floor is already underwater. The work above it still pays.
What is left to humans is not a fact to discover. It is a space to claim.
Exposure ≠ dividend.
Everyone is measuring where AI touches work. Almost no one is measuring what a business actually keeps.
Exposure
Where AI could touch the work. Easy to count, easy to headline, and quietly misleading.
The dividend
What a business keeps after a human checks the work, the workforce comes along, and you count every task, not just the easy-to-automate ones.
Where the hours come back.
Belief maps the terrain. To act, you need the other half: across every industry and vertical, who actually nets the most from adopting AI?
We modelled it the honest way: pricing the human review, respecting the workforce, and counting every task instead of cherry-picking the automatable ones.
Where the hours come back. The share of working hours returned to a business after the agent runs, a human checks the work, and the workforce comes along. The ceiling is set by the review costs we refuse to assume away.
Unglamorous admin and decision work tucked inside hands-on industries. They look low-exposure, so whole-industry averages miss them, yet they quietly net the most.
Built on macroeconomic and survey data and stress-tested so the rankings hold. These are scenarios, not forecasts, and hands-on physical work remains the atlas's open clock.
Partners & investors: the full thesis →Two readings, one path.
Set the atlas beside the dividend and a direction appears. Automation goes where it pays, into the sleepers the dividend surfaces. Augmentation goes where it matters, because the atlas shows human value concentrating wherever machines can't follow.
And because belief is causal, building toward the humane reading is part of how it comes true. That is the whole point of a commercialisation lab.
What rises in scarcity as the tide climbs
The dial every R² engagement is calibrated against, set by evidence rather than hype.
The outcome isn't written yet.
AIPE is how R² reads the future of work, and helps choose it. A living foundation for a unified vision. Start with the map.