Diffusion / bucket 4
Adaptation
- Stock
- Workflows, institutions, labour markets and regulation reorganised around the technology.
- Speed limit
- Complementary innovation, regulation, safety brakes and the pace of institutional change.
- Valve measured
- return arrow
Indicators
The wider economy
Official statistics, and figures companies publish from their own records, read beside the claims about what happens next. They carry no status of their own, and none measures AI alone; where one is also the reading behind an indicator on this page, its caption says so.
Electricity generated is the Federal Reserve's index of output from US power plants, each month's last twelve months against the twelve before, since a single month swings with the weather; real output is gross domestic product after inflation, each quarter against the same quarter a year before, from the Bureau of Economic Analysis. If building and running data centres were pulling power demand ahead of the rest of the economy, electricity would grow faster than output.
Derived (electricity_generation_yoy, real_gdp_yoy) from: FRED (official statistics). Federal Reserve Board, Bureau of Economic Analysis and Bureau of Labor Statistics series, retrieved from FRED.
The numbers
Gross domestic product per person after inflation, from the Bureau of Economic Analysis. Economists who expect AI to change the rate of growth itself look for this to climb well above its recent pace; those who expect an ordinary new technology expect it to stay near that pace for years.
Derived (real_gdp_per_capita_yoy) from: FRED (official statistics). Federal Reserve Board, Bureau of Economic Analysis and Bureau of Labor Statistics series, retrieved from FRED.
The numbers
Labour's share is the part of business output paid to workers as wages and benefits (the Bureau of Labor Statistics' index for nonfarm businesses); median real pay is the usual weekly earnings of the middle full-time worker, after inflation. If machines took over work and its income went to their owners, labour's share would fall while output kept rising. Labour's share is also the reading behind the labour-share indicator on this page.
Derived (bls_labor_share_yoy, real_weekly_earnings_yoy) from: BLS productivity and costs; total factor productivity; FRED (official statistics). U.S. Bureau of Labor Statistics, Productivity and Costs (PRS85006092/93, PRS85006173), Total Factor Productivity (MPU4910012). Federal Reserve Board, Bureau of Economic Analysis and Bureau of Labor Statistics series, retrieved from FRED.
The numbers
Payroll jobs in private education and health services and in leisure and hospitality, as a share of all US payroll jobs, from the Bureau of Labor Statistics; public schools and hospitals count as government jobs, so they sit only in the total. The two sectors stand in, roughly, for work built on people dealing with people, which some economists expect to be in more demand, and better paid, as machines take on other tasks: that view expects this share to rise. Many such jobs sit in other sectors, and not every job in these two is of that kind.
Derived (relational_work_share) from: FRED (official statistics). Federal Reserve Board, Bureau of Economic Analysis and Bureau of Labor Statistics series, retrieved from FRED.
The numbers
Average weekly hours paid to all employees in professional and business services, from the Bureau of Labor Statistics: a sector that holds much of the office work most exposed to AI, alongside administrative and waste services. A new tool can first make work more intense before it frees any time, and paid hours falling here would be one sign of time being freed; salaried staff are paid for their contracted hours, so time freed for them would not show.
Source: FRED (official statistics). Federal Reserve Board, Bureau of Economic Analysis and Bureau of Labor Statistics series, retrieved from FRED.
The numbers
| line | as of | value | from |
|---|---|---|---|
| average weekly hours | 2026-09-30 | 36.7 hours | the row itself |
| average weekly hours | 2026-08-31 | 36.8 hours | the row itself |
| average weekly hours | 2026-07-31 | 36.7 hours | the row itself |
| average weekly hours | 2026-06-30 | 36.7 hours | the row itself |
| average weekly hours | 2026-05-31 | 36.7 hours | the row itself |
| average weekly hours | 2026-04-30 | 36.7 hours | the row itself |
| average weekly hours | 2026-03-31 | 36.6 hours | the row itself |
| average weekly hours | 2026-02-28 | 36.6 hours | the row itself |
| average weekly hours | 2026-01-31 | 36.7 hours | the row itself |
| average weekly hours | 2025-12-31 | 36.5 hours | the row itself |
| average weekly hours | 2025-11-30 | 36.6 hours | the row itself |
| average weekly hours | 2025-10-31 | 36.6 hours | the row itself |
| average weekly hours | 2025-09-30 | 36.5 hours | the row itself |
| average weekly hours | 2025-08-31 | 36.5 hours | the row itself |
| average weekly hours | 2025-07-31 | 36.4 hours | the row itself |
| average weekly hours | 2025-06-30 | 36.4 hours | the row itself |
| average weekly hours | 2025-05-31 | 36.5 hours | the row itself |
| average weekly hours | 2025-04-30 | 36.5 hours | the row itself |
| average weekly hours | 2025-03-31 | 36.4 hours | the row itself |
| average weekly hours | 2025-02-28 | 36.3 hours | the row itself |
| average weekly hours | 2025-01-31 | 36.2 hours | the row itself |
| average weekly hours | 2024-12-31 | 36.3 hours | the row itself |
| average weekly hours | 2024-11-30 | 36.5 hours | the row itself |
| average weekly hours | 2024-10-31 | 36.3 hours | the row itself |
| average weekly hours | 2024-09-30 | 36.3 hours | the row itself |
| average weekly hours | 2024-08-31 | 36.4 hours | the row itself |
| average weekly hours | 2024-07-31 | 36.4 hours | the row itself |
| average weekly hours | 2024-06-30 | 36.4 hours | the row itself |
| average weekly hours | 2024-05-31 | 36.4 hours | the row itself |
| average weekly hours | 2024-04-30 | 36.4 hours | the row itself |
| average weekly hours | 2024-03-31 | 36.5 hours | the row itself |
| average weekly hours | 2024-02-29 | 36.5 hours | the row itself |
| average weekly hours | 2024-01-31 | 36.3 hours | the row itself |
| average weekly hours | 2023-12-31 | 36.5 hours | the row itself |
| average weekly hours | 2023-11-30 | 36.5 hours | the row itself |
| average weekly hours | 2023-10-31 | 36.4 hours | the row itself |
| average weekly hours | 2023-09-30 | 36.5 hours | the row itself |
| average weekly hours | 2023-08-31 | 36.4 hours | the row itself |
| average weekly hours | 2023-07-31 | 36.5 hours | the row itself |
| average weekly hours | 2023-06-30 | 36.5 hours | the row itself |
| average weekly hours | 2023-05-31 | 36.5 hours | the row itself |
| average weekly hours | 2023-04-30 | 36.5 hours | the row itself |
| average weekly hours | 2023-03-31 | 36.4 hours | the row itself |
| average weekly hours | 2023-02-28 | 36.6 hours | the row itself |
| average weekly hours | 2023-01-31 | 36.6 hours | the row itself |
| average weekly hours | 2022-12-31 | 36.4 hours | the row itself |
| average weekly hours | 2022-11-30 | 36.6 hours | the row itself |
| average weekly hours | 2022-10-31 | 36.7 hours | the row itself |
| average weekly hours | 2022-09-30 | 36.7 hours | the row itself |
| average weekly hours | 2022-08-31 | 36.6 hours | the row itself |
| average weekly hours | 2022-07-31 | 36.6 hours | the row itself |
| average weekly hours | 2022-06-30 | 36.6 hours | the row itself |
| average weekly hours | 2022-05-31 | 36.6 hours | the row itself |
| average weekly hours | 2022-04-30 | 36.7 hours | the row itself |
| average weekly hours | 2022-03-31 | 36.7 hours | the row itself |
| average weekly hours | 2022-02-28 | 36.7 hours | the row itself |
| average weekly hours | 2022-01-31 | 36.7 hours | the row itself |
Crosswalk to the capture lens
- Labour & consumers (same valve) — Productivity, wages, employment and consumer surplus.
- Deployment & application (leak from) — Deployment/FDE services capture part of the surplus that adaptation creates.
- Deployment & application (same valve) — Adaptation measured inside deployed products: regulated devices built on language models and paid robotaxi rides.
- Model (input to) — Backlash, safety brakes and paused runs are adaptation acting back on the model layer.
- Adopters (same valve) — Adaptation inside adopting organisations and sectors: regulated devices, state bills, call-centre uplift, reported executive impact.