Evidence and method
AI language, spending and headcount: different evidence
AI mentions, capital expenditure and employee counts measure different things. A company can discuss AI without disclosing an AI-only investment amount, and a headcount change alone does not establish an AI-caused job cut.
- 1. Filing languageWhat a company says.
- 2. Financial disclosureWhat an amount covers.
- 3. Workforce disclosureWho is counted, and when.
Read the scope before the number
Say versus Spend uses original company filings and same-fiscal-year financial context. Total capital expenditure includes many purposes. It cannot be labeled AI expenditure unless the source identifies an AI-only amount. Employee counts are company-wide where disclosed; contractors, acquisitions and reporting dates can complicate comparison.
Separate adoption from attribution
Census business AI-use shares are survey measures, distinct from filing mentions and capex. IC 100 is a fixed reading cohort, not the whole economy. Passage context labels are model estimates. Open the original passage and the eligibility rules before comparing companies. Evidence that AI is used is weaker than evidence that it caused a particular workforce outcome.
Limitations
These tools do not supply an investment ranking or causal estimate of AI job loss. Compare like periods and definitions; an unavailable value is not zero.
Sources and verification
Follow the original source for its scope, definitions and updates. Review date refers to this companion, not a new data release.
- SEC · Accessing EDGAR data
Original filings and structured company facts; compare filing periods and accounting context. - U.S. Census Bureau · BTOS
Survey coverage, timing and published business cuts. - Core Loop Press · sources and method
Actual cohort, passage classifications and financial definitions.