Share of work hours assisted by generative AI
Share of all US work hours in which employed adults used generative AI, from the Bick–Blandin–Deming Real-Time Population Survey (quarterly series via FRED).
Reason through thisThe reading
Broad but shallow: most workers have touched it, few of their hours run on it. The share using it weekly rose much faster than we expected, so this is the reading to watch.
8 readings drawn from fred.us_workers.hours_assisted_share.q; 8 observations in the evidence.
Source: FRED (Federal Reserve Economic Data). Bick, Blandin & Deming, Real-Time Population Survey, retrieved from FRED.
Why it matters
Saying you use AI is not the same as working with it. Adoption surveys count anyone who has tried it; this counts the share of working hours actually done with it, which is what any productivity gain has to be built from.
What cuts against it
Self-reported survey; the instrument changed between waves; hours assisted is not hours saved. One instrument (the Real-Time Population Survey) is published three ways (FRED, the authors' tracker, the FRED Blog), which count as one source, so the status stays emerging until a second survey measures hours.
How it is measured
- series
fred.us_workers.hours_assisted_share.q - source FRED (Federal Reserve Economic Data) · default tier 6 · FRED terms of use; RPS series by Bick, Blandin & Deming
- proxy behaviour, deployment · cadence quarterly · valve product to adoption
Normal = single-digit share of hours in year three (PCs and the internet were here at the same age); fast = more than 20% of hours assisted.
Applied to fred.us_workers.hours_assisted_share.q.
Latest readings
Why this status, and its history
6.3% of US work hours assisted (Q2 2026, FRED) sits inside the normal band, but every reading comes from one instrument, the Real-Time Population Survey, published three ways (FRED, the authors' tracker, the FRED Blog). Three publishers are one source, so the two-source rule caps the status at emerging until a second survey measures hours.
- 2026-09-11consistent with normal to emergingconf 80 → 80 · Claude (model)
6.3% of US work hours assisted (Q2 2026, FRED) sits inside the normal band, but every reading comes from one instrument, the Real-Time Population Survey, published three ways (FRED, the authors' tracker, the FRED Blog). Three publishers are one source, so the two-source rule caps the status at emerging until a second survey measures hours.
source1 - 2026-09-10unmeasured to consistent with normalconf — → 80 · initial seed (model-drafted)
Bick-Blandin-Deming RPS, May 2026 wave via FRED: 6.3% of work hours assisted by generative AI, inside the single-digit normal band. Weekly work use is 39.2% and any-use 61.8%, both well above the brief's expectations - broad adoption, shallow intensity. Initial seed.
source1
Why this confidence
Rests on 8 observations from FRED (Federal Reserve Economic Data); best evidence tier 6, grade B; current.
Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.
Predictions that lean on this
- HappeningAI's big economic effects arrive over decades, not years. · Narayanan & Kapoor
- Not happeningAI is not an ordinary technology · This site
- Too early to tellNo single leap in ability unlocks AI's economic value. · Narayanan & Kapoor
- Too early to tell
- Too early to tell