๐ช๐ฎ๐๐ต๐ถ๐ป๐ด๐๐ผ๐ป ๐ถ๐ ๐ฎ๐ฏ๐ผ๐๐ ๐๐ผ ๐บ๐ฒ๐ฎ๐๐๐ฟ๐ฒ ๐๐ ๐๐ต๐ฒ๐ฟ๐ฒ ๐ถ๐ ๐ฎ๐ฐ๐๐๐ฎ๐น๐น๐ ๐ต๐ฎ๐ฝ๐ฝ๐ฒ๐ป๐:
๐ช๐ฒ ๐ท๐๐๐ ๐๐ผ๐น๐ฑ ๐๐ต๐ฒ๐บ ๐๐ต๐ฎ๐ ๐๐ผ ๐ฎ๐๐ธ
The US Bureau of Labor Statistics is proposing something no statistical agency in the world has done: adding Artificial Intelligence questions to the American Time Use Survey (ATUS), linking AI use to a continuous 24-hour record of people's actual days (91 FR 42775, OMB 1220-NEW).
Adoption surveys tell us whether people have tried AI. A time diary tells us how much, for which tasks, and instead of what. For anyone managing AI risk inside an organisation, that second set of facts is the one that matters.
The Human Ai Instituteยฎ has today submitted a formal public comment applying Enterprise-wide AI Risk Managementยฎ (EW-AiRMยฎ) ewairm.com to the proposal, authored by our Founding Director, Prof. Markus Krebsz.
Our position is one of strong support, with ten specific recommendations. Four stand out:
๐น ๐ ๐ฒ๐ฎ๐๐๐ฟ๐ฒ ๐๐ต๐ฎ๐ฑ๐ผ๐ ๐๐. One question, asking whether the AI tool used for work was employer-approved or the respondent's own, would produce the first nationally representative estimate of unsanctioned workplace AI use. Every enterprise governance framework currently has to assume this number. BLS could let us know it.
๐น ๐ ๐ฒ๐ฎ๐๐๐ฟ๐ฒ ๐๐ฒ๐ฟ๐ถ๐ณ๐ถ๐ฐ๐ฎ๐๐ถ๐ผ๐ป. Did people check AI outputs before using them? Time spent verifying is the hidden denominator of every AI productivity claim, and no national survey anywhere measures it.
๐น ๐ฆ๐ฒ๐ฝ๐ฎ๐ฟ๐ฎ๐๐ฒ ๐ด๐ฒ๐ป๐ฒ๐ฟ๐ฎ๐๐ถ๐๐ฒ, ๐ฒ๐บ๐ฏ๐ฒ๐ฑ๐ฑ๐ฒ๐ฑ ๐ฎ๐ป๐ฑ ๐ฎ๐ด๐ฒ๐ป๐๐ถ๐ฐ ๐๐๐ฒ. The 2027 to 2028 fielding window of two years will span the mainstreaming of AI agents. A module designed around "chatting with an AI" will misdescribe 2028.
๐น ๐ ๐ฒ๐ฎ๐๐๐ฟ๐ฒ ๐๐ต๐ฒ ๐ฎ๐๐ฎ๐ฟ๐ฒ๐ป๐ฒ๐๐ ๐ด๐ฎ๐ฝ. AI is disappearing into ordinary software faster than respondents can recognise it. Instrument the drift; do not pretend it away.
Our submission also makes a point we believe applies to every AI collection now being designed worldwide: national statistics are the missing denominator of enterprise AI governance. Organisations, insurers and regulators are currently calibrating against vendor marketing surveys. BLS can replace that foundation with statistics.
As one of our EW-AiRMโข Governance Maxims puts it: "Governance that looks right is not the same as governance that works." The same discipline applies to measurement.
See our submission here:
https://enterprisewideairiskmanagement.grigora.app/usatus/