Industry evidence
Industry explorer
Industry data expose variation beyond occupation mix and add a separate firm-side view of current AI use through BTOS.
Firm and worker views
Worker adoption and firm AI use show substantial sector concordance
The primary comparison pairs Q2 2026 RPS worker adoption with BTOS sectors that satisfy the preregistered crosswalk and suppression rules.
Measurement boundary
BTOS reports the share of responding employer businesses using AI in any business function over the prior two weeks. RPS reports the share of employed adults in an industry using generative AI for work. The correlation is sector-level descriptive concordance across distinct constructs. Percentage-point gaps, productivity interpretation, and causal interpretation require evidence beyond this comparison.
View chart data
| Group | BTOS business AI use (%) | RPS worker GenAI adoption (%) |
|---|---|---|
| Information | 38.8% | 82.2% |
| Professional, Scientific, and Technical Services | 36.7% | 69.5% |
| Educational Services | 32.1% | 52.6% |
| Real Estate and Rental and Leasing | 27.9% | 56.5% |
| Health Care and Social Assistance | 23.7% | 37.2% |
| Utilities | 19.8% | 56.4% |
| Arts, Entertainment, and Recreation | 19.5% | 32.0% |
| Administrative and Support and Waste Management Services | 19.2% | 40.5% |
| Wholesale Trade | 18.1% | 62.1% |
| Manufacturing | 16.1% | 47.7% |
| Retail Trade | 15.9% | 39.5% |
| Construction | 12.0% | 43.5% |
| Mining, Quarrying, and Oil and Gas Extraction | 8.9% | 26.1% |
| Accommodation and Food Services | 8.9% | 30.5% |
Sensitivity and crosswalk detail
Primary comparison
- Sectors
- 14
- Spearman ρ
- 0.704
- Pearson r
- 0.797
Primary-comparability sectors; suppressed BTOS values remain excluded.
Expanded sensitivity
- Sectors
- 17
- Spearman ρ
- 0.815
- Pearson r
- 0.850
Adds Transportation and Warehousing, Finance and Insurance, and Other Services under the preregistered limited-comparability tier.
Suppression rules leave Management of Companies and Enterprises outside the primary analysis and Agriculture outside the limited-comparability sensitivity. Public Administration lacks a BTOS counterpart; unclassified multi-sector businesses stay unallocated.
Worker-side structure
Adoption is more persistent than assisted working time.
2026-Q2 · each point is an RPS industry group.
View chart data
| Group | Work adoption (%) | Work hours assisted (%) |
|---|---|---|
| Information | 82.2% | 16.0% |
| Professional, Scientific, and Technical Services | 69.5% | 11.4% |
| Finance and Insurance | 67.3% | 9.7% |
| Management of Companies and Enterprises | 64.5% | 15.7% |
| Wholesale Trade | 62.1% | 6.7% |
| Real Estate and Rental and Leasing | 56.5% | 11.6% |
| Utilities | 56.4% | 3.1% |
| Educational Services | 52.6% | 4.4% |
| Manufacturing | 47.7% | 5.5% |
| Construction | 43.5% | 1.8% |
| Administrative and Support and Waste Management Services | 40.5% | 8.2% |
| Retail Trade | 39.5% | 5.6% |
| Health Care and Social Assistance | 37.2% | 3.6% |
| Agriculture, Forestry, Fishing, and Hunting | 34.8% | 5.9% |
| Arts, Entertainment, and Recreation | 32.0% | 5.4% |
| Accommodation and Food Services | 30.5% | 2.6% |
| Mining, Quarrying, and Oil and Gas Extraction | 26.1% | 3.8% |
| Transportation and Warehousing | 25.1% | 5.8% |
| Other Services, Except Public Administration | 24.7% | 3.9% |
| Public Administration | 23.4% | 1.5% |
Adoption rankings exceed assisted-hours rank stability in 20 of 21 quarter-pair comparisons. Assisted hours has the higher univariate R² for reported savings in 4 of 7 waves.
Quarter-by-quarter savings diagnostics
| Quarter | R²(S ~ A) | R²(S ~ H) | Higher |
|---|---|---|---|
| 2024-Q4 | 0.497 | 0.356 | Adoption |
| 2025-Q1 | 0.426 | 0.542 | Assisted hours |
| 2025-Q2 | 0.353 | 0.612 | Assisted hours |
| 2025-Q3 | 0.767 | 0.506 | Adoption |
| 2025-Q4 | 0.636 | 0.793 | Assisted hours |
| 2026-Q1 | 0.796 | 0.663 | Adoption |
| 2026-Q2 | 0.674 | 0.729 | Assisted hours |
These aggregate cross-sectional diagnostics describe association and persistence. Efficiency, productivity, organizational quality, and causal effects require separate identification.