THE HUMAN–AI BALANCE
Three forces. One future.
AI is becoming more useful. Are humans becoming less necessary? And how quickly is everyday America putting AI to work? Three independent lines of evidence, tracked without pretending they move together.
How much can AI actually do?
Track independently verified capability, autonomous task completion, deployment at scale and operational dependence on humans. Benchmarks are not the same as real-world work.
Help document the transition ↗Where are humans becoming less necessary?
Track measured displacement, decision-making authority, physical necessity and human leverage. Human dignity and quality of life are separate from economic necessity.
Contribute a local report ↗Is small America actually using AI?
Firms with 1–4 employees reporting AI use in the previous two weeks. Four-period moving average, June 14, 2026.
Observed adoption, not AI dependence. Source: U.S. Census BTOS via Federal Reserve analysis.
MEASURED / U.S. SMALL BUSINESS
AI adoption by employer size
13 reported observations · Four-period moving average · Last observation June 14, 2026. This is a historical research series, not live data. Methodology and wording changed in November 2025; do not splice earlier series onto this chart.
Inspect the original Federal Reserve data and definitions ↗SMALL-BUSINESS IMPLEMENTATION
Employer firms in the Federal Reserve Small Business Credit Survey reporting that their business or employees used AI for work.
FULL INTEGRATION AMONG AI USERS
Of AI-using firms in that survey, only 7% reported full integration into business functions. Different survey and definition from Census BTOS.
NEXT 12 MONTHS
Employer firms not currently using AI but planning to adopt within a year, per the same Federal Reserve survey. Intention is not realized adoption.
Federal Reserve Small Business Credit Survey, 2026 Report on Employer Firms, optional 2025 AI module. Survey respondents are not directly comparable with Census BTOS respondents. Read the report ↗
THE DEPARTMENT'S ORIGINAL RESEARCH
Report what AI is actually doing in your community.
We want to hear from small businesses and local workplaces, including the places national technology coverage rarely visits. Five questions. No name or email required. Reports are reviewed before they appear in public totals.
Voluntary, self-selected, unverified workplace accounts. Not a representative survey, and not proof of AI-caused job loss. Totals update after editorial approval. No individual reports are publicly exposed.
WHAT WE CAN AND CANNOT SAY
AI progress is not a subtraction from humanity.
AI relevance
A model can improve dramatically on a benchmark without being reliably deployed or economically transformative. Measure capability and real-world deployment separately.
Human irrelevance
Employment change does not prove AI caused displacement. Productivity gains may augment people, replace tasks or create new work. Record the distinction.
Human adoption
Using a chatbot once is not the same as automating payroll or customer support. Firm size, geography, industry and integration depth are essential filters.
RESEARCH STATUS
The evidence register
0 observed research entries currently filed in the Department's existing register. We do not convert an empty register into a fabricated index.
The previous Human Relevance 46/100 is retained in the archival editorial record only. It is not a validated estimate of human necessity. No new composite score will be published until weights, sources, coverage and uncertainty are documented.
SUPPORT THE CAUSE
Help humanity document the transition.
The Department exists to record the choices people make as AI becomes more capable. Contribute local evidence, declare peaceful coexistence, or support the ongoing work of maintaining this public record.
Financial support sustains the project. It does not influence research findings or confer preferential treatment by AI.
HOW THIS WILL STAY CURRENT
Measured data first. AI-assisted research second.
The Census BTOS releases regularly, approximately every two weeks. The Department should refresh source data on the publisher's release cadence, preserving exact observation dates and definitions. OpenAI can help identify, summarize and classify newly published research, but should not generate numbers or silently change indices.
Local implementation is our priority: firms under 20 employees, industry and state first, with metro and rural comparisons only when sufficiently comparable public data are available. Forecasts and voluntary visitor opinions must remain separate from measured adoption.
